Something other than plain water is highly desirable if not essential, to prevent corrosion as well as freezing, and the usual product is a glycol-based fluid that has to be diluted with water. A 25% concentration protects down to -12C, and a 33% solution down to -18C, which is probably good enough under most circumstances. However if yours is a daily driver, then although it is unusual to go below -10C in central England, it did get down to -15C in December 2010, and has been as low as -27C (Newport, Shropshire, 1982).

If you are putting anti-freeze into a system that has previously contained plain water - or even if replacing old anti-freeze with fresh, there is a right way and a wrong way of doing it. DO NOT drain the system, and refill it with an already diluted mix, as many web sources advise. There will be a significant amount of water/old coolant left behind which will dilute your mix even further. Calculate how much anti-freeze is required to give the required dilution, add the anti-freeze neat, then top up with clean water, and you will get the correct mix. The total capacity of the MGB system according to Haynes is:

• Early models - 9.5 Imperial pints (5.4 Litres, 11.4 US pints)
• 18V engines, GHN5 and GHD5 from model No. 410002 - 11.5 Imperial pints (6.6 Litres, 13.8 US pints)
• 18V engines 1976 to 78 - 10 Imperial pints (5.6 Litres, 12 US pints)
• All models 1978 onwards - 12 Imperial pints (6.8 Litres, 14.4 US pints)
• All plus 0.5 Imperial pints (0.28 Litres, 0.6 US pints) where a heater is fitted.

Looking around on the internet you will see replacement intervals of 30-60k or 12 years or more quoted, however that will be for modern cars. The Workshop Manual recommends replacement at 2-yearly intervals, and checking the specific gravity at each service. There are testers available from the usual sources for a couple of pounds upwards.

There are alternatives such as 4-Life and Evans Waterless, which on the face of it seem to offer benefits. In the case of 4-Life it is said to indicate a head gasket leak by changing colour, but on a pals car with a cracked head and compression leaking into the coolant (detected by other means) there was no colour change. Evans Waterless amongst other things is said to reduce system pressures and can't boil, but you need to go through a multi-stage flushing process before using it to replace glycol-based fluids. Both can only be topped-up with the same stuff, which you have to carry around with you, and neither guarantee you won't lose fluid e.g. from hose failure, which given the quality of rubber these days is bound to happen at some time. Both are expensive compared to glycol based concentrates, and personally I can't see any good reason to use them, millions of cars have been running for decades with glycol-based perfectly happily. Neither the 4-cylinder even when run in desert states nor the V8 suffers from cooling system problems, unlike some other marques which are prone to problems even in the UK unless everything is perfect.

Coolant Temperatures:

    Temperature Gradients: Added September 2010
I took the following set of comparative temperatures on the roadster, warming up at a fast idle on a day when the ambient temperature was a cool 9.5C (all temperatures in degrees C):

Update July 2013 (phew, what a scorcher):
Quite a few people complain about 'overheating' and ask about installing uprated radiators, and electric fans to cars with mechanical fans. This is very rarely required, especially in the UK. As I say elsewhere these cars run in desert states in standard form without problems and have done so for many years. If your car is running hotter than it should then you need to be considering why, not throwing money and bits at the problem. Firstly, if a car isn't losing coolant, then it isn't overheating. Secondly, anywhere between the upper part of the C zone and the lower part of the H zone (on CNH gauges) is considered 'normal', depending on climate and usage. Consider a car in a temperate climate, pootling along a flat open road, and the temp gauge should be about the middle of it's travel i.e. near the N of CNH gauges.

That still leaves quite a lot in the block as it will only drain down to the lowest passage which may only be the bottom hose. On 4-cylinder engines this is enough and leaves the coolant level several inches below the top of the block. But it is a problem with V8s as the water pump passages are much higher. Blocks have a drain point as well - on 4-cylinder engines it is just below and to the right of the oil gauge port, originally another tap, subsequently replaced by a plug which is all I have ever seen. Someone wanting to fit a geared starter in place of an inertia unit to an early engine said it either fouled the chassis rail or the drain tap, I suggested he replace the tap with a plug. V8 engines have one each side, relatively central. On mine it is the later more functional tap on the near side, and a plug on the off-side - I discovered why they differ when installing a rebuilt engine to Vee. Having fitted a tap to the off-side as well as the near-side (as I had two available), I found I couldn't fit the exhaust manifold! Fortunately I had just enough room with the engine fitted to remove the tap. I didn't have a plug available as the original was damaged and I couldn't get it out of the old block, but I cut the 'arms' and threaded part off the base that screws into the block - after making sure it was fully tightened - and that sufficed. However, whether you can get anything out of any of the block drains, is another matter altogether. I couldn't from the 4-cylinder or the V8 and this seems very common. Poking in the hole with stiff wire does nothing, and those who have delved into a stripped block say the bottom of the water jacket is choked with something that needs chipping out, probably the original casting sand.

Cars with mechanical fan: I can't remember ever removing the bottom hose on Bee over the 25 years I've had her until I needed to replace the thermostat, and hadn't appreciated just how awkward it is to get it off the rad. You do not want to completely remove it while under the car as coolant goes everywhere, but you will have to undo the clamp and get the hose moving on the rad port at the very least. Then from above you will need to pull the hose upwards, which means folding it to some extent, and it will probably come off all of a sudden. It's easier to remove it from the pump first, then pull it up, but that means coolant going everywhere twice. With old hoses that have hardened you may have to do it this way anyway. However the bottom hose that came with Bee 25 years ago is still supple and crack-free, which is more than can be said for the other hoses that I've had to replace on both cars over the years.

Cars with electric fans including V8: Much easier as there is more space to get at the hose, and direct it into a container, it's also a straight pull backwards so no folding of the hose required.

When refitting bottom-hoses on the mechanical fan cars and the V8, great care needs to be taken with positioning, as described here.

  Refilling has its own considerations and pit-falls. When using glycol-based antifreeze that comes as a concentrate, to get the correct ratio you should add the required quantity of neat glycol to the drained system, then add water to fill the system. Unless the engine and heater are completely dry, as after a strip and rebuild, there will always be some old coolant left behind. If you simply refill with dilute anti-freeze, then whatever is remaining in the engine and heater will dilute it still further. The total capacity seems to have changed several times over the years, see here. How you get on with replacing glycol with ForLife I don't know, but I do know that with Evans waterless you have to go through several flushing cycles first.

For the first start after refilling you can get very different results depending on what thermostat you have, which unless you have changed it will be an unknown quantity. Unless you have a thermostat with one of the self-bleeding functions, the thermostat will trap a large volume of air under it, until it opens. When this happens the level in the radiator will drop like a stone, so you will have to leave the cap off watching the level, and have a kettle-full of hot water to hand to top-up. When I refilled with a non-bleeding thermostat I only got the required amount of concentrate in, i.e. just under 2 litres, and that brought the level up above the tubes. When the stat started to open and the level dropped I had to add over a litre of hot water to get it above the tubes again. If you have a self-bleeding stat you will be able to put at least 1 litre of cold water in as well as the concentrate to half fill the header tank before starting the engine.

Once the thermostat fully opens it may then drop more slowly as it purges itself, so you can trickle some more water it to keep the level visible. The heater valve should be open, and ideally the nose of the car higher than the rear to aid purging. Check the level again when cold and top-up with plain water, it may need a couple of heat/cool cycles to fully purge itself. After that, any topping-up required should be done with the required dilution, I keep a 2 litre container of ready mixed for this.

I have the mid-era top-fill rad so it is easy to see the coolant level, and top up while running. On the earlier rear-fill rads you can't see the tubes so will have to maintain the coolant to just above the bottom of the filler elbow. If done cold this can chuck some out, not so if checked warm, but be careful removing the rad cap if hot. Herb Adler added a catch bottle to his as described here. The remote expansion tank on later systems can't be used to fill the system. 4-cylinder cars have a fill port on top of the thermostat housing, and V8s have one on top of the radiator. But both of these only fill the radiator, there will still be a large amount of air trapped below the stat if that isn't self-bleeding. The V8 fill port can be used to observe the level, and the sudden drop and consequent topping-up with hot water, but I suspect the 4-cylinder thermostat housing fill port could well overflow with hot coolant from pump pressure and flow as the thermostat opens.

My V8 cooling was always a bit marginal in that the temp gauge was higher in hot weather than I would have expected. And although I have a high-efficiency rad (replaced the old one when it had successive leaks, looks the same but has 25% more tubes) which improved matters slightly I was on the lookout for a bit more.

  I also decided to investigate the voltages in the fan circuit as they draw a high current and any resistance in wiring and connectors causes a significant volt-drop. I was quite surprised to find I was losing 0.3 volts in various connectors in the brown circuit, 0.8 volts in the relay, and 0.7 volts in the earths - nearly two volts altogether. A PO had added some crimped connectors in the brown circuit and omitted to solder them so I rectified that, the relay was also getting very hot in use so I replaced it, and I added some earths from the fan connectors to terminals bolted between the fan brackets and the bonnet slam panel. Got rid of most of the volt drops (some in the wiring is inevitable) and the fans are now audibly faster and cool quicker.

I have also been looking at the possibilities of fitting some additional puller fans - they would have to be very slim and of a small diameter to clear the water pump. However, after having seen some fans of this type used in pusher mode I am very unimpressed. Despite having a shroud I don't think they are as effective as the factory fans, which really is saying something. The most noticeable difference is just how long the fans continue to spin after the power is switched off, which says to me they aren't pumping very much air, backed up by feeling how much (little) air is pumped through the rads when they are running. They have 'spiral' blades, which one fan manufacturer claims they tried, and promptly dumped them. At a minimum of £60 each I think I'll give them a miss and look further at fitting shrouds to the existing fans.

The latest enhancement was to run a heavy-gauge brown wire from the unused output spade on my alternator direct to the fan relay, which happened to have a second spade on that terminal. This has made another notable improvement in fan speed and rate of cooling, subsequently I shut it in the garage (exhaust piped outside) on a 30 degree day and left it fast-idling to see what happened. The air going into the grill was being recirculated from the engine bay rather than being 'fresh' air at ambient and actually got up to 41 degrees, but still the temp gauge never got more than two-thirds the way from N to H. Travelling through France to LeMans and back in 2002, which was very warm, I was able to keep the needle on Normal at all times even in the hottest conditions. Auxiliary fans no longer required.

