Tuesday 29 September 2020

Part 2: Commissioning and Performance of the Marshall Supercharger on the MG PA

The first attempt at starting the car proved to be an anti-climax and something of a damp squib. Although I had never slackened off the distributor clamp during the fitting of the supercharger, or at any other time, my buddy noticed that the distributor appeared to be loose as I attempted to start the car. Indeed it was …so, time to check and reset the timing.

With the distributor top off we found that the car had been fitted with electronic ignition as there was a sensor and a 2-wire pickup rather than a set of points. Luckily the manufacturers name was on the side and so, after getting the wiring data for setting up the timing with a multimeter and using the TDC mark on the flywheel (observed through the inspection cover on the bell housing), the distributor position was reset to the equivalent of “points just opening” (multimeter voltage falling to zero), to give a static (fully retarded) timing at TDC. The electronic ignition takes care of the advance curve.

Next try at starting: Nothing. Not even a misfire, a backfire or anything. Nothing. So, plugs out and checked for sparks. There was a spark, although not as strong as the fat blue spark that I would have expected from an old-school points system, but as the car was running perfectly before taking off the twin SUs I assumed this was just a characteristic of the electronic ignition.

The next check was: “is fuel getting into the cylinders?”. With soaking wet plugs the answer to that was obviously “yes”. Now, in my experience and also that of my buddy (who has also spent his entire working life in the automotive trade) is that if you have fuel and sparks you will get something to happen. It may not be the right thing but there will be an ignition event of some kind - it may be when the exhaust valve is open, when the inlet valve is open or, if the engine has been timed on the wrong stroke, then at “phut” during TDC valve overlap. But, here, there was absolutely nothing.

Now, I have mixed feelings about electronic ignition. I am a mechanical engineer by profession and pretty well all my non-professional experience is with magnetos on vintage motorcycles, which I understand and have got to trust over the years. We have good people that will re-magnetize the magnet, rewind the armature and take care of the few mechanical bits and pieces of points, bearings and the advance/retard. They are also straightforward to set up with a timing disk on the crankshaft. Advance/retard is manually controlled with a lever on the handlebars connected to the magneto with a bowden cable.

But with electronic ignition you never quite know what is going on. Out of my two dalliances with electronic systems I have had one good and one bad experience. The good experience was to replace an original Lucas magneto (because it needed rebuilding) with a BTH self-generating magneto on my Vincent Black Shadow. It came with clear instructions and worked “out of the box” and, many years later, I have never ever had a problem with it. It starts, hot or cold, with just a lazy push on the kickstart.

The bad experience was with a (name removed) electronic system fitted on a Velocette Thruxton.  This was similar in concept to that on the MG in the respect that it consisted of a sensor that replaced the usual magneto and required a coil and battery. This was not my bike but that of a good friend of mine who was, at the time, a very competent restorer of Velocettes in Chicago.The ignition system and pretty well everything else in the engine was brand new. We kicked and checked all morning – there was fuel and a good spark at the plug (at least externally) but all we could get out of the engine was an occasional “phut”. In spite of the fact that there appeared to be plenty of power in the battery – a good spark, lights worked etc. we decided to connect in another battery. The bike then started on the next kick.

The lesson learned from this experience was that the battery must not only be good (it was already a brand new battery) but also absolutely fully charged to ensure that the electronic ignition system would function correctly. It transpired that the battery had been charged a few weeks earlier but had lost sufficient power in the intervening period to preclude a good in-cylinder spark. The corollary to this is that electronic systems can appear to be functioning OK, judging by a strong spark outside of the engine, but are failing to produce a spark under compression pressure to ignite the charge.

So, back to the MG. I reasoned that the additional power required to turn over the engine with the supercharger could reduce the voltage sufficiently that there would be no in-cylinder spark and decided to replace the entire ignition system with a good old-fashioned points-based distributor. A new distributor, leads, plugs and Lucas coil were ordered and fitted. Now, I know that advocates of electronic systems will decry this move as a retrograde step (and me as a dinosaur) but I personally have more confidence in these “old fashioned”  systems and ….they have been powering engines successfully for an awful long time.

