Flying Scotsman Speed

I agree with most of the above regarding Rous-Marten's claims, however the LNER dynamometer car underwent continuous development throughout Gresley's reign as CME of the LNER (it starting off, of course, with the NER) and it underwent regular calibration to the extent that it was used to prove that the Midland's own dynamometer car had not in fact been calibrated correctly when used behind the at the time new Royal Scots.

So whilst I agree it may not have been specifically calibrated regularly for 100mph, it was in regular use with known margins of error and for Mallard's run had been seen behind a number of A4s recording high speed runs (and of course, prior to the A4 development had been used behind a number of A10s and A3s). So the LNER were not without comparative analysis going forward for all of their high speed runs.

I would argue the more scientific approach of the LNER is likely to be accurate compared with Rous-Marten's timekeeping - but that's one man's opinion based on his limited knowledge of such practice in comparison to that he knows of the dynamometer car's work.

 

Surely the most that can be said of all these claims is that the locomotive did x mph, +/- y -- in other words, there is a speed, and then a margin either side that might represent, say, 90% confidence that the "true" speed was between those limits. The key is then looking at what the possible sources of error would be, and how much that might alter things.

For example, with C R-M and his stopwatches: I suspect the watches themselves were pretty accurate. (For example, if they were accurate to one second per day, the difference that could account for in speed would be neglible - about 1/1000 mph). However, what about the mileposts? A ten yard error in the position of one quarter mile post relative to its neighbour would make a roughly 2mph difference to the measured speed. Then there is the reaction time as he pressed his watch at each point. The reaction time itself is not important if it is consistent, but what if it is not consistent - sometimes reacting within, say, 0.2s and sometimes within 0.3s? That might mean his quarter mile times can't really be taken to better than a tenth of a second either way: another 1 or 2mph variation. So you might feel that CoT did 102mph, but with reasonable confidence that it was within +/-4mph of that. Probably faster than any other locomotive had travelled up to that point, but not definitive proof of 100mph.

As for the LNER dynamometer car: the inaccuracies would be different, but still there. In particular I wonder about the calibration. I assume there was some linkage that was converting the rolling wheel diameter and converting to some kind of output signal. I don't know the specifics, but maybe, say, an electric generator, with a device to record the output current? Whatever the mechanism, there is no absolute certainty that the response bore a linear relationship between speed and output: the fact that calibration might have shown, say, a 1% error at 50mph (which you could correct for) does not mean that there would also be a 1% error at 100mph. The best you could do would be to plot a calibration curve, and then extrapolate to speeds higher than were calibrated for.

All of which is to say that in the case of CoT, Flying Scotsman and Mallard, in all cases I am sure they went very fast - but in all cases, I'd feel more comfortable accepting a narrow range of potential speeds, rather than the quoted values that appear to give a precision that I suspect is not justified by experimental standards available.

Tom

 

I hope I didn't give the impression I thought anything else, its just that calibrating instrumentation can be a decidedly tricky task, and it would be interesting to know how it was done. CR-M only needed to run his stopwatches for a few hours against a known good source to check they were as accurate as they were going to be. The LNER team had a trickier task and desired, I am sure, far greater accuracy than the limitations of a 1/5 second stopwatch against mile posts. The GWR car got much data with an extra wheel on the track I believe. Presumably the LNER one was similar. I wonder how either company ensured that wheel didn't bounce or skid enough to affect the measurement? It seems a non-trivial exercise to me. But no doubt the actual traces would tell them a lot.

 

Tom, don't forget the stopwatch granularity was 0.2 seconds. So If the quarter mile was done in precisely 8.9 seconds then the stopwatch would record either 8.8 or 9.0, and both would be correct. Having charted all the numbers with error bars I'm inclined to think that the last 1/5 of a second can't be guaranteed, and that the last quarter was probably more likely 9.0, or nominal 100mph. Rous-Marten's system meant successive errors balanced out, so another quarter mile would have told all. The other thing a chart tells is that the locomotive was clearly still accelerating, and a reasonable fair curve drawn on the chart tops out in the 105 region - or alternatively in the great mail train accident of 1904...

 

105 would also be within my "102 +/- 3 or 4 range"

If as you say, CoT was still accelerating, why are there no subsequent quarter mile times? Did C-RM stop recording at that point, or did he record but never publish the times? Surely, given that his numbers were in the region of 9s, he must have realised the significance of what he was timing, and therefore the importance of having a complete record from before, during and after the peak speed?

Tom

 

Oh, didn't you know? Driver Clements hit the brakes. According to Rous-Marten there were platelayers on or near the track, but there's also been a suspicion there was a disinclination on the part of Clements (who of course didn't have any speed measurement) to "go and break his bloody neck".