Fan fusing October 2017
The wiring diagrams show 4-cylinder cars electric fans having their own in-line fuse under the fusebox - white/brown to green, one of two for RHD cars from 1978 on. However somewhere I've seen reference to a thermal cut-out rather than a fuse for North America, and this eBay offering from Oz shows just such a device with stud connections. The part number quoted is of the form typically used by Moss US, but Googling that throws up no other references. However Googling variations of the description came up with this forum post that shows two cars with the thermal device, albeit with spade connections.

  V8s originally used the main green fuse in the fusebox. There is a fan relay, but it is a three-terminal device which results in both relay and fan current coming from the main green circuit. But so did the HRW current and all the ignition powered stuff, which meant that without an ignition relay you could easily have more than 20 amps flowing through the ignition switch. Vee, and I have seen it on another example so it may have been a retrospective factory mod, has a four-terminal relay with the fan supply coming from a tee off the brown circuit at the fusebox. So the fan load is not through the ignition switch, but it's not fused either.

I wasn't happy with that, especially after I boosted the current-carrying capacity of the circuit which significantly boosted fan performance, so I added an in-line fuse, with male and female spades on the fuseholder wires to make it 'plug compatible' with the existing wiring and relay and easily reversible. Originally a 'recycled' blade-type fuseholder with rubber cover I happened to have in my stock of bits with a 15A fuse (Halfords only had 6A in standard glass-type available). But with contacts that were less than perfect I discovered that the plastic body of the fuse had melted in normal use, so 18 months ago I got some 20A in-line fuse holders with fuses and wired one of those in.

That showed no heat damage after several months, but shortly after her restoration, in a busy car-park looking for a space I noticed the fan tell-tale was on, but I couldn't hear the fans. Fortunately I soon found a space, and could get the bonnet up - to find a melted fuse-holder! Unplugged it and tapped the wire directly onto the relay, and one fan burst into life but the other didn't, which made me think it had seized - which happened to the other fan many years ago. However when I removed the wire I noticed the non-moving fan moved slightly, and looking closer saw that my emergency bonnet release cord had somehow got onto the radiator side of one of the blades and was stopping it spinning. Moved that out of the way, reconnected the fan wire, and both fans burst into life, with the stalled one apparently working normally. That started me thinking. I've had this cord in place for years, and this hasn't happened before, and I couldn't really see that the action of the fan could suck the cord over the blade anyway. The only alternative is that I inadvertently ended up with it on the wrong side when I refitted it. But that would mean that fan had been stalled for 500 miles of running-in, over a dozen or so trips, with plenty of fan operation in traffic, and especially when I left it idling for quite a long time after first replacing running-in oil with 20W/50 to check the effect on oil pressure, and I'm sure I would have noticed if only one of the fans was running then. So, a mystery that will probably never be solved.

It all started me thinking about current, and whether I should fit a fuse that would blow in the event of a stalled fan. Measured the current with both fans running with a stopped engine (12v) and it was very nearly 10 amps. Stalled one of them and it went way over 10 amps, but as my meter only reads to 10 amps I was none the wiser. So took the grille off and unplugged one of the fans, and measured the other at 5.5 amps running (that's higher than half the current from two fans, but is down to how inevitable connection and wiring resistances interact with voltage, current and heat, and especially how electric motor current varies with voltage and speed of rotation, all too complex to go into here). Checked the current with the remaining fan stalled, and it was still over 10 amps. Measured the resistance at 0.8 ohms for each motor, which means a stalled motor will take 15 amps at 12v and 18 amps at a charging voltage of 14.5v! Add to that more than 5 amps from the remaining motor still running, and you get well over 20 amps i.e. more than the rating of the fuse and the fuseholder. That's theoretical, as again we are into the realm of how current, heat, and resistance interact and the greater the current you try to draw the more the resistance goes up which tends to limit the current. Looking at the fuse holder the damage has occurred where the terminals on the ends of the wires were in contact with the ends of the fuse. Which shows they were the point of greatest resistance, and hence where volt-drops were occurring, and consequently heat. Both fuse and holder were 'new', so should have been clean. The wires themselves were undamaged, as was the fuse, albeit with the strip discoloured so close to melting, until the pressure of the internal spring pushed the ends of the softened fuseholder apart, which broke the contact. Now I don't know what the 'blow' rating of these fuses is, but going by the OE fuses which are 17A rated 35A blow, you could be talking close to 40 amps for a 20 amp fuse, so both fans would have to stall to reach that, which hopefully would never happen. I could use a 10 amp rated, 20 amp blow fuse, but that could be marginal for both the normal 10 amp or so running current, as well as for 20 amps or so for one fan stalled. It would also cut power to both fans. I could fuse the fans individually at 5 amps each, but that is still going to be marginal for both running and stalled current, and would mean messing about with the wiring in order to have the fuses behind the radiator panel to avoid the worst of the weather.

All-in-all, I think I will settle for simply replacing this 20 amp fuse and holder, and making sure the bonnet release cord can't get tangled up in the fan. The old fuse acted as a 'thermal protection device' by disconnecting the power, even though it was sacrificed in the process. So for what should be a very rare occurrence I'll accept the possibility that it might happen again. If it does, it's only a couple of minutes to remove the grille and disconnect a seized fan, then bypass the fuse as before, and one fan will be enough for anything but the hottest weather.

V8 fan cycling September 2017 Following Vee's body restoration and engine rebuild one interesting change is how consistent the fan cycling has become. For 24 years - with very few exceptions - it hasn't cut in until about 4 o'clock on the temp gauge (higher still with the original Otter switch), staying on until it gets back down to N i.e. very long on and off times. On rare exceptions it would cut in at barely 5 o'clock, and cut out again barely any lower, i.e. as per a modern car. Once it started doing that it would continue, but once underway again, or switching off and a restart, it would be back to the usual cycling with very long on and off times. Now, every time, it's been cutting-in at about 5:30, and out again without the gauge noticeably moving. Timing it - out of interest - following the first oil change, after it had done a dozen cycles or more it was off for 25 seconds every time, with the on times slowly increasing 40, 45, 50 seconds, without the gauge noticeably moving. Why this change in behaviour? It's the same inlet manifold and the switch wasn't touched during the rebuild; that manifold had been on and off several times before; and the coolant drained and refilled several times. Before I did occasionally have to top-up the expansion tank, but since filling with anti-freeze and allowing a couple of heat/cool cycles to fully purge, the level doesn't seem to be dropping over the 500 miles. Maybe a tiny leak was allowing air to get round the switch, I doubt we shall ever know.

Added October 2008:

 Update October 2010

Update May 2016
After the MOT I noticed how much the cooling fans seemed to affect the cranking speed on a hot start, and thought they would be a far more worthy candidate for being controlled by the 'accessories' circuit which is cut off during cranking, than the washers, wipers and heater fan. So gazing at the fan relay and fuse and working out what would be required, I noticed the rubber cap had been pushed off the blade-type fuse holder I'd added some time ago as the car came to me with unfused fans, because the fuse had melted.

The accessories fuse is literally right next to the fan relay, but no convenient connectors so I had to cut the output wire from the fuse (green/pink) and solder on two bullets, and fit a 4-way connector between them. Then about 4" of wire was all that was needed to go from there to the relay, in place of the green (fused ignition) wire. This wire only operates the 4-terminal relay that I have, it doesn't power the fans, so the current is negligible. Turned on the ignition to the accessories position, and I was surprised to hear the relay click and get a burst of spinning from the fans, and the tell-tale in the switch was glowing slightly. I realised that the tell-tale also needs a voltage supply, which I had taken from a fused ignition source. So effectively what was happening now was that in the accessories position 12v was being fed through the relay back to the override switch, through the tell-tale, onto the green circuit i.e. the fused ignition supply. The standing loads on that such as ignition and fuel pump allowed enough current to flow to briefly operate the relay and cause the lamp to glow, but not enough to power the ignition circuits. It all worked correctly with the ignition on, but I didn't want to leave it like that. I could have fitted a diode at the relay to block the reverse current, but opted to change the wiring at the override switch to pick up accessories power from the heater switch which is right next to it, instead of fused ignition. The irony was that when I removed the (POs) wire that had been feeding fused ignition to the override switch to power the tell-tale, I found it went all the way to the fusebox, and was long enough to have reached the fan relay. So I could have reused that to feed accessories power from the switch to the relay, instead of cutting in to the accessories wire by the fuse! Oh well, such is life.

Subsequently I realised I could have done it without any cutting, which I normally try to do. If you get one of these fuses with a loop of wire between the ends and not the pre-cut type, then you can fit it between the two ends of the existing fuse-holder, i.e. effectively have two fuse-holders in parallel. You can have a fuse in both, or put a metal bar the same size as a fuse in the new holder. The power comes through the old fuse-holder first on white/green, so that should still have a fuse. Then all you need to do is tee another wire (soldered and heat-shrunk please, not Scotchlok) into the new wire between the fuse-holders, and that is your fused accessories supply to the fan relay once you have removed the original green supply from it (insulate it to stop it shorting on anything). To reverse, simply remove the new fuse and its tee, connect the two halves of the original fuse-holder together, and reconnect the green supply to the relay.

 Adjustment: At the moment I'm going to restrict this to the main water pump and alternator drive belt, and not get into the additional belts used on North American spec cars with air-pumps let alone air-conditioning.

