This page is an attempt to bust some common misconceptions about Brunel's atmospheric traction system for the South Devon railway and why it failed.
One, which may be a misconception or may be just laziness, is the failure to realise that it wasn't actually Brunel's atmospheric traction system at all. It was Joseph Samuda's. He invented it, and he had the contractual responsibility to maintain the installation, but he didn't have the resources to actually sort it when it became clear that the whole length was fucked. He had experimented with it on what became the West London line at Wormwood Scrubs, which is probably what brought the idea to Brunel's attention originally, and he had done a couple of operational installations - one at Dalkey in Ireland, and one at Croydon, which was basically a demo/development system (ie. installed mainly for that purpose, although it was used in regular service), and had all the same problems as the South Devon one, only the South Devon problems got all the attention. The Dalkey system actually pretty much worked, but there were under two miles of it and it was Irish (see later), so it wasn't really comparable.
This is probably the most common misconception and it gets repeated everywhere, including on sites that ought to know better.
The usual assertion is that the tallow mixture used to condition the leather caused rats to consider it a tasty snack and they promptly munched the lot. It's bollocks. No doubt the odd rat did have a cheeky nibble; sometimes they got sucked into the traction pipe. But it wasn't a significant problem. And if it had been, it would have been a doddle to solve it - simply add arsenic to the tallow mixture, in accordance with the general Victorian love of putting arsenic in bleeding well anything on the slightest excuse. (All the things considered in Feet of Clay as possible sources of the arsenic that was poisoning the Patrician are things that people did get arsenic poisoning from in real life because the Victorians put arsenic in them. Even the really weird ones that the Watch ruled out straight away. Indeed the more weird and stupidly improbable they are and the more likely you'll think "oh, he must have made that one up", the more likely it is that the Victorians actually did it loads and did it all the time.) Rats learn which food sources are dodgy by waiting for more adventurous rats to try it first and seeing if they get ill or not, so a few dead rats equates to a lot more rats deciding to leave it well alone; this makes poisons kind of shit (well, more shit) for the purpose of getting rid of rats in an area, but also means they do work well for putting them off eating specific things.
The above is not to say that the leather valve was problem-free. On the contrary, it was a pain in the arse, and the need to replace the whole bloody lot and the hassle involved in doing so constituted the major excuse for taking the final step and junking the whole system.
The basic problem was that leather as a material just wasn't up to the job. They used it because it was the best available material at the time, but that "best" wasn't good enough. Why Not Use Vulcanised Rubber, you may ask. Answer, because they didn't have it. Its development period pretty much coincided with the period the atmospheric system was in use on the line, and when they were putting it together there was no such thing as vulcanised rubber apart from experimental laboratory samples which were shit (burnt on one side and gooey on the other, with just enough good stuff between the two to show that it would be useful if you could figure out how to make it all go like that). The abovementioned valve replacement plan was going to use vulcanised rubber, which by then had reached a useful state of development, but it was too late and nobody (including Brunel) could be arsed with it any more.
In hot weather the valve would dry out, become inflexible, and crack along the line of flexion where it was clamped to the iron pipe. The "remedy" was said to be "an immediate application of seal oil", but of course this could never do more than slow the progression of the damage. In cold weather, it would soak up water, freeze, and again become inflexible, with the same result. The cracking would then develop into longitudinal tears, which made it leak like a bastard and ruined it. Such conditions also of course caused immediate stoppage from the whole thing just freezing solid. Simply being repeatedly doused in sea water didn't do it much good either, of course.
Most seriously, chemical reactions between iron oxide from the hardware and the tannins in the leather in effect "de-tanned" the leather and allowed it to rot. Again this action was most pronounced along the line of flexion, and the decomposition of the base material rendered all attempts to preserve its qualities with fatty compositions ultimately futile.
The Dalkey system had, in addition to the leather flap valve, a hinged iron flap over the top of it to act as a weather shield. It wasn't completely effective but it did make a major difference. The South Devon system, on the other hand, was built without this iron flap, partly because of the huge inertial forces to be contended with in flipping it out of the way at the high speeds envisaged, and partly to cheap out. The Croydon system didn't have it either, presumably because it was only a demo and they didn't think it was worth the extra effort. Accordingly, both the South Devon and the Croydon systems suffered the same weather-induced deterioration, only the South Devon one had it worse because the Croydon one didn't have to cope with being inundated with sea water in the winter.