With the new distributor and coil fitted, time for the next attempt at starting. Nothing. Absolutely nothing. Each starting attempt was getting to be quite an anti-climax.

Back to the fuel system. After another phone call to Oliver Richardson it transpired that I had an incorrect jet and needle in the carburettor. The diameter of my jet was 0.125” whereas it should have been 0.1”. It might not sound much but it’s a 56% increase in area. A new jet and RA needle were ordered from Burlen Fuel Systems. However, the needle is fitted into a short sleeve at the bottom of the dashpot and a new sleeve was also required and so this was ordered too.

Unfortunately, the sleeve is a tight interference fit in a blind hole and the existing sleeve cannot be removed easily. I tried! It is also hardened steel and it’s not possible to bore it out in the lathe and so another trip to my tame spark-eroder to take out the inner of the existing sleeve leaving a ~0.010” wall thickness that could then be collapsed without damage to the dashpot sleeve hole.


Although the engine would not have run well being so much over-jetted it should still have run and so there was not really any optimism that this would be the solution.

I also found that there was a large pool of petrol at the bottom of the blower resulting from the many failed attempts at starting and this was soaked up with blue-roll and allowed to dry out. It would be more than unfortunate to get a backfire into the inlet manifold and set fire to the supercharger after all this effort….

What next? Having checked and rechecked everything there comes a time when you run of ideas as to what to look at next. By this time I had added another nearly brand new and fully charged battery off my Austin Healey in parallel with the MGs battery to ensure there was no shortage of power during cranking. I went back to the ignition system and, although everything was new, again checked the spark at the plug. I noticed that the spark on the plug of cylinder #1 was a bit more intermittent and anaemic than I would have expected. I took all 4 plugs out and checked their sparks with them resting on the blower. I was hoping to see 4 strong sparks and hear 4 electrical “cracks” during cranking …but that was not the case. There was an occasional weak spark from one or other of the plugs but nothing that would inspire confidence to start and run an engine.

I removed the HT lead from the coil to the distributor at the distributor end and, as hoped for, there was a massive, nearly 1” long continuous stream of blue sparks during cranking. There is only the (brand new) rotor arm and the carbon and brass pickups between the input and output of the HT leads at the distributor so the resistance of the rotor arm was immediately checked and it turned out to be an open circuit. It should have been a 5 kohm resistor to supress interference but that was not the case here. It would seem the problem had been found and another rotor arm was immediately procured (and checked!).

Next try. Nothing!

At this point I was running out of ideas and called Oliver. He came over a few days later and quickly went through the setup. It turned out that I had not set up the new needle and jet quite correctly (not centred properly - the dashpot should bottom with a satisfying “clunk”), the number of turns of the jet adjusting nut was wrong and I had screwed-up the timing.

Next try. It bust into life. What a relief! This was the first time that the engine had started with the supercharger fitted. We spent an hour or so checking various things until an oil leak from the top of the engine brought proceedings to a halt.

A few days later, with a new cam cover gasket and putting the remaining bits of the car (radiator tie rod, bonnet etc.) back together it was time for a first road test. Although the car went reasonably well, the performance was not markedly different from the original twin SU setup and was running significantly rougher. It had all the characteristics of being far too rich, which was confirmed by a check on the plugs. In fact, they were so black (as was the inside of the exhaust tailpipe) you could have mistaken them from being inside the chimney of Battersea power station for 10 years. It was time for someone that had a greater understanding of SU carburettors than I to set this up properly.

A slot was booked on a reasonably local rolling road and I duly drove the car over with the hope that the dynamic timing and especially the carburation would get sorted out.