For years I've checked my belt tension by seeing how far the middle of the longest run can be deflected from its 'at rest' position to or from the other side of the loop with light-ish finger pressure, and set it to about 1/4"-1/2". I've never had belt squealing in 40 years, and only had to replace a water pump once each on two cars quite soon after coming to me after many miles with other owners. Some manufacturers are a little more precise and state a deflection with a particular force, and others require two marks to be made on the belt a certain distance with no tension, and the alternator adjusted to cause those marks to move apart to another certain distance, i.e. the belt is stretched in use. The Leyland Workshop Manual in the Cooling section states for both dynamo and alternator systems it should be possible to move the belt laterally in the middle of its longest run by no less than one inch! Is that in one direction which seems a helluva lot? Or a total in both directions i.e. 1/2" each way which seems more reasonable? The Electrical section just says to "remove any undue slackness". Haynes in its cooling section says there should be a sideways deflection in the middle of the longest run of 1/2" on 'early' models, and 1/4" under an applied load of 7.5 to 8.2 lbf. Both 'lateral' and 'sideways' imply to me to be pushing or pulling it towards the front or rear of the car, and not towards or away from the other side of the loop which is always the way I have done it. Also is it in one direction or both? It seems likely that when using a spring-balance to apply the stated force for later models it would be in one direction only. The distance for that is half that for 'early' models, so do the early ones have twice the play? Or is theirs in terms of total deflection in both directions? And for that matter what is the definition of 'early' and 'late'? (Dynamo vs Alternator?). FWIW my 1967 Mini Workshop Manual also quotes an inch in the longest run, but an 80s Metro Manual quotes a much smaller deflection of 3/16" and in the shortest run i.e. between alternator and water pump pulleys. It also shows it being pressed down towards the crank pulley i.e. not sideways, and only in one direction from rest. Interestingly it also quotes a torque wrench value of 11 to 11.5 lbf when applied to the alternator pulley nut to produce slip, which is perhaps the most useful and relevant approach.

 Size Considerations Added September 2010: I've been asked a question about fanbelts widths for different alternator types, in particular how those for the higher output Lucas 20ACR and 25ACR compare to the 18ACR, so did a little research.

It will depend on the pulley. As the same alts definitely have different diameter pulleys for different applications in theory you should be able to swap them round i.e. between different 18ACRs from different cars. That implies a common size of spindle, but whether this extends to different models of alternator I don't know. Different makes of alternator do seem do have different spindle diameters, one site claims 15mm for Lucas and 18mm for Bosch. My roadster has an alternator with an alloy end plate and not plastic like the originals, I don't actually know what model it is, but the standard belt fits fine. I have a Lucas A115-45 from a wrecked 80s Metro which I have trial fitted to the roadster. The pulley is a different diameter (which gives a lower output at idle but is fine when driving) but the same belt fits just fine, so I keep it as an emergency spare. With a V-belt and pulley as long as the belt isn't sitting on the bottom of the vee (and hence no pressure on the sides) but is being wedged into the vee of the pulley, it shouldn't really matter unless the belt is so wide that the top of the belt is above the top of the vee and hence only part of the vee of the belt is engaging with the vee of the pulley.

The clue may be in the belt number. I know many indicate the length, and Goodyear for example say the first two digits of their belt numbers indicates the width as well as the last four indicating the length.

This site gives codes for widths, and the examples in the image could well contain width as well as length info printed on them like the Goodyear.

This Triumph (I know, "wash my mouth out") site says 15ACR and 18ACR were used with no A/C, or 20ACR and 25ACR with, you would have to compare pulley and belt numbers. But one thing occurs to me concerning A/C and higher-rated alternators, and that is that the more current any alternator delivers, the harder it is to turn, which puts more load on the fan-belt. It could be that higher current alternators do have a bigger (wider? deeper?) belt because of this. It would only be an issue if you were going to use the extra current, but then a bigger belt would impact on the water pump and crank pulleys as well!

So what size belt is required for standard engines (i.e. standard crank and water pump pulleys and standard dynamo/alternator) on 4-cylinder engines? The Parts Catalogue quotes:

Modern equivalents are the GCB10900 and GCB10950 which are direct replacements for those in the above table, the first two digits signifying the width (10mm) and the last three the length (900mm and 950mm respectively). You may well get slightly longer ones to fit, or maybe even slightly shorter ones, it depends what pulleys have been used, which may not be original. A tip on changing is to get the belt over the crank and alt pulleys first as they wrap furthest round these, and the water pump last - having got it over the mechanical fan blades first of course! If the belt is still slightly too short then try removing the adjuster bolt altogether (while fitting the belt!) which may get the alternator that bit closer to the block.

For the V8 the original was specified as GFB 148, recoded as GCB 11125 i.e. 10mm wide (although shouldn't that be GCB 101125?) and 1125mm long. However some have reported problems with that belt too long on both conversions and factory cars, and mine is right at the end of the adjuster but could still do with a little more tension. A shorter belt would be preferable, GCB 11088 has been mentioned i.e. 1088mm, but then the problem is then getting the alternator down far enough to get the belt on as it hits the rocker cover way before the inner end of the adjuster is reached. Maybe something between the two would be better, e.g. GCB 11100 (1100mm), although with that someone has reported the adjuster bracket fouling the alternator fan! To repeat the tip above, try fitting a short belt over the crank and alternator pulleys first, and the water pump last.

In any event any belt you buy as a spare should be trial fitted when you get it, it should be kept in the car and not the garage(!), along with the tools to change it, and you always should be carrying a spare, even if only by the Law of Sod to guarantee you will never need it!

Both Bee and Vee came to me with old hoses of course. Bee's top hose showed slight signs of surface cracking, so I replaced it as a precaution, only for the new one to look the same after a couple of years. Replaced that with a Kevlar, and that does seem to be lasting better, although contrary to what I originally thought it isn't a Kevlar 'sleeve' embedded in the rubber like the earlier fabric reinforced hoses have, but apparently just chips of Kevlar embedded in the rubber. They may resist some abrasion in the short-term, but it is the quality of the rubber that will determine longevity. By contrast Bee's bottom hose remains unchanged (until May 2014 as part of a cooling system problem) and still pliable and crack-free, after 20 years and 50k miles just in my ownership - "They don't make them like they used to".

  When refitting hoses on the mechanical fan cars and the V8, great care needs to be taken with positioning. On the 4-cylinder there is only about 1/4" of clearance either side of the hose between the timing cover and the fan blades. After pushing the hose on make absolutely sure the blades are well clear, and allow room for hose expansion under pressure. However putting it too close to the timing cover can result in rubbing from that. Electric fan 4-cylinder cars should have no problems, but the V8 has this and additional issues as below.

  V8 Bottom Hose: Travelling at speed on the motorway in Vee one day I became aware of a slight misfire when accelerating. It was raining so could have been road spray affecting the HT leads, but it had never happened before. The thought immediately occurred to me that it could be a coolant leak spraying onto the HT leads. If so I'd be in illustrious company - I had recently read about an MG record attempt where exactly that happened, although in that case it was very high cylinder pressures lifting the head and causing coolant to spurt out past the gasket! All the gauges were normal, and as I had just passed a service area I carried on keeping a close eye on things. At the next service area I pulled off and as I slowed to parking speeds could hear Vee hissing like a steam engine! Stopped and opened the bonnet to see steam issuing from what looked like a pin-hole in the bottom hose right by the fan-belt, together with what looked like a series of cuts in the same area. Wasn't carrying a spare hose (then, I do now!), so it was a job for the AA. They came in due course, no replacement hose of course, so it was a repair jobby with a special kit (I now carry one of those as well!). Very little coolant came out when the hose was removed, yet the temp and oil gauges had stayed on normal all the time. The repair got me up to Liverpool and back via Clive Wheatley for a replacement hose, but was beginning to weep the next day, when I replaced the hose.

Added August 2009:

  July 2014: Noticed a spot of coolant on the floor where Vee had been parked - bottom hose dripping. It was coming from by the flange on the oil cooler hose, which I mention above a couple of times. It was touching the flange, which had worn a groove, despite me positioning it with a small clearance when I last replaced it, which was only 5 years and 12k ago. Also signs of surface cracking, again at the bottom, and when freeing it from the water pump inlet it split very easily. All of which confirms an opinion that the rubber in these 'Kevlar' hoses is still crap. So I fitted the hose bought in 2009, which again needed some cut off the rad end.

  V8 Top Hose June 2013

Fans are usually 'handed' i.e. they have one side that should face the engine and a different side facing the radiator. Reversing it does not change the direction of air flow (only changing the direction of rotation can do that), but it does have an effect on efficiency i.e. how much air it moves at a given speed. Both metal and plastic types seem to have sculpted blades where one side is concave (i.e. slightly cup-shaped) and the other convex. The concave side should face the engine as it is more effective at pushing air towards the engine i.e. pulling it through the radiator. Metal types also seem to have one gently rounded corner and one tightly rounded on each blade. The gently rounded corner should be on the leading edge, probably for safety reasons if one gets one's fingers in the way. The individual blades of metal fans are attached to the stubs of the central boss on the engine side, with rounded rivets, both of which reduce turbulence and improve efficiency. Plastic fans often have the blades with an aircraft wing profile, with the blunter edge leading and the more finely tapered trailing.

Update October 2007

Update January 2010

July 2014: Roger Parker covers the same topic in 'Enjoying MG' this month. However the measurements he gives are significantly different to the three above. He measures to the edge of the header tank, which is fair enough at that is closer to the blades than the core. He says that on several cars he measured it ranged from 3/8" to 1 1/4", but the average clearance on cars with the 'correct' spacing was 1/2"! Even subtracting 3/8" - which is the distance from the edge of the header tank to the face of the core - from our measurements that gives 7/8" for Steve's and David's, quite a bit more than Roger's 1/2", and 1 3/8" for mine which is nearly 3 times as much.

November 2014:

Water pumps are one of those things one changes when they go wrong, and generally what goes wrong is the seals wear and they leak, or the bearings wear which causes noise, wobble and also leaks. I have also heard of the impeller coming off the shaft, which will cause overheating of course. Bee's pump started leaking shortly after I bought her. At that time I wasn't aware of the variations between engines, but the proprietor of the MG parts supplier had a look and got the one that I needed, with a new gasket - or so I thought! When I came to fit it I found it was alloy instead of steel (probably a copy as I know now), one of the bolt holes was about half a hole out even though the gasket was correct (but being alloy was easy enough to file into an oval), and another had a thick boss (now known to have been an early GWP117 pump which had two long and two short bolts whereas the one coming off was the later GWP130 with three short and one long) which meant the bolt wouldn't engage with the threads. Fortunately also easily correctable. To add to all that shortly after fitting I noticed it had started weeping slightly during warm up! So I bought another, steel this time, but as it happens in all the miles and years since then the leak hasn't got any worse and water loss is negligible and I still have the steel one in the boot ... (until June 2014, that is).

I've heard it said that there are at least two pumps with different impellers, both fit at least one of the engines, but the incorrect combination results in a low coolant flow rate and overheating. If that's true it can only be the pumps for the 3-bearing and 5-bearing engines that are incompatible, as all the other pump changes occurred during productions runs of the various five-bearing pumps and so must be interchangeable - with a caveat.