The other major weather problem concerned the method by which the valve achieved a seal. Mere contact between the leather valve and the metal surface not being good enough, it was necessary to improve the seal with some kind of grease. This had to be fluid enough to make a seal and to allow the seal to be broken easily by the mechanism on the train, but also solid enough to maintain the seal and not all be sucked into the vacuum pipe; and it had to retain this behaviour over the full range of operating temperatures from summer to winter. It didn't. They rang the changes on the mixture of fats and soaps but never got anything better than an unsatisfactory compromise. (It still wouldn't be a straightforward problem to solve even with all the subsequent developments in grease chemistry that we have now.)
The early proposals and experiments tried to get over this problem by using radiant heat from a charcoal burner underneath the train to melt the grease ahead of the piston connection, and then allowing it to re-solidify after the train had passed. It did very little to widen the operating temperature range, and of course was no good at all at any speed over a crawl. The Dalkey installation apparently had it, though I doubt it really did anything; the South Devon installation, being always intended for fast running, did not use it.
Finally, it is worth noting that even the minutest deficiency in the seal, such as you would never even notice in most applications, nevertheless when integrated over several miles of pipe will easily add up to a much larger passage for air than the pipe itself is.
It's more or less the other way round. The route was decided, the plans submitted to Parliament, and the Act obtained, BEFORE anyone had raised the idea of using atmospheric traction. The Act got the Royal Assent in July 1844; the plans at that time were, per Brunel, for a double-track railway using conventional traction. It wasn't until August that the Board of the company began to discuss the possibility of using atmospheric traction, and the decision to actually do it was made in September.
Brunel's idea was that since the route would inevitably have to gain and lose a lot of altitude, it was better to concentrate the necessary gradients into specific sections which could be worked under their own operating arrangements, thus allowing the majority of the route and the engines for it to be designed for express speed without the compromises called for by the introduction of gradients. This is of course the principle on which he had laid out the Great Western main line with the Box and Wootton Bassett inclines. Similarly his route for the South Devon Railway was laid out, from the start, on the basis of four "inclined planes" and two long stretches laid out as best the landscape would allow. It was always the plan to use some form of assistance over the "inclined planes", and originally this was to be by means of cable haulage (as used by the L&BR on Camden bank).
When the decision to use atmospheric traction was taken, it was seen as an opportunity to relax some of the specifications for gradient and curvature and modify the alignment accordingly to reduce construction costs. However since a route had already been decided and approved by Parliament, it could only be modified slightly. The original route still served all the same places and of course had the four "inclined planes". Unfortunately I haven't found any more detailed description of the exact differences than one very brief summary of the chief gradients, which although it was given with the intention of showing that the original route was easier, in fact falls nicely within the range of average gradients different people quote for the route as built, so if it means anything at all it implies that it wasn't significantly different.
This is a misconception from which the Victorians suffered too, but it is hard to see why, because it only takes a trivial amount of mental arithmetic to show that it is bollocks. If the system ever had got beyond Newton Abbot, though, it would have hit them in the face. They were running trains of up to 120 tons along the level stretch where they did install the atmospheric system. If they had tried to take those over Dainton they would quite simply not have made it: they would have ground to a halt, and only dividing them into several smaller trains would have got them going again.
Pretty well no popular accounts mention this and few "serious" accounts do. It is far too common to see statements to the effect of "atmospheric traction makes light of gradients" with no mention of the limited circumstances under which this is valid. The advantage of atmospheric traction on gradients is that it is not limited by the point at which the driving wheels lose their grip on the rails. But it is limited by the product of the piston area and the pressure differential across it. This, in practice, is a much more severe limitation. Moreover, there is no possible remedy other than to rip out all the stationary plant and replace it with something bigger.
