Well, I've been annoying the neighbors all weekend. Still trying to get the bike running acceptably, am finally feeling like I'm getting somewhere... I'm writing this up now while I remember what I've been doing.
I noticed a while ago that the bike has different length headers prior to the cross. The exhaust system is fairly simple: a curved header runs from each cylinder head into a four-way cross, which then feeds twin silencers. The headers enter the cross facing each other, but there are curved guides halving and feeding each header ninety degrees each way into the twin exhaust outlets. The header pulses end up running side by side in the same directions, not into each other. There is no provision for expansion, as more modern systems tend to do where two pipes join. Effectively it looks like the outgoing blast from each exhaust pulse is intended to venturi / siphon the other header pipe whilst keeping header diameter constant all the way to the muffler, also halving the resistance to flow by using both mufflers simultaneously. This may not be what's actually happening though. I've dredged up a few old photos, hopefully these show what the cross looks like.
It's quite different to modern exhausts. Most current performance V-twins use a two-into-one-into-two configuration, with Y joints and pipe expansion. Headers appear to be equal length - curled if necessary - and the Y joint angle is fairly shallow, not 90 degrees. Ducati themselves have moved away from the cross, finally.
I've spent a few days reading The Scientific Design of Exhaust and Intake Systems, Philip H. Smith. The book's rather dated but good. There were some key points I took from it concerning this particular exhaust system:
1) The whole system will pressurise every time there's an exhaust pulse, from header right through to tailpipe. The piping gives a flow resistance, the silencers as well, plus there's pressure wave behaviour to consider.
2) The waves can be around a third of an atmosphere in pressure, both positive and negative magnitudes. Negative pulses in the exhaust help to start to fill cylinders when there's valve overlap, positive pulses on both induction and exhaust sides help to tamp cylinders extra-full just before inlet valves close. The effect on tuning can be pronounced and timing is critical.
3) Getting accurate charts of these waves (pressure vs crank angle) could be done in 1967 or earlier, via a very clever arrangement of crank-driven disc valve, single pressure tap on exhaust piping, and around 32 mercury manometers being fed one by one off the disc valve.
4) The cross may not be functioning as a one-pipe-helps-the-other venturi properly. Smith was very clear that pressure pulses can go around corners, and certainly any pressure pulse coming back from the silencers and tailpipes would go to both cylinders.
5) The different header lengths may have been an attempt to compensate for the different dwell intervals inherent to V-twins in the exhaust timings. This would work perfectly at one engine speed, although it might be possible to broaden the response over a fairly wide range of RPM via tricks with conical megaphones (visible on the front of the silencers, with parallel perforated pipe inside) and tuned interference or resonance with the non-flowing exhaust header.
6) If there are differences in exhaust vacuum at the start of the induction cycle, when the valves overlap, this would directly affect fuel/air intake and thus carburettor tuning. The two carburettors won't be tuned the same.
With this last in mind, I connected the vacuum gauges to the inlet manifolds and tried winding the damping valves all the way out so that the needles tracked variations in vacuum, not the average. This very clearly showed that the vertical cylinder has a wider variation in vacuum than the horizontal, at least at idle. The gauges couldn't respond quickly enough to keep up with variation at higher RPM. I swapped gauges between cylinders to confirm. I tried photographing this: the vertical cylinder's gauge is the lower one in the photo. The needle motions are blurred but there's an impression of the angles they have both swung over.
I've been wondering why the vertical cylinder keeps on fouling spark plugs, while the horizontal comes through just fine. It's been doing this the whole time I've had the bike. I've checked or changed pretty well everything except the exhaust system. High vacuum spikes coupling through to the induction would tend to pull fuel, maybe quite sharply. No wonder it's had problems with running rich on the vertical for apparently identical settings and timings.
While running more tuning attempts, I swapped the wide-band O2 sensor over and confirmed the effect. It's a difference of at least 1 full AFR point. This will only show up if the sensor is connected directly into a header, it wouldn't show on tailpipe sensors since the mufflers are shared and thus carburettor tuning becomes averaged out.
It now looks like I have to road-tune one cylinder, swap the sensor, tune the other, then confirm that tuning on the first is still OK.
Bookmarks