 

Still seems strange not to publish the times for at least the next quarter or so, even if they were heading out to the 10 - 11 s mark.

Tom

 

Papyrus of course did run at over 100 mph for 12 miles, so while the peak was 'approximately' 108 it did, without any doubt whatsoever, exceed the 'ton'.

 

Can somebody with knowledge of the log clarify, when Rous-Martens refers to five successive quarter miles, does that mean over 1.25 miles total distance covered, or five successive timings of, say, alternate quarter miles?

I've been doing some rough calculations on required power output (shamelessly borrowed from Wardale's 5AT fundamental design calcs) and the numbers are quite large (1750 dbhp, 2500 ihp and 90psi bmep). I can set them out in full if anyone's interested, but it would take some tidying up.

 

there are 8 timings - consecutive from milepost 173 (first after tunnel) to 171
tare weight about 90 tons, gross about 115 tons

What happened was this: when we topped the Whiteball Summit, we were still doing 63 miles an hour; when we emerged from the Whiteball Tunnel we had reached 80; thenceforward our velocity rapidly and steadily increased, the quarter-mile times diminishing from 11 sec. at the tunnel entrance to 10.6 sec., 10.2 sec., 10 sec., 9.8 sec., 9.4 sec., 9.2 sec., and finally to 8.8 sec.

 

There was never a speed trace. The dyno car made marks every second on paper moving at 12"/mile (24"/mile for Mallard). This was then interpreted and plotted by hand on paper as a speed distance graph. It appears that for FS, Papyrus and Mallard the interpretation became "optimistic" around the maximum speed. I believe a trace survives of the original dyno car roll for Mallard - but not for FS and Papyrus.

 

Presumably people (outside the LNER) saw the logs at the time? Did anyone question them back then?
I am always highly unconvinced by arguments which suggest new interpretations of old data without new science. People aren't any cleverer now than in the past. We just have the benefit of a few more years of data and experimentation.

 

Yes - page 383......

http://www.gracesguide.co.uk/images/8/8a/Er19350412.pdf

 

Interesting and well reasoned letter! Out of interest, was the author (J.S. Dines) an "interested amateur", as it were, or was he involved with one or other of the railway companies in a professional role? Is the name a familiar one? (Obviously it isn't to me, hence the question!)

Tom

 

Fascinating clip. I must confess I was more interested in the turbines and that Auto Sync panel than the railway stuff having spent many years sync'ing alternators and systems.

 

have a look at page 8, I think that's him

https://www.rmets.org/sites/default/files/hist08.pdf

 

Must admit 3440 was possibly the punchiest beast out of the blocks of any kettle I have had the luxury of riding behind. I think it might 'ave, there was a rerun which proved it was theoretically possible at least?

 

In 1934 Cecil J Allen was on the Flying Scotsman run with his stop watches. He disagreed with the LNER publicity machine with the results. Using JimC's reasoning the stop watches were perhaps more accurate than the Dynamometer Car.

The blips at milepost 91 on Stoke Bank have never been properly explained. Modern computer simulation would perhaps help.

Cheers,
Julian

 

I absolutely do not accept that human reactions in the use of stopwatches can be in any way more accurate than a regularly used, tested and calibrated dynamometer car.

One humans reactions will not be as good as another's, but the rolls recorded in a dynamometer car kept in excellent order by its company will have less margin for error - they're being run through the same machine after all.

Cecil J Allen was not infallible and to suggest his views override that of the dynamometer car suggest to me a lack in understanding of how the NER dynamometer car works and was used by the LNER.

It was their greatest asset in the development of their steam locomotives and the research was taken very seriously and up for constant scrutiny.

 

I hope I didn't quite imply that: it wasn't intended. I suspect its much easier to accurately calibrate stop watches than the dynamometer car, but I imagine the owners were well aware of that and worked on it.

But no matter how well calibrated your 1/5 second stop watch is, how well spaced the quarter mile posts are and how competent you are at physically doing the timing, there's still an absolute limit of resolution of 1/5 second (+- 1.1 mph at 100mph), and you only have data points every 9 seconds at 100mph. To be absolutely sure that the locomotive has hit precisely 100mph, for example, you actually need to record 8.8 quarter miles, because 9.0 quarter miles might only mean 99mph. Then you also have to consider the "calibration" of the quarter mile posts, and the ability of the operator. AIUI sophisticated operators like CR-M used a multiple stopwatch system which had a certain error correction capability.

Courier tells us the dynamometer car marked points on the paper every second on paper driven presumably from the extra wheel, so its an analogue mark, so the potential accuracy of the dynamometer car is far greater, and more important perhaps the resolution is much finer. The only question mark is over the accuracy of the calibration of the car, but I am sure that was just as obvious to their engineers as it is to us today.