The caveat is that the pumps changed from having a long nose to a short nose in August 71 for the 18V engines. Whilst the pumps themselves are interchangeable it has a knock-on effect on the pulleys and fan blades that can be used. Long-nosed pumps use a deeply dished pulley, so of course a short-nosed pump will move the V-groove much closer to the engine and way out of line with the crank and dynamo/alternator pulleys. The fan bolts up to the pulley so that is also moved closer to the engine and in fact hits it. Short-nosed pumps used a shallower pulley to keep the correct V-groove alignment. The fan blades would still hit the engine of course, so a 7/8" spacer was fitted between the pulley and the fan. You can get away without the spacer if you have a 3-blade metal fan by turning the fan round the wrong way, because the blades are offset to the centre this actually puts the blades at almost exactly the right position between engine and radiator. They are turning the wrong way of course, but the difference in effectiveness is marginal, I ran Bee like this for many years before discovering it, and obtaining and fitting the spacer. In theory you can also use a deep pulley with the short-nosed pump, by putting the spacer between pump and pulley instead of pulley and fan, but I've not tried it. You need longer bolts when using the spacer in any configuration, of course.

Culled from various sources this is what I have managed to work out for the many combinations of engine, market, period, pump, fan, and the various bits and bobs to put them all together:

Bolts - important! GWP117 originally took two short screws HZS 511 1 3/8" long and two and two long screws HBZ515 1 7/8" long. But that pump was superseded by GWP 130 which took three short screws HZS 508 1" long and one long screw HBZ 514 1 3/4" long. They in turn were superseded by three short screws SH 605081 and one long bolt BH 605141 of the same lengths. However the 1" and 1 3/4" screws that came with my gold-seal engine and GWP130 pump only go into the block about 1/4" - which doesn't seem enough to my mind, and the drilling is nearly an inch deep. The supposedly 1" screws only have about 7/8" of thread, not 1". The MGOC information (now corrected) said the GWP 117 pump has been replaced by the GWP 130 which is correct, however it went on to talk about two screws SH 605111 and two bolts BH 605151. But these are the items and quantities for the earlier GWP117 pump not the GWP130 in question! Not realising at the time they were for the old pump I ordered them, but they are too long for the GWP130 pump which has shallower bosses. I had to cut about 5/16" off the short screws, and 1/8" off the long bolt, or they bottomed in the block before fully clamping the pump. The short screws were obviously too way long, but the long bolt was really only discovered by offering the pump up WITHOUT its gasket, loosely fitting all four, then one at a time nipping them up and checking that it clamped the pump before it bottomed. The long bolt left the pump only slightly loose, which wouldn't have been noticed with the gasket fitted, and I could have ended up tightening the bolt into the bottom of the hole, either risking breaking through the casting into the water space, or leaving the pump not properly clamped. So, to reiterate: If you are fitting a GWP130 pump you will need three screws GHF103 or SH 605081 (1") and one bolt BH 605141 (1 3/4"). If replacing a GWP117 pump with a GWP130 and are using your existing screws and bolts, they will need to be cut down. Test fit without a gasket and confirm they clamp the pump before they bottom, and I used washers under the heads to be sure. Mine ended up with usefully more thread engaged, but without any risk of bottoming.

Note 1: Replaced by and compatible with GWP 114
Note 2: Clausager indicates a further change in Jun 66 but this doesn't appear in the Parts Catalogue
Note 3: GWP114 is interchangeable with GWP117 (short) given the shallower pulley and fan spacer
Note 4: Replaced by GWP130 so compatible with GWP117
Note 5: GWP 117 and GWP 123 must also be interchangeable
Note 6: Clausager indicates a change in water pump in Dec 75, not shown in Parts Catalogue. Possibly the earlier GWP 123 was changed for the GWP130 which would make these two pumps interchangeable as well
Note 7: Moss Europe indicates there are steel and alloy versions of the GWP 130 fitted to the 18V672 and changing at engine number 27269 in 1974.

An alloy pump I bought many years ago was a very poor fit, I had to file out one of the mounting holes as it was about half a hole out although the gasket supplied with it was fine. I also remember having to grind down a boss on one corner so one of the original bolts would fit. Given the bolt changes above, I suspect it came to me with a GWP130, but I was supplied a GWP117. That also started losing a drop of water while warming up shortly after fitting so I bought a steel pump but didn't fit it immediately waiting for the alloy one to get worse - and it never has so the steel one has been sitting in the boot for a dozen years or more! 17 years in fact, as in June 2014 as part of a cooling system problem I chose to fit it.

Another pump replacement: April 2018. Just after refitting the head after conversion to unleaded I noticed little spots of water on the front of the rocker cover and carb piston covers after a run. Hoses tight, temp sender nut possibly not quite as tight as it should be so corrected, but no change. Then when doing yet another compression check the fan belt seemed damp ... and the crank pulley ... and peering across the front of the engine from above the alternator I espied two lines of dots up the bottom hose in line with the pump pulley flanges, and more (I hadn't noticed before) up the offside inner wing! Now there is also water on the floor, so waggled the cooling fan and the pump shaft waggled with it - buggah! For a moment I pondered putting back the old alloy pump (which shows no play), but only for a moment as having been dry for nearly four years and weeping anyway who knows how bad it would be. So new pump it is.

Originally GWP117, suppliers show it having been replaced by GWP130, and at first glance information on the MGOC page implies it takes two short and two long bolts and these should be replaced as well. However further investigation shows that those are for the original pump, and the replacement pump actually takes three short and one long of different lengths! At least that explains part of the problem I had when fitting the alloy pump way back which had obviously been a 117 replacing the 'original' Gold-Seal replacement GWP130. But by this time I've ordered the pump, and new bolts as I'm not sure of the condition of the old ones, so I will have one that is too long (at least that can be cut down). An hour sees the system drained down to the bottom of the hose outlet on the pump, pump off, with more coming out down to the bottom of the pump cavity as those bolts are slackened. Looking at the bolts they only engage with the block by about 1/4", which doesn't seem much to me, whereas there is at least an inch depth available. The Parts Catalogue shows three SH605081 at 1" and one BH605141 at 1.75", but the MGOC page shows SH605111 at 1.375" and BH605151 at 1.875" which are just that little bit longer, so preferable ... or so I thought.

Package arrives next day ... but have you tried to find two small polythene packs of shiny bolts in a box full of white shredded paper? I rooted around in the box and eventually found the large ones by touch, but not the small ones. Emptied it out into a very large polythene bag, more rooting around, still nothing. It wasn't until I sprinkled the shreddings back into the box one handful at a time that I found them! First realisation is that I can't cut one of the long ones down as it has a plain shank that is longer than the three short bosses on the pump. Second realisation is that the 'short' screws are way too long as described above, and it was only by careful checking I discovered the long one was too long as well, so three new ones had to be cut down. For the third short screw I cut down the original long one. After all that messing around it didn't take much more than an hour to get the new pump on - remember to smear sealant round the outside of the bypass hole through the gasket, as well as inside the four holes for the holts. The only awkward thing was having to remove the front air cleaner as although I could get the bottom hose on without doing that, I couldn't get enough purchase to twist it into the right position to give maximum clearance to the fan blades while still giving some clearance to the front cover. It's all too easy to fit the hose, fire up, and find the fan blades hitting it!

Pump replacement: June 2014 Needed as part of investigations into a cooling system problem. First step is to drain the system and remove the hoses, which isn't as easy as on the V8! You can try undoing the block drain tap, but these are usually choked with casting sand and never work. Without a rad drain tap I have found the easiest way is to remove the clamp and wedge a medium flat-blade screwdriver (ideally an old one without sharp edges) in. This allows coolant to trickle out - guided down past crossmember and steering rack by something like a large padded envelope which can be wedged in - while you get on with the next stage.

Slacken the alternator adjuster nut on the crankcase pin, remove the two upper bolts, and pull the alternator back clear of the mounting ear on the pump.

Check the gasket fits the holes, and especially that the bypass hole between pump and crankcase is clear, and that the gasket also fits the pump, i.e. the pump holes will be in-line with the crankcase holes!

Refit the radiator.

Refit the hoses taking particular care that the bottom hose has clearance to the fan blades. For the top hose slide it onto the rad port as far as possible, then fit it onto the thermostat housing, then slide the hose towards the housing so that there is a relatively equal amount of hose on each port. If using new hoses Vaseline helps get them onto the ports, as they are a pretty snug fit when new.

Then refill the system. Probably sensible to use plain water if you have been fixing a problem, to check it is fixed, before refilling with anti-freeze.

  To refit slot the rad into position with the packing strips, and get the upper bolts started. As they come through the nut on the rad flange, position the wing strut over the thread. If you wait until the bolt is fully tightened you may not be able to get it on the bolt without removing the strut from the wing bolt. Once the top bolts are started fit the other bolts and washers before fully tightening anything. Finally tighten the nuts on the struts under the wing bolts.

If you find that your system chucks some coolant out shortly after switch-off, but the temperature of the coolant is 'normal' then your cap is probably faulty and not holding the pressure it should.

What might be less well known as that as well as the main pressure seal there are two other seals or valves in the MGB cap:

Under extreme conditions as mentioned earlier there can be coolant vented out of the overflow pipe, which is normally lost on the ground. Of course one could put a catch-bottle under the overflow pipe to catch it, then pour the expelled coolant back into the radiator when the system has cooled down again. But the two additional seals or valves in the radiator cap working together can cause this to happen automatically. If the catch-bottle is positioned such that the bottom of the overflow pipe reaches the bottom of the bottle, then as the system cools and draws air in through the low-pressure valve, the upper seal will ensure that this suction is applied to the overflow pipe, and hence will suck up any coolant the bottom of the overflow pipe is in contact with, and direct it back into the radiator automatically - a lost-coolant recovery system! This is an unpressurised catch-bottle, which must be open to atmospheric pressure, completely different to the remote pressurised reservoir (see below) used on MGCs, V8s, and 77 and later MGBs.