It turned out that the fully advanced timing was at 24 deg BTDC so that was reset at 33 deg BTDC and then attention was turned to the carburation. The CO measurement of 12% @ 4000 rpm confirm just how rich the setup was. Another needle that should have been only very slightly weaker was tried (based on the flow areas of each needle/jet vs needle position in the jet - details here ) but the engine would barely run on this needle. By this time we had to call a halt to testing as, among other things, I now had a broken exhaust manifold that was leaking at the front cylinder. This was probably a delayed consequence of dropping the engine at the front to put on the new pulley (before I had decided to make the 2-piece pulley)


At this juncture I was learning that setting up SU carburation properly requires a lot of experience. This approach, ie selecting a needle from the literally 100s of needles with different profiles is quite different from setting up the carburation on vintage motorcycles where the jet size, rather than the needle, is the variable. I am quite accustomed to making my own jets for ancient Binks and AMAC carburettors using a set of drills in 0.1mm steps. In fact some of these early carburettors don’t even have needles  ….but I digress.

And so my first foray into rolling roads was not entirely successful. If I summarize the day, it took me 11 hours from when I left home to when I returned (I only made it 3 miles towards home from the rolling road before I had to call out the RAC to rescue me), I had a car that performed worse than when I started the day, now had a broken exhaust manifold and it had cost me over £300. Not a happy bunny!

Another call to Oliver along the lines of “help!”, as I was running out of ideas as to what to try next. He strongly recommended that I take the car to Sigma Engineering in Gillingham down in Dorset, who had the right expertise to set this up.

The first step was to sort out the damaged exhaust manifold. A quick call to Barry Walker soon had a replacement plus gaskets on the way and, after grit blasting and a couple of coats of Frosts heat-resistant paint, was fitted.

As the car was at least now running again I called Sigma Engineering and made an appointment for a slot on their rolling road. It’s about 100 miles from my home and so we trailered the car down and they duly set about sorting out the engine setup.

I cannot praise enough the work that the 2 guys did on the engine. Pete Lander is regarded as the guru on building and setting up straight-6 Jaguar engines for performance but I believe that they also see quite a few little MGs. He mentioned that he hadn’t seen one of these with a supercharger fitted since Friday  …and this was Monday!

In addition to the strobe, the engine was also instrumented for fuel pressure (this is important so that the fuel level relative to the jet is maintained) and the lift of the dashpot is monitored with a mirror. The first run confirmed the level of richness and another needle was selected. This turned out to be a bit too lean with spitting back into the intake and an inability to rev. The 3rd needle was spot on and, after resetting the timing to 37 BTDC fully advanced and checking the points and distributor bob-weights, we ended up with clean combustion throughout the rev range, confirmed by the much lower levels of CO, and maximum power of 42.6 BHP at the crankshaft (the losses through the transmission and tyres is estimated from a coast-down test).

Although I was quite happy with this result, Pete was disappointed and expected more power. I was satisfied because we had more than the 36 BHP that is quoted for the original N/A engine (although I have also seen a figure of 34.9 BHP quoted) and we had clean combustion. It seemed that the amount of boost provided by the blower was only ~3 psi and this would be reason for the lack of power, but without changing one or other of the pulleys there is not much that can be done about the boost pressure. Without a baseline measurement of power from this engine before the supercharger was fitted there is no way of knowing exactly the power increase; it can only be surmised by reference to the quoted power output of this engine type from various sources.

If we take a value of 36 BHP for the original engine the increase in power at 5000 rpm is 18% and if a value of 34.9 BHP is used then the power increase is 22%.

I could not find any references to the expected power increase for a Marshall supercharger on a PA but I did find here advertising data for a Marshall supercharger fitted on TD and XPAG engines. Running at 6 psi boost pressure these gave increases of 24% and 26% respectively at 5000 rpm. This is not that different from the figures obtained from this project.

The litmus test, of course, is how it drives on the road. Although I have only so far taken the car out on one decent run (I am reporting this the day after its rolling road outing) the performance increase is very apparent. I had been driving this car for more than 4 years before fitting the supercharger so I have a pretty good feel for what it will and won’t do. The increase in torque at lower engine rpm is very noticeable; I can now accelerate up hills in a higher gear than would have ever been possible previously. The boost pressure at a steady 3500 rpm is between 3.5 and 4 psi; this is quite close to the data on the XPAG engine quoted previously which showed 4.5 psi at that speed. As boost pressure is increasing with engine speed this could easily translate into 5 – 6 psi at 5000 rpm on this engine. The car has also not lost any of its driveability and is just as “sweet” to drive as before.