The system as installed was operated at a depression of about 0.6 bar, and Brunel specified a 22-inch diameter traction pipe for the inclined sections. This gives a maximum static tractive force of 1.5 tons, which corresponds to the static gravitational resistance of a 55-ton train on the 1:36 maximum gradient of Dainton bank. Even if it had been possible to achieve the theoretical maximum 1 bar depression the figure is still less than 92 tons. And since these figures neglect all frictional and other resistance, the practical limit would be less.
In short, if they had got as far as trying atmospheric traction over the banks, far from making light of them, it would have been really really shit.
It did work, eventually. When they finally got it into service on the level section it seemed to work quite well, at least in terms of impressing the passengers with the speed unaccompanied by engine noise or showers of hot cinders. It just didn't stay working for more than a few months.
As we have seen, what finally did for it was the mechanical and chemical failure of the longitudinal leather valve. It needed completely replacing over the entire length, and being unable to get Samuda to pay for that they decided to bin the whole thing off instead. But that was only the final straw following a lot of other problems and difficulties.
In Brunel's original proposal of the idea to the South Devon board, he starts off by stating two axioms:
How he felt able to state the second of those eludes me since it patently wasn't true. It was only "proven in service" in the little Dalkey installation. The London & Croydon was still a work in progress, and indeed Brunel was trying to leave it as late as possible to finalise aspects of the South Devon design so that he could watch how well they worked at Croydon and modify the design if necessary.
With the first, though, he was simply rather unfortunately far ahead of his time. It's a pretty safe bet that a stationary power plant can be made more efficient, more reliable, less demanding on maintenance, etc, than can a locomotive version, in large part because the lack of restrictions on weight, size and layout makes many enhancements possible which you simply can't fit on a locomotive and still make them useful. The contrast is particularly vivid with a steam plant, where a stationary plant can achieve the highest efficiency of any kind of heat engine, while a locomotive is among the lowest that are still practically useful. And by means of electricity it is now possible to transfer a good two-thirds of the mechanical output of a stationary steam plant to tractive force at the locomotive wheels.
You also, of course, save most of the weight of a locomotive and have it available for payload instead. This was of particular moment given that with the low power-to-weight ratio of locomotives at that time, it was quite possible for half the weight of an express passenger train to be in the locomotive.
(It is tempting to conclude at this point that "all Brunel needed was electricity, haha", and I have seen one writer continuing to the ludicrous elaboration that he perfectly well could have used electricity but he had a bee in his bonnet about it (as supposedly shown by the GWR being slow to make serious use of the telegraph). Bollocks, say I, to both assertions.)
In Brunel's time, though, not only were stationary steam engines not more efficient than locomotives, but the direction of development was towards improving the efficiency of locomotives more than stationary plant. At that time stationary installations were still happily using crude boilers designed for simplicity above all else at very low steam pressures, and compensating for any deficiency in power output by making the whole thing bigger and shovelling more coal in. On locomotives, by contrast, the need for a substantially better power-to-weight ratio in order to make the things useful on any sort of scale, along with the desire to avoid hauling around unfeasible quantities of fuel, compelled the use of measures which conduce towards greater efficiency, such as multi-tube boilers with induced draught, jacketed fireboxes, a steady trend towards higher steam pressures, improved valve gears, and attention to details such as adequate sizing of steam passages and valve openings. It would be a while yet before stationary (and marine) engines began to catch up.
[ Page edits in progress ]
Brunel cocked up the calculations for sizing the pumping engines. Brunel was a very successful engineer in civil and structural matters; he was great at static things, but he had an unfortunate tendency to drop a bollock when anything dynamic came along (perhaps most notoriously he couldn't design locomotives, but at sort of the opposite end he also failed to realise that track is a dynamic structure, which is a bit odd as he really should have spotted that one.) I can get similar figures for the required pumping engine power by doing very simplistic calculations in my head from the only readily accessible "results" figure for the system output (force x distance along the track) - and I know they are wrong, because they ignore everything I don't have figures for; all the stuff like leakage, pump inefficiency, engine manufacturers quoting higher output figures than you actually got, and every other source of inefficiency, inaccuracy and loss.