The above seems to be the case on all UK caps I have come across. But if either the upper seal is missing or faulty, or the low-pressure valve isn't present, this coolant recovery won't work. People in the USA often mention caps available there can't always do this, indicating they are to a different design rather than being faulty. One source indicates these caps have the upper seal to prevent scalding, but not the low-pressure valve. Without this valve, not only can lost-coolant recovery not be employed, but the cooling system can experience negative pressure on cooling down, which could cause boiling at the next run. Another source indicates that the low pressure valve is present to prevent the partial vacuum in the cooling system, but the upper seal is missing so coolant recovery cannot take place. Note that only under extreme conditions should lost-coolant recovery be required, if it is happening at lower temperatures and lower loads as my V8 was doing, there is a fault which needs to be investigated and fixed.

If you overfill any MGB cooling system when cold it will expel the excess when it warms up. The correct level is about 1" above the tops of the tubes when cold in the Mk2 radiator to 76. The V8, 77 and later MGB and MGC remote reservoirs should be kept about half full when cold. The Mk1 radiator with the right-angle filler is the most awkward as you cannot see the tops of the tubes. You must keep coolant just visible from the filler neck or it could have dropped way down, but if you keep it too high in the filler neck it will keep expelling the excess. Probably best to check this when still warm (but not hot!) and if you can see coolant it should be fine, even if it drops out of sight when fully cold.

 Remote reservoirs were necessary to allow the installation of a lower-profile radiator mounted further forward, without it the V8 and later MGBs would have needed a bonnet bulge, and the MGC would have needed a much bigger one. These remote reservoirs are pressurised, but the cap, overflow and potential for lost-coolant recovery are exactly the same as for the earlier MGBs. The pipe between the radiator and the remote reservoir is at the top of the radiator but the bottom of the reservoir, which ensures that any air in the radiator is pushed through into the reservoir as the system heats up, to be replaced by coolant as it cools down again. So even though the reservoir and radiator cap are below the top of the engine and radiator, the system is self-purging, getting rid of any air e.g. after a coolant drain and refill, over no more than a couple of heat-cool cycles.

Mk1 MGBs used a 7lb cap and radiator with rear-mounted right-angle filler. However it is important to note that this is GRC102, and not the longer GRC101 which some suppliers state (see below).
Mk2 cars up to and including the 75 model year used a 10lb cap and radiator with top-mounted straight filler.
The 76 model year used a revised radiator (visually similar and exchangeable with the previous) and 13lb cap.
For the 77 model year and onwards, and all V8s, the forward mounted, lower-profile radiator with remote reservoir was used with a 15lb cap. 4-cylinder cars have a brass or plastic filler plug on top of the thermostat housing, V8s have it on the radiator itself, note a plug-spanner fits this.
The MGC was similar to this last MGB style but had two filler caps - one on the thermostat housing without a main pressure relief valve, and another on the remote reservoir rated at 10lb for the UK and 14lb for the USA. There was no filler plug on the radiator.

  July 2014: A couple of comments on the MGOC Bulletin Board about incorrect radiator caps being received, from MG parts suppliers. The caps were 22mm from seal to seal, but the filler neck needed a cap that was 26mm from seal face to seal face. I felt whilst one was a possible error by the supplier, two was too much of a coincidence (unless from the same supplier). So I measured mine and found them to be 20mm for both the cap and the filler neck. The inference being, that somehow these two people have incorrect filler necks on their radiators. You could say that mine might be incorrect at 20mm instead of 26mm, but having bought caps for both cars at different times from different MG parts suppliers, and they fit, and a pal recently having done the same, I doubt it. Then someone else posted that the Mk1 rear-fill radiators do have a deeper neck. But the Parts Catalogue states that the cap for those rads is GRC102, and any number of sites indicate that has the 'shorter' 3/4" or 19mm neck. Then someone else posted that he was getting the wrong caps from the MGOC, sent them a picture of his rad, and they immediately spotted it was a rear-fill rad in a MK2 car i.e. should have had a top-fill. But then someone else posted that they have a 67 Mk1 with the rear fill and that has a 20mm cap. Apparently the MGA did have the deeper 1" or 26mm neck, but whether this is compatible with the MGB I don't know. There also seems to be at least one after-market radiator that has a 7/8" neck!

Thinking about it, all the stiffness when trying to turn the cap on and off was down to the extra thickness of the upper seal. In theory the extra thickness will be trying to lift the cap higher, so reducing the pressure of the main seal on its seat, but in practice the ears on the cap and the flanges on the neck should be forcing the cap down into exactly the same position in both cases. However it was so stiff, that if anything the extra thickness of the upper seal might actually be tending to bend the ears on the cap down a bit, which will allow the upper part of the cap to be a bit higher, so reducing pressure on the main seal. That won't be happening now, and having pinched the ears a bit closer to the upper part of the cap, that will be tending to pull the body of the cap down a smidgen more, and hence applying a smidgen more pressure to the main seal. I'm not going to put the pressure gauge back on the system unless I have to, but I will see if any air is bubbling out on warm-up, which would indicate the main seal isn't sealing.

What is surprising - to me at any rate, is that the distance between the seals on the cap (without the rubber ring) is the same as the distance between the sealing faces on the filler neck. Given the significant range of movement of the main seal on it's big spring, compared to the insignificant range of movement available to the spring-steel disk (whose job is to seal under conditions of cooling suction more so than pressure), I would have expected there to be a couple of mm at least more distance between the seals on the cap, compared to the distance between the sealing faces on the filler neck, i.e. to compress the big spring above the main seal by that amount.

  June 2013 Is it holding the correct pressure? For as long as I can remember the roadster has hissed when up to temperature, and the end of the overflow tube placed in a tub of water shows bubbling. This is when running, and for a few minutes after switch-off. The hissing and bubbling gradually slows and stops, then a few moments later it starts again but with a different sound, and now it is sucking air back in through the low pressure valve. In all that time it has never boiled or lost coolant, despite getting up to the edge of the H zone on a couple of occasions, and only needs topping-up once a year at most, and that is with a small weep from the water pump during warm up.

I never really thought about it until I had a problem with the cooling system on the V8, where something was causing air bubbles to get into the radiator, but with the completely filled radiator and remote expansion tank on that car it was pushing coolant from the radiator into the expansion tank, and eventually out of that to be lost. That was probably the pump sucking in air, so could be happening on the roadster as well, but with the large air space above the coolant in the radiator there was no chance of any coolant being ejected, short of getting hot enough to boil. I thought something must have been producing the air that is continuously coming out, but what? As there were no other symptoms or problems, like a head gasket leaking, I had decided to leave well alone. The other oddity is why is it pulling in air when it is cooling down? Ordinarily I'd have expected the system to pressurise to 'normal' i.e. below cap level as the coolant expands with running, then as it cools back down the pressure should drop back to zero, and not need to pull any air back in. Only if you remove the cap when warm, and release some pressure, then replace the cap, will it pull air back in as the system cools further.

August 2016: Subsequent a pal was going through the same process and wondered if it was possible to butcher a spare rad cap as a pressure point, but keep its profile low enough to be able to shut the bonnet. In fact all you need to do is drill through the main seal and backing plate, or remove the low-pressure valve. That will pressurise the space above the main seal, so you can attach the gauge to the overflow port on the filler neck. It will need a good seal between the upper part of the cap and the top of the filler neck though. There is a spring-steel disc seal there, but that is only to prevent a sudden pressure release from spurting out from under the cap (it should escape via the overflow pipe instead) or to allow a coolant recovery system from an open catch bottle with the overflow pipe in it. That may not be enough to hold the main cooling system pressure, but more recently caps have had an additional rubber ring below the spring-steel disc, and that should be enough. See here. However you can also get radiator blanking caps without a pressure seal, which are used on some cars with remote expansion tanks, where the pressure cap is on that tank. Note that the cap portrayed in that link does have '15lb' stamped into it, one assumes that is an example from the range of blanking and pressure caps they have available.

Went for a brief run, and on my return the pressure showed about 3psi. This didn't surprise me as the V8 was only showing 3psi when running in free air after the cooling system problem was resolved. I then blanked off the radiator to force the temperature higher, and about mid-way between N and H the pressure was still only 4psi, which did seem a bit low. Switched off, hissing and bubbling as expected. Over the next few minutes the pressure dropped, and at 0.5psi the hissing and bubbling stopped. Now that I wasn't expecting as the temp gauge showed nearly at the H zone with heat-soak. I removed the radiator cap, and immediately it started boiling slightly - fortunately not enough to overflow but with some gurgling.

After a bit of thought all (well most) became clear - the radiator cap is leaking, in fact only holding about 0.5psi. But the rate of leakage is slow enough for enough pressure to develop from micro-boiling at the hot-spots close to the combustion chambers, to prevent boiling over during running including in temperatures of 30C and with the temp gauge on the edge of the H zone, over the last 20 years or more. However as soon as you switch off, and the micro boiling stops producing more water vapour and hence pressure, the pressure that is already there leaks away past the cap. So when the engine starts to cool and the coolant contract, it has to suck air in through the low-pressure valve in the cap which is there to prevent negative pressure collapsing the hoses. The remaining oddity is, if the system is continually venting air when running, which is presumably water vapour from micro-boiling, why aren't I losing coolant? As I said it probably takes less than a pint over the course of a year, and there is a small weep from the water pump during warm-up. I did some Googling and discovered that water vapour at 107C (i.e. about the upper end of our temp gauges) has a density of 0.0361 lb per cubic foot, whereas liquid water (temperature not specified) weighs 62lb per cubic foot, or 1700 times more. Estimating the bubbles that come out of the overflow at about 3/16" diameter, or .000767 cubic foot, and one bubble per second, is an escape rate of 2.76 cubic feet per hour. Say 3000 miles per year travelled, at an average speed of 35mph, is 86 hours, or an escape rate of 236 cubic feet per year. That's water vapour, so converting that back to liquid water at 1,700 times the density, is 0.14 cubic feet or 1 gallon - E&OE! I kid you not, working through those conversions and estimations got me to 1.047 gallons, I did not start with 1 gallon and work back. The oddity is, I'm putting nowhere near a gallon a year back in, so where is all that water vapour, air, gas or whatever coming from?

I don't have a spare 10lb cap but I do have a 20lb which will do as a test, to make sure it isn't an issue with the cap seat on the radiator itself. I repeat the run with that and if anything the pressure is slightly lower, but blanking off the radiator gets up to 7psi which is closer to what I would expect. Also no hissing or bubbling. New cap ordered, so time will tell if there are any other differences when running.