So, how to summarize this project:

Cost: I don’t keep exact records but if I exclude the cost of another exhaust manifold (which would probably not have broken if the engine had not been lowered so far to fit the new pulley) and the first rolling road outing the total cost is probably about £8,500.

Effort to fit and commission the supercharger: More than I thought initially and I hit of lot of “gotchas” along the way. I have well equipped workshop facilities at home, I enjoy engineering challenges and I have learned a lot along the way but it was more effort than I anticipated.

Would I do it again – was the required effort and money spent worth it? It wasn’t cheap and quite a bit of effort but I am very pleased with the eventual result. Based on my limited testing so far, it has transformed the performance of the car. So, Yes, I would do this project again.

….except that I don’t have another MG PA, so now back to building my AJS V-twin world record attempt replica…

 Again, Thanks to Oliver Richardson for invaluable advice and for sorting out various things on this project. I’d still be stuck in the mire somewhere if he hadn’t helped.

Also thanks to the guys at Sigma Engineering for applying their knowledge in the final stage of getting this to run properly. I would wholeheartedly recommend them to anyone that needs real “old school” experience to set up their engine.


Sunday 27 September 2020

Where to Start….?

Having decided to build an approximate replica of this vintage world speed record attempt beast the first question is …..where to start? The engine dominates the bike and so the obvious place on which to embark on this project would be the V-Twin engine itself. And so the top-down plan was ….build a V-Twin OHC engine that is similar in design and appearance to the original AJS engine, find a chassis to put it in plus an appropriate gearbox and then build the rest of the bike. Clearly, this wasn’t going to be a 2 month project…!

As with the AJcette, the intention was to use Mk1 cammy Velocette cylinder barrels, heads and camboxes, as I had plenty of them, and to graft those onto bespoke crankcases, using chain drive to the camshafts. There are variants of these early OHC engine cylinder barrels; the earliest barrels for the road bikes had the part number K22 and these evolved into barrels with the part number K22/4 which have a noticeably thicker section where the barrel joins the base flange, seen below.



I chose 2 of the latter, thicker barrels. These are distinct from the KTT barrels of the period which have a thicker base flange and part numbers K22/2, K22/3 and K22/5. 

As I mentioned in a previous blog, I do not make drawings and do the detailed design “up-front”. I do make scruffy drawings (more like dimensioned sketches) as I’m going along so that I know what dimensions to machine but I do not, as a rule, know those dimensions until the previous “bits” have been made. However, for the V-Twin, I made an exception and made some general layout drawings before I started. Why? Well because there is otherwise a high probability of getting halfway through the manufacture and build and finding some silliness, such as a screw thread needing to be in the same place as a bearing.

The 3 drawings, which were made at 1:1 scale, are shown below:

The yellow line down the middle of each drawing is where 2 pieces of graph paper are stuck together with tape. Design features of the single-cylinder AJS OHC counterparts are carried over including:

-       2x drive-side main bearings separated by an oil distribution ring

 -      The same gear and sprockets arrangement for the camshaft drive, the only difference being that there are now 3 sprockets in total – the 2 inner sprockets to drive the camshafts and the outer for the magneto

-       Although it can’t be seen in these drawings, the same Weller chain tensioner arrangement would be used.

I planned to use Alpha Bearings to make the entire crankshaft and conrods, as I had with the AJcette, and after a lot of discussion with Max Nightingale, I decided to use side-by-side connecting rods rather than a knife-and-fork arrangement and this can be seen in the drawing in that the cylinders are staggered. Why? Because it is easier to make con-rods for a side-by-side arrangement rather than the more complicated knife-and-fork design. As it turned out, this fairly major design decision was changed during the build and knife-and-fork con-rods and in-line cylinders were used ….but more of that later.

Apart from the design, the project kick-off also signalled the time to start collecting bits and pieces that would be needed. As this is a 500 V-Twin an appropriate magneto would be required and, by luck, a lovely 500 BTH magneto with anti-clockwise rotation turned up on ebay; it wasn’t cheap but was in excellent condition.


And so this was the starting point: some layout drawings, cylinder barrels, heads and camboxes and a magneto. The barrels would need appropriate pistons and the heads and camboxes would need refurbishing ….but it was a start. As not much could proceed without crankcases the first task would be to figure out where these would come from …which was the next part of the project.

Saturday 19 September 2020

The AJS V-Twin Project

I wish I could pinpoint the moment when I decided to do this project. I have plenty of projects-in-waiting (8 in fact – 5x Mk1 cammy Velos, 2x Big Ports and a 33/7 OHC AJS ….oh, and I nearly forgot, a Velocette Thruxton that I need to bring back to life) but after I had completed the AJcette project, with which I was very satisfied, I decided to use more of my stock of early Velo bits and make a V-Twin version. This is another decision-on-a-whim that has, so far, engaged me for more than a year of my life.

In addition to the complete project Velo bikes, I have also acquired many Mk 1 cammy bits and pieces over the years. One often stumbles across these purely by chance. One day I was at my grit-blasters premises getting some bits cleaned up and I just happened to mention Velocettes and he immediately said that another of his customers had mentioned to him that he had some Velo bits that he wanted to get rid of. I phoned this guy – he only lived about 5 miles from me and, sure enough, he had bucket-loads of bits. My luck was in - they were all Mk 1 OHC parts - crankcases, barrels, heads, crankshafts, camboxes, bevels, pistons – including a lot of racing pistons with huge domes ….and so I bought the lot. He had no connection with Velocettes and I have no idea how he came by these parts. His hobby was collecting Sinclair C5s so maybe he was planning to develop a go-faster kit…

Now, making an OHC AJS V-Twin would not be an entirely original idea because AJS made a one-off of such bike in 1930 for an attempt on the world speed record. The engine is essentially two 500cc R10 cylinders/heads/camboxes on V-Twin crankcases. The project has been recorded by Steven Mills in his excellent book “AJS of Wolverhampton”, which documents AJSs history up to the time they were bought by the Collier brothers.


There is also a good precis of the history of the bike subsequent to the speed record attempts here

One can only imagine how awesome this bike must have appeared (and sound!) to onlookers 90 years ago.

Interestingly, and apart from the obvious differences of being a V-Twin, the details of the internal design are identical to the single cylinder counterpart from which it was derived, although the rear cylinder head is a mirror-copy of the front head to avoid the rear exhaust interfering with the front carburettor.


Acknowledgements to whoever owns the copyright of the above 2 pictures.

The engine started life as naturally aspirated (as in the above picture) but, following its first failed attempt at the motorcycle land speed record (it made around 130 mph), it was supercharged. This increased its top speed only marginally and it failed again at the land speed record (those pesky Germans kept going faster) and it was eventually sold and ended up in Launceston, Tasmania. In fact, I was on holiday in Tasmania in February 2020 and tried tracking down anyone that might have a recollection of it but I think this history is now lost. The garage when it “lived” is now a coffee shop.

It ended up back in the UK in 1981 and was restored and now resides in the National Motorcycle Museum. Will it ever be seen (and heard!) running again?

And so this was the project on which I embarked. It would not be an exact copy and it wouldn’t be 1000cc but I would try and make something that closely resembled this amazing looking bike.  

The idea, in its basic conception, would be to use 2x Mk1 OHC Velocette cylinder barrels, heads and camboxes and to make the rest of the engine. A chassis and gearbox etc… would then be added to make the rest of the bike. What could be simpler….?

This blog is now the story, in detail, of how this project evolved.

Wednesday 16 September 2020

The AJcette Shakedown

Anyone that has rebuilt a vintage motorcycle will know that, when you think it is finished and can just fire it up and start riding, the fun really starts! There is always something that needs adjusting or fixing and the AJcette was no exception.