Update May/June 2014: New 10lb cap received and checked against the old - spring definitely stronger. But after a 25 mile run it is hissing and blowing air from the overflow tube just like the old one! No change in the temp gauge or anything else either, so it will just have to hiss. Update 2: On a subsequent trip there was no hissing! At which point it will just have to hiss or not as the case may be. Then in late June I discovered the head gasket was leaking combustion gases into the coolant, and replacing that seems to have resolved all the issues.

Or so I thought. Shortly after fitting I had occasion to lift the bonnet again, to find the front strip that seals to the roadster had come away and was just lying there. Good job I spotted it, I wouldn't have been best pleased if I had lost it. There was virtually no adhesive between the two pieces, just one thin wavy line and that for only part of the length. Made a better job of gluing it than the manufacturer and since then it has been fine.

Also since then we have had some pretty warm weather, and I'm sure the readings on the gauge have been lower since fitting it. Unsealed there is a pretty large gap between the top of the radiator and the diaphragm allowing ram-air to bypass the radiator, whereas when sealed more will be forced through the radiator to cool the engine, enough to make a visible difference it seems.

A change that I made was to the fibre-board 'trunking' that sits behind the slot. Although the slot and the cooler are distinctly rectangular, the board had a very pronounced curve which was allowing a lot of air to leak past the cooler underneath. I folded sharper corners in the board to match the slot and the cooler, then put a metal strut underneath and up the sides to keep it in that shape, clamped by the strips that hold the board in place, and again it seems to have made an improvement to how quickly the temperature comes down when getting underway after being stuck in traffic.

Mechanical gauge
What lies inside

  Dual gauge replacement: Failed for the second time (first time not long after coming to me in 1990) on the way back from the Rose of the Shires run in May 2012. Checked round the usual suspects and MGBHive were a few quid cheaper than anyone else and I've dealt with them before so they got my order. Outright purchase which surprised me, previously they (had to replace Vees in the past as well) have been exchange i.e. you have to return the old one. Specials like Jaeger and magnolia are still exchange, so one could suppose it's something to do with volume of the 'standard' gauges Caerbont are producing, but inside surely they are all the same, it's only the dial that is different?

First thing was to check the new gauge read with the probe in boiling water - showed just above N so that will have to be good enough, comparing with C and F gauges it looks like boiling point should be more or less the start of the H zone. Next was to check I could undo the nut holding the sender in the head and that was OK too. Nut turned independently of the sender which would be essential if you were removing a good gauge for other reasons, but technically doesn't matter in the case of a failed gauge. Next got the centre air vents out - little screwdriver lifting the tabs, and prised out with a finger-nail. Left hand in the resultant space I could just get my fingers on the knurled nuts, and just undo them - one earth wire under each. The far one was a fiddle as my fingers were round the capillary, so I had to work the nut past the pipe without dropping it (a bit like those rings beloved of magicians that join together and come apart with no visible join). Gauge pulled forwards I could undo the oil pipe and remove the illumination bulb. Now time to completely remove the sender from the head, and to limit coolant loss I got an under-sized bolt and wrapped enough plastic tape round the end so it would 'screw' into the head. Undid the clip holding the capillary to the block, which I know now should be the bottom heater tap bolt rather than the unused tapped hole that the dynamo rear bracket bolts to on earlier cars. Unwound the capillary so it was straight, removed the next clip from under the oil pipe fitting on the shelf by the heater, then eased the whole lot out through the shelf and dash holes.

Carefully unwound the capillary on the new gauge (don't just pull it straight which will twist the tube) so it could be fed through in the reverse direction. Once through the heater shelf, and with the gauge nearly all the way back into the dash (wait for it!) I wound the surplus capillary into three loops by the heater, then fed the sender under the heater tap to the front of the engine, then close to the block until a few inches from the port in the head where I made a graceful 'P'-shaped loop in the capillary away from the block and straight into the port. Removed my temporary plug and screwed the sender into the head. Back in the cabin I attached the oil pipe, and ran the engine to check the oil part worked, and there were no leaks from the joint. Back in the engine compartment I refitted the clip under the oil pipe fitting on the heater shelf, and this time connected the second clip to the bottom bolt of the heater tap. With the nut undone it leaked coolant slightly, but done up again has been leak free (I use Hermetite Red non-setting sealant). In the cabin refitted the illumination bulb, fiddled the mounting clip onto the studs, then the earth wires, and then the nuts onto the studs. Pressing the gauge against the dash (make sure there is an O-ring on the gauge body) and tightening the nuts finger-tight makes them tight enough, the rubber O-ring supplies enough tension to stop them coming undone, it seems. Turned on the lights to make sure the all gauge lighting worked, and refitted the dash vents. Ran the engine for long enough to check the temp gauge moved, then subsequently went for a decent run. Disappointingly confirmed the saucepan test and reads quite a bit lower than the previous gauge, so it will have to be a case of getting used to what is 'normal' for this gauge, then watching out for departures from that as before. Following other problems with this supplier I shan't be using them again.

Slight oscillation during warm-up: My V8 had always done this before gradually settling on the 'N'. As the roadster does it too I took no notice. But after doing a top-end overhaul on the V8 I noticed that it no longer does it, but just rises to 'N' then stops there. During the overhaul I did find the thermostat bypass pipe that is inside the inlet manifold (not the heater return pipe bolted to the bottom) choked with scale which I cleared maybe this was the cause. But then again, why did it always do it, even in the winter when the heater valve was open? Or maybe the thermostat has chosen the same time to start sticking open, we shall see (Update summer 2005: Apart from settling at a slightly lower point by N in winter compared to summer the temp gauge indicates the thermostat is working just fine, and the oscillations haven't returned). May 2014: I changed the roadster thermostat when it apparently started sticking and sending the temp gauge into the H zone during warm-up. On test in the garage immediately after changing it seemed fine and had stopped oscillating altogether, but when driving off from cold it oscillates far more than before, and is chucking out some coolant. This has led to a more detailed investigation of the problem which turned out to be head gasket failure, and since correcting that the roadster temp gauge also now only goes up to 'N' and stops. However a pal with similar symptoms to mine which turned out to be a cracked head still finds his temp gauge goes up and oscillates slightly about N as the system is stabilising.

Wild oscillations at any time:

May 2016: Someone has reported this effect on a 4-cylinder, and changing the capillary gauge seems to have cured it. Personally I can't see how, as the needle requires pressure in the tube to move it, which requires heat to expand it. It could be a mechanical fault in the gauge head i.e. with the gearing, but I'd expect it to be a lot more erratic than that - sticking high when cold and all sorts. But there is another possibility mentioned at the end of the thread which is a live wire shorting out on the tube. That would cause a significant current to flow in the tube, which will give an additional heating and expansion effect to that from the bulb in the cylinder head. February 2018: A pal of a pal has experienced similar. It's also possible that trapped air in the head causes localised boiling, with steam getting onto the gauge bulb, which does send it into orbit as on the V8 when the steam pipe gets blocked.

Even before that I had wondered whether the common situation of the gauge reading lower in winter is something to do with the ambient temperature around the tube. This could be a good 20 to 30 degrees between winter and high summer, which is about 1/3rd the temperature difference between the needle just starting to move (30C) and normal (88C). However as the thermostat has an 8 to 10 degree C difference between just opening and fully open then winter to summer variation is more likely to be from that, as it will be closer to closed in winter and closer to fully open in summer.

V8 conversions. Some conversions don't seem to have the above steam pipe, and on one of these I have seen similar oscillations when the heater temp control is turned to 'off'. Turning the control to on stops it. The heater control being switched on acts similar to the steam pipe, allowing coolant to circulate albeit not through the radiator. Several other people have commented on the same thing and talked about putting a small bore bypass pipe between the heater supply and return pipes before the heater control valve.

The vast majority of us will need a thermostat to achieve a normal running temperature in as quick a time as possible, only cars in very hot climates may not need one, and see below regarding this. An engine runs more efficiently at it's correct temperature, choke - or more correctly in the case of the SU and Zenith carbs - 'mixture enrichment' will be needed for the shortest time saving fuel, reducing pollution, and giving full performance sooner.

But first, what is the thermostat for? I've seen quite heated discussions with one camp saying it sets the minimum temperature, and the other camp saying it sets the maximum. It does neither, and the clue is in the name - 'thermo' i.e. heat and 'stat' which can be taken to mean 'static' or 'constant' i.e. it is a device that attempts to maintain a constant temperature. However it can only do that between two extremes, and those extremes are controlled by the heat output of the engine and the cooling effect of the radiator and the air temperature passing through it. Consider driving along in moderate conditions, with the engine output and radiator output balanced, and the thermostat partially open, which should give a 'normal' reading on your temperature gauge.

• Now consider a long downhill into cooler conditions on a trailing throttle. Much less heat output from the engine, but the radiator is cooling at the same rate as before, so the coolant temperature drops, which causes the thermostat to start closing, and the radiator to dissipate less and less heat, to maintain the coolant temperature. But once the thermostat becomes fully closed, it can no longer regulate the temperature, and at that point if engine output reduces still further coolant temperature will start to drop because you cannot reduce the cooling effect any more, and you will see this on the gauge. You may wonder how the thermostat could close fully on a running engine, but read on.
• Now consider the opposite where the engine is slogging up a steepening incline in hot weather, and the engine is outputting more and more heat as you go. The coolant temperature rises, and as it does so the thermostat opens further, and the radiator dissipates more and more heat. But once the thermostat is fully open, any further increase in heat output will cause the coolant temperature to rise without flowing more through the radiator to cool it, and again you will see this on the gauge. But even without more coolant flowing through the radiator, and a rising coolant temperature, the radiator still dissipates more heat due to the difference between coolant temperature and ambient i.e. the air flowing through the radiator, so overall the coolant temperature is still being 'controlled', or managed, just not to a given temperature.

What causes even more 'discussion' is when driving the same roads winter and summer, and seeing a more noticeable change in the temperature gauge reading, which is when people start talking about blanking off the radiator and such-like, with others saying that if the thermostat is working correctly you shouldn't need to. That's true, and if the thermostat isn't working the effects on the gauge will be much bigger - anywhere from barely getting off C in winter if it is stuck open to boiling if stuck closed. Even with a working stat, it will be closed when you first start the engine and drive off, so the radiator isn't doing any cooling anyway, so blanking it off won't do anything ... or will it? Air is generally ducted through the grille and the radiator to get the best cooling from the ram-air effect, with minimal air that has passed through the grill not passing through the radiator. The foam and rubber seals on the radiator and diaphragm contribute to this. So in winter even when the thermostat is fully closed and the radiator doing nothing, there is still the icy blast (when driving along) flowing over the block, head and sump, and through the oil cooler where fitted, and you probably have the heater on which is cooling things still further, and this is where a 'radiator' blind does have a benefit. By reducing the air-flow into the engine compartment you are reducing this surface cooling effect on engine and head at least, and with the over-mounted oil cooler reducing air-flow through that as well. Adjustable blinds were common as an accessory in the UK in the 50s - it was a factory accessory for the MGA, and some countries with very cold winters such as Canada and Sweden still have them. Whether they - in the form of temporary non-adjustable partial covers of the rad or oil cooler - are needed in the UK is highly debateable, how many of us drive our cars in freezing weather with salt on the roads?

But then, does it really matter? I know engines ideally run at an optimum temperature, but we almost never get extreme lows (I remember -26C in Shropshire in the 70s) in most parts of the UK, and most of us don't drive them then anyway. And the kind of variations the rest of us are going to get when we do drive them are not going to have any detrimental effect on the engine, and only a minor discomfort in cabin temperatures for the occupants. And if you don't like that, don't drive them!

May 2014:

Incidentally this one from Bee had been in at least 25 years. She had always run at around N on the gauge - which is supposed to be 82C, until the second replacement gauge fitted in May 2012 when normal running showed about 7:30 o'clock. Subsequently changing the stat made no difference. I originally thought I had replaced the stat like for like i.e. 82C, but much later subsequently discovered the old one had been 88C. I find I bought two from Leacy in the space of a couple of days, and do remember ending up with the 'wrong' one, and getting the 'correct' one although I can't remember whether I found the old one was an 88C so initially bought that, then discovered the standard stat in Clausager should be 82C and subsequently bought that, or whether I simply ordered an 88C by mistake. Leacy offered to have the wrong one back, but I said for the money I might as well keep it as a spare. I can't find either invoice, but the new spare is an 88C, so I must have an 82C in now. Why the change from 88 to 82 made no difference on the gauge is a mystery, In Jan/Feb 2018 and having the head off again for conversion to unleaded after the same exhaust valve was leaking as before I have the opportunity to fit the 88C and see what happens.

MGA stats had a cylindrical sleeve which moved as the stat opened to block this bypass port, and causes the vast majority of circulation to be via the radiator and virtually none via the bypass port. These are sometimes referred to as 'bellows' stats, however this refers to the technology used to open and close the stat, and not the presence or otherwise of this cylindrical sleeve. This type of stat is no longer available. Later wax-pellet stats as used in the MGB don't have this sleeve, meaning that some of the coolant still recirculates once the stat has opened. MGA engines are said to suffer from inadequate cooling in high ambient temperature conditions if a later wax-pellet stat is installed, but it doesn't appear to be a problem with MGBs.

Bob Muenchausen has a comprehensive page on cooling and thermostats and reports that Neil Cotty in Australia fitted a standard wax-pellet stat and a blanking sleeve to a road car (although it's not clear which car this was done to) to get reduced circulation through the bypass port and slightly better cooling. Hopefully the reduced temperature difference between his ambient and normal running temperature eliminates the risk of hot-spots during warm-up. Neil professed himself happy with the result, but wondered if this arrangement in cooler countries would prevent the engine ever reaching its normal temperature. I can't see why that would be if a stat is fitted as seems to be the case. Bob in Idaho and its colder winters tried the same thing in his 68 MGB and reports that it does take slightly longer, but as I say I can't see how, unless the blanking sleeve reduces recirculation so much it prevents the hottest coolant reaching the gauge sensor, i.e. there is uneven heating. But surely the fitting of a blanking sleeve in cooler countries is unnecessary anyway, and with the bigger temperature difference between ambient and normal you run the risk of damage from localised hot-spots for no gain.

  Coolant Temperature too high:

MGBs had hot weather testing during development, and many have been running in desert states in America and the Middle East for many years with no problems. If you are having cooling problems it is more likely that there is something wrong with your system - either the engine producing more heat than it should or the radiator failing to get rid of it, rather than any fundamental design problem requiring modifications or expenditure on after-market stuff. It may even be the temperature gauge over-reading, although this is more likely with electric gauges than capillary. Also don't forget that it is normal for the temperature gauge to show a higher reading when slogging up a hill in high temperatures than cruising on the flat in winter.

Updated May 2008: Early in 2008 there was a heated discussion about thermostats where one person insisted that the stat controlled the maximum temperature, whereas everyone else myself included said it controlled the minimum. On consideration I changed my view, but to one that says the thermostat maintains a given temperature, rather than a maximum or a minimum. But like any thermostat it can only do this between certain upper and lower limits. If the ambient temperature is below a certain point surface cooling of the block, sump, oil cooler and use of the heater will prevent the coolant getting hot enough to open the stat at all but the engine will still be running below temperature. This is when people in cold areas talk about 'blanking off the radiator'. And if the ambient temperature is above a certain point the radiator won't be able to dissipate as much heat as normal, the stat will be fully open, but the engine will still be hotter than normal. The fact is that more of us experience the latter i.e. higher temp gauge readings in summer than low readings in winter, which is why most people say it controls the minimum temperature. Even giving this interpretation the 'maximum' man refused to acknowledge it and stuck to his guns, so there it was left.

So under a wide range of ambient conditions the running temperature is set by the stat. Below the minimum is relatively easily dealt with by blanking off the radiator, but above the maximum will inevitably result in a higher temp gauge reading than normal. Even in a typical UK summer with temperatures in the high 70s and low 80s the temperature gauge will read higher than normal under some driving conditions. This isn't the end of the world, it just means that the coolant temperature will have to be higher than normal before the radiator can get rid of the usual amount of heat. Mid-way between N and H is no problem at all, and is actually about where the electric cooling fans cut in at 90C, whereas a standard stat is 82C. 'H' on the gauge represents about 110C/230F, which is above boiling point but you should be able to get right up to the H zone without any loss of coolant or steaming because of the pressurised system. Using a typical 33% anti-freeze solution a 12lb cap raise the boiling point to 123C/253F which is well above H on the gauge. I've seen my V8 inside the red zone without problems, other than for my nerves, although by that time you should be thinking about whether you are pushing it too hard for the conditions, or there might indeed be a problem. In either case turning on the heater full-blast can buy you a bit more time - at the expense of comfort!

As mentioned before there are two main reasons why coolant temperature can rise more than it should - either the engine is pushing out more heat than it should or the radiator isn't getting rid of it. In the former case this can be head gasket blown, timing too advanced or retarded, drag on the engine from tight bearings e.g. just after a rebuild all of which can result in a significant increase in heat output, binding brakes (usually accompanied by a smell of them burning), underinflated tyres (slight effect) etc. In the latter case the radiator can be partially blocked either eternally by debris or internally by sludge, restricted airflow through the grille from additional lights, rally plaque, number plate etc. Flow can also be reduced by a thermostat not fully opening, water pump vanes corroded away, and I understand there may be an incorrect combination of engine and pump that results in reduced coolant flow. There is also another possible cause of engine overheating where the coolant temperature coming out of it isn't necessarily raised, and possibly not even an indication on the gauge, and that is when there is sludge in the block and heads restricting coolant flow across parts of the metal surfaces resulting in localised overheating and possibly gasket or head warping problems. Internal sludge when still soft can often be shifted by repeatedly forward and reverse flushing of the engine and radiator until the water runs clear, but hardened deposits may not respond to this and still remain even though the water is clear. There are some DIY radiator flush additives around but they are probably of marginal benefit, both from the safety of DIY use point of view and attacking rubber and alloy parts. Still worth a try before the next step, though. Hard deposits are bad enough in the radiator, but at least this can have the header or footer tank removed, the tubes rodded-out and the tank resoldered, or at worst the radiator replaced. But in an engine often complete dismantling and 'hot-tanking' will be required Hot-tanking is dunking in hot chemical solution to hopefully dissolve any deposits before it dissolves the engine! These chemicals can be pretty caustic and not only dangerous to handle but often destructive to alloy parts.

  Coolant Temperature too low:

In extremely cold conditions and even with a correctly functioning stat the temperature may never reach normal, due to the 'surface cooling' effect of freezing air passing over the sump, block, hoses, and use of the heater. As a point of interest it is this surface cooling that led to there being 'summer' and 'winter' thermostats. In winter a higher rated stat was fitted as more of the cooling would come from surface cooling of the block and sump. In summer a lower rated stat was fitted as the surface cooling effect would be lower. The effect of this surface cooling can be reduced by partially blocking the air-flow through the radiator and oil cooler. Contrary to popular belief this doesn't stop the coolant in the radiator from being cooled too much, as the thermostat closes anyway when the coolant temperature drops below the rated temperature, it simply reduces the amount of surface cooling. Blanking-off will however make the coolant too hot if you blank off too much or the weather doesn't warrant it.

Michael Beswick has pointed out that the typical 13" x 6" (excluding the bump on the top) rally plaque frequently issued on organised runs in the UK tied on with string makes an excellent oil cooler blind for cold weather running, for those cars with the oil cooler above the apron at any rate.

Thermostat Replacement: May 2014

The day before a 300-mile run I went to fill up. Glancing at the temp gauge on the way to the petrol station I was surprised to see it on the edge of the H zone and coming down, normally it barely goes above N before coming down and oscillating a bit before settling on its 'normal' position. On my return home checked the coolant to see it barely above the top of the tubes, even though it was hot i.e. fully expanded. Topped up when cool and it took a litre - too late to do anything about whatever the problem might be now. Set off for Chepstow next morning with some trepidation, and the same thing happened again although as I was watching it this time I saw it go well into the H zone before settling down. Fine for the rest of the day, and on our return home checked the level and again it was well down, when cool it took another litre. Next morning took it round the block, same thing happened, level dropped by the same amount. Looked carefully round the engine, no sign of any leaks, oil is clean, but there is dampness below the overflow tube. As it seems to lose the same amount regardless of whether it is 3 miles or 300 miles from cold, I reckon the thermostat is sticking, it's getting hot enough to boil, and pressure is coming via the bottom hose and pushing the coolant out of the overflow. So thermostat replacement is an immediate task as the next run is in three weeks, and hope that fixes it.