Although the engine felt tight for the first couple of miles, it was encouraging that the bike would go very well when run in BUT I got back from my standard ~3 mile first-run-on-the-road course covered in oil. It was “breathing” more than a bit and if I had ridden another ½ mile the oil tank would have been empty. The problem was that far too much oil was going into the cambox and was simply pouring out and covering the back of the bike. I had already plumbed a small tap into the cambox oil feed and calibrated this to what I thought would be an appropriately small amount of oil but this was obviously far too much. The entire plumbing can be seen below with the tap for the cambox oil feed on the RHS.


The first obvious step, therefore, was to reduce the amount oil reaching the cambox by closing the tap further. Unfortunately, this was starting to move into dangerous territory as the tap was already nearly closed – it was about ½ a turn open and the flow rate at this tap setting is very sensitive to the slightest change. Another simple flow rig was built to test the tap setting and the tap closed down appropriately. Same problem as before - too much oil. Next try ….and now with the tap nearly closed (and lock-wired) and a very low flow rate on the rig. This resulted in ZERO oil flow rate on the engine and 2 cooked rocker skids; luckily the cam was OK. The picture below shows the 2 destroyed rocker skids together with a new one.


It became clear at this point that using a model steam engine tap to control the oil flow rate would not work and a new design would be needed. The difficulty in knowing how much oil is required is that it is pretty well impossible to measure the oil flow rate entering the cambox in a cammy Velocette engine although it would be possible to measure the flow rate of the oil being returned to the engine ...but I didn't have one handy at the time. Oil is forced upwards into the upper bevel drive from “down below” and then passes through slots in the shaft and bush that overlap once per revolution. Any leakage through the shaft and bush, which depends on their clearance, needs to be added to this.

One way to control flow rate is to use an orifice, the simplest type of orifice being a hole in a plate. The downside of using an orifice is that if the hole gets blocked there would again be zero oil reaching the cambox. This can be mitigated by using an appropriate sized filter upstream of the orifice and my revised solution therefore consisted of a self-contained flow control device that consisted of an orifice and a filter.

The first incarnation of this is shown below.



This was set up initially with a 1mm diameter hole as this gave a rig flow rate equivalent to the last (reasonably) successful flow rate, ie a flow rate that did not destroy bits of the engine!

In spite of reducing the orifice diameter to 0.4mm (testing each time on the flow rig) and introducing a much finer 100 mesh filter (this should, theoretically, stop any particles that would block a 0.4mm orifice) there was still too much oil “breathing” out of the cambox.

As I did not want to use a hole less than 0.4mm, attention was now turned to the oil drains in the cambox and how to ensure that oil was not prevented from coming out due to the downstream pressure. On the Velocette engine, the oil from the various gullies is collected and fed back into the crankcase and into the inlet valve guide. This was obviously not entirely successful as Velocette introduced a scavenge pump on the end of the camshaft on the KTTs to feed excess oil back to the oil tank.

On the AJcette I had the option of either feeding the oil into the bottom of the timing case or about halfway down the timing case where there is a hole to check operation of the chain tensioner.  In both cases, there are pressure pulsations that result from the crankcase pressure and there is therefore no guarantee that oil will flow, under gravity, into either.

Another flow rig, this time on the engine itself, was therefore constructed to check independently the oil flow rate out of the cambox by collecting it directly in a container and also to check that oil would flow into the engine and, in the case of the latter, by inserting a one-way check valve into the oil line. The check valve is another item from the model steam engine world and uses an extremely low inertia shuttle. The hope was that, even if the check valve was not perfect with the high frequency oscillating pressure, it would allow oil to flow in only one direction and might even provide some suction.


Testing of this arrangement showed that it worked and the final solution was to incorporate the check valve and a small catch tank into the oil line. It also provided confirmation that oil was reaching the cambox …and approximately how much.


The cambox lubrication was the only major problem that needed solving and the bike has subsequently been ridden a few hundred miles, including a trip to the 2019 Manx GP, without further issue.