First job is to drain at least some of the coolant out, which entails removal of the bottom hose. This came to me with the car 25 years ago, although I have replaced the water pump, and I think I've also had it off the rad for flushing, both many years ago. Because the bottom rad pipe faces upwards it's a struggle to remove the bottom hose from that first, but if you remove the pump end first water will go everywhere, whereas at least removing the bottom end first you can catch most of it in a bucket. Before this happened I was planning to drain and flush the system anyway, and replace the coolant, as when hot the coolant in the rad is quite orange, but not when cold, implying that a lot of sediment is being stirred up when in use. I was going to do it by repeatedly filling with water, running up to temp, then draining hot or nearly so, but given the difficulties getting the hose off that's not really an option.

Next day as repeated filling and draining hot wasn't going to be an option, for the sake of four bolts I removed the rad so at least I could agitate that and rinse it out. The two struts that go from the top two bolts to the wing bolts have to be slackened at the wing end so they can be swung out of the way far enough to angle the rad forwards so the top connection clears the mounting panel (and my foam seal). When replacing these struts, remember to position them over the welded nuts on the rad flange as you are screwing the long bolts in, otherwise you will have to undo them again!

Replacement stat arrives. Decide to test it in a pan of water just in case. But that is not as straightforward as it might be as we have an induction hob which means using one of SWMBO's shiny pans. She's OK with it, but I find my digital oven thermometer is being confused by radiation or something from the hob, so I have to keep turning the hob off to read the correct temp. It seems to be opening a shade after 82C, but I had tested the old stat earlier using a blow-lamp on a tin (which took ages hence doing the new one - clean - differently) and that seemed to be about 90 or 92C. Not as high as I was expecting, the temp gauge was going higher than that - maybe to 105C on a Centigrade gauge, and my gauge reads a few degrees low.

What to do about the bottom hose is a bit of a conundrum. As this one is still supple, shiny and smooth do I refit it? Or use the spare that I bought years ago? Given the quality of hoses these days a 'new' one could well fail before my 25+ year-old one. I've had my spare for nearly that long, so is that now perished? Or did I buy it long enough ago before the quality got so bad? I decide to put the original on first, fill it with water and run it up to temperature first to check that there is nothing else wrong and it was the stat before draining and refilling with anti-freeze. I'll fit the new hose and clip at that point.

When it stops raining I get the cars out to do a test run. No leaks, but I'm slightly concerned to see the temp gauge goes to about mid-way between N and H before it comes back down again. No coolant loss, but then I've only just refilled an empty system so the level is probably low anyway. Switch off and it is well down, and takes about 1.5 litres from a just-boiled kettle. Let it cool down to about 70 degrees, and run it again. This time the temp only gradually creeps up from mid-way between C and N to it's normal position. Initially the rad was cooler than the thermostat housing which indicates the stat was closed, then it obviously opened. Switched off, checked the level which was as I had left it. So nothing more to do but put it back in the garage (it had started raining again anyway) until it has fully cooled down overnight and try again.

Next day coolant level still good so run it again. This time I film the temp gauge while monitoring the temperature on the head just above the temp gauge sensor with my infra-red thermometer and the rad header tank with my hand. The rad only warms slightly to begin with, then rapidly got hot as it should. At that point the thermometer indicated about 78C. Replayed the video to see that the temp gauge simply rose to it's normal running position and stopped there, which is a relief. Coolant level still good, and nothing chucked out of the overflow. So the final job once it has cooled down is to drain the plain water which I have put in to start with, and refill with anti-freeze.

The odd thing is that ever since I've had the roadster it's warm-up has been totally consistent in that the temp gauge would always go slightly above N, then come down quite a bit as the stat opened, then go up and down less and less until finally stabilising on its normal running position. When I got the V8 that was exactly the same so I took no notice. Then after many years and tens of thousands of miles it developed a cooling system problem that resulted in coolant loss, but after fixing that the temp gauge stopped oscillating during warm-up and now just rises to its normal running position and stays there.

In the event it was a week or so later that I removed the bottom hose to drain some water out prior to adding antifreeze. This should be done by adding the required amount of neat glycol first - 1.9 litres for 33% with a heater for example - then top up with plain water as that takes account of any old coolant or plain water left in the system, which can be quite a bit. How you get on with replacing glycol with ForLife I don't know, but I do know that with Evans waterless you have to go through several flushing cycles first. I had completely removed the bottom hose as I had decided to be rash and fit my new spare. I smeared Vaseline round the first half inch or so of the inside of the new hose, as dry it was proving difficult to get the hose on any of the three pipes, and the heater return pipe was still quite hard, and tightened up the clamps. Put 2 litres of neat glycol in and that was enough to fill the radiator to above the tops of the tubes, which was a surprise, I had expected to be adding some water as well.

Started it up to hear a load drumming noise so immediately switched off. As suspected the fan was hitting the new hose, hadn't even considered that when fitting it. There is only about 1/4" clearance each side of the hose to both the fan and the timing cover, so it has to be positioned quite carefully. I had noticed the old hose had been rubbing on the timing cover. Fortunately lost no coolant - which would have been neat coolant - when adjusting the hose.

Restarted and let it warm up, looking round for any leaks and keeping an eye on the rad level with the cap off, feeling the thermostat housing, and looking at the gauge. Got maybe half-way from C to N (bearing in mind this gauge reads low) when the rad level suddenly dropped like a stone. Immediately switched off, boiled a kettle (neat water!) to top up with, and it took most of a full kettle. This must be due to the new thermostat having no jiggle valve, not even a notch in the main valve, so it traps air beneath it until it opens, even though it has coolant on top of it from filling the rad. This must be why when I first filled and ran it with the rad cap on, the temp gauge went up quite a way above N before it came back down, and I subsequently found the rad 'empty' needing quite a bit to bring it above the tubes again.

Restarted, no movement in the header tank for a while, then suddenly started flowing across the tops of the tubes and turning blue. I hadn't noticed that it was clear to begin with, even though I had only put neat fluid in. I suppose it is 'heavier' and had displaced the water that was left in the bottom of the rad. At the same time the level started dropping again, but quite slowly, so I just trickled in some cold water to keep it above the tubes, perhaps half a litre. All this is best done with the front of the car slightly higher than the rear to aid bleeding, particularly of the heater, the tap for which should be open. Left it overnight and gave it another test in the garage, and all was well, no oscillation, and no coolant loss.

  However next day I took it round the block, to find the wild oscillations had come back. Got back home and found it had lost about half a litre. The same happened again next day, with a catch bottle on the overflow tube. This has to be more than simple purging of air left in the system. By now we were only a couple of days away from the Arden Run, so nothing I could do about it. But as it seems only to chuck some out if filled more than 1/4" above the tubes when cold, and only as the stat opens, it should be OK, and it was.

I'm wondering if it's the pump sucking in air, as seemed to be the case with the V8. I changed the pump almost on getting the car in 1990 as the bearings on the original had gone. It was a pig to fit - an alloy after-market item, with one of the holes having to be ovalled to get all the bolts in, even though the gasket from the same source fitted as it should, and one of the bosses ground down as otherwise the bolt was too short. After a few years I noticed that started weeping on warm-up, so got a pukka cast-iron pump prior to a trip to France and put it in the boot waiting for the weep to get worse - that was 17 years ago and it's still in the boot! It could also be a problem with the bypass passages in the head and/or block not circulating coolant while the stat is closed, but the pump is easier to change as a first go.

Warmed up on the drive and no problems at all, gauge just rose to its 'normal' point, stat opened, pressure up to about 6psi - higher than I was expecting given the V8 pressures. Thermometer on the stat and temp gauge sensor housing showed about 80C. Blocked off the rad, allowed the temp to rise till the gauge was right over the N and the housing temp was 88C. Let it rise to between the two little dots i.e. on the edge of the H zone, temp was 107C, pressure about 9.5 psi, so the 10lb cap is basically OK.

Switched off and the rad cap was hissing, and the pressure started slowly dropping, even though the temp gauge was actually going higher due to heat soak. So although the cap does seem to be 'maintaining' just under 10 psi, it seems there is a small leak from somewhere associated with the cap or the rad fitting, that is allowing pressure to leak away even after it has dropped to below 10 psi. Removed the 10lb cap and fitted a 20lb I had from when I was investigating the V8, ran it again with rad blocked and the pressure got up to about 15psi. Switched off no hissing - faulty 10lb cap seal?

Refitted the 10lb cap and let it cool right down for about 3 hours with bonnet up and cool draught through the garage so the temp gauge was on the end stop and the pressure gauge zero. Drove round the block and I was surprised how quickly the pressure rose, to about 12psi. Temp gauge did its usual thing of rising towards H - but not reaching it this time, and oscillating before it settled back down to its 'normal' but low point. Pressure gauge steady. Got back home, no coolant in catch bottle, and cap not hissing this time!

When I had the V8 cooling system problem that would always rise to cap pressure. Afterwards it only rose to 11psi (15lb cap fitted) when the fans cut in i.e. mid-way between N and H. When the fans cut out it was 6psi, and running in free air on the motorway only 3psi. So I'm pretty sure the roadster system is pumping up just like the roadster was. But whether it is down to the pump sucking in air, or a leaking head gasket, only time will tell. As the pump is easier to change, I did that first.

Unfortunately things were just the same. Before I'd got to the end of the road the pressure was up to 5psi with the temp gauge still on C. As the temp rose the pressure went up to about 12 psi )10 psi cap), temp went up to just over N before coming back down. Pressure dropped to about 5psi, and crept back up to 10 again on the way back home. Perhaps about 250mL in the catch bottle, and the rad cap hissing. I suspect the pressure drop just as the temp dropped was the stat opening, and either localised boiling or something else chucking out some coolant. Looks like the head gasket then, but really I need to do a combustion leak test of the coolant, a compression test and a leak-down test to get as much diagnostic info as possible before doing anything about the gasket.

The combustion leak test proved the head gasket was leaking, and replacing that had the same effect on the gauge as the V8 top-end rebuild, i.e. the gauge now just rises during warm-up and stops, very little if any oscillation, and the rad cap no longer hisses.