So the engines installed were undersized, and could not cope with the task they were called on to perform - unless they were thrashed. Of course, one of the great things about a steam engine is how well you can thrash it. But doing so absolutely canes the thermal efficiency, which is ropy enough already. So it is not something you want to be doing as a matter of course. Thrashing a steam locomotive to overcome a heavy gradient is OK, ish, unless you're the fireman, because it's a temporary expedient, but using a stationary steam engine in thrash mode all the time just makes it use stupid amounts of coal, and this is exactly what happened.
And of course when they weren't pumping they still had to burn coal to keep them ready for action. Since the proportion of idle time was vastly greater than for a locomotive, that too resulted in disproportionate fuel consumption.
They never sorted out the telegraph. There should have been a telegraph installed in all the pumping houses, so they could let them know when a train was due and they needed to start pumping. But they fucked around and pissed about and never got their fucking act together to install the bloody thing. So the pumping houses just had to start pumping whenever the timetable said a train was due, and then carry on pumping until it actually turned up. This of course led to even more waste of fuel when the trains weren't running to time.
A significant part of the justification for installing the atmospheric system in the first place was that it was expected to be cheaper to run than locomotives. Quite why they had any confidence in this, given that there was plenty of data to estimate the running costs of locomotives but fuck all for atmospheric railways so the figure was basically guesswork and wishful thinking, is obscure, but that's what they did expect. In the event, with all the thrashing and idling and redundant pumping, it ended up costing about three times more than locomotives to run. (How it would have worked out if everything had worked as intended is again guesswork, but I strongly suspect it would still have been a lot worse than locomotives; if nothing else, a locomotive connects the steam engine directly to the wheels, whereas atmospheric traction interposes a lossy and inefficient method of power transfer. It would take a lot of efficiency advantage of stationary engines to make up for that - and at that period stationary engines hadn't started taking advantage of their stationariness to incorporate efficiency measures which aren't practical on a locomotive (apart from condensers), so unlike the situation obtaining today, where stationary steam plants are the second most efficient (behind ship-sized two-stroke diesels) heat engine in use, they were even worse than a locomotive. The boiler pressures, for a start, were ridiculously low, whereas locomotives at least were trying to get them up, even if that was for reasons of power-to-weight rather than efficiency at that time.
So what with the grossly excessive running costs and the expense and disruption of replacing the entire length of the flap valve, they ended up deciding that it was far more trouble than it was worth and got rid of it.
Anything other than plain running was a pain in the arse. The atmospheric system can't do two very important things: it can't go through pointwork, and it can't go backwards. So while it's fine for tanking along plain track hand on cock and mind in China, it can't do any kind of shunting or marshalling operations, and it can't put a train in a loop part way along a single line section (which it all was) to let another one come the other way. For shunting, they had an auxiliary traction pipe in stations that ran alongside the track with a piston that you could tie on to the train with a bit of string. That let you pull one way, but not, I think, the other; or at least not in any remotely practical way. Regardless, there still was plenty of "other", and for that it was either use a horse or shove it along by hand. (Either of which, I rather suspect, would probably have been less fucking around most of the time than using the pipe. Shunting horses, after all, remained common long into the 20th century.)
For two-way working, they were just plain fucked. They had to run trains through in one direction, then reverse the whole system and run trains through in the other direction. There was just no way to send trains through in both directions at the same time and stick one in a loop to let the one coming the other way go by.
Perhaps surprisingly, they did manage to sort out a way of handling level crossings. Something or other went up and down to bring the road surface and the top of the traction pipe to the same level when the crossing was set for road traffic. I'm not sure quite what, but I'm pretty certain it was some kind of movable ramps in the road rather than having the pipe itself go up and down, because making that seal properly would have been an utter cunt.
The Dalkey installation in Ireland was not the abject failure that the South Devon was. It ran for 11 years, and was torn up not because it didn't work, but in order that they could convert the track to standard Irish gauge and connect it to the rest of the rail system.
So what was different? Well, for one thing, it was much smaller. It was less than two miles end to end, and didn't stop on the way. So it was vastly less of a burden to operate, with only one pumping engine, and not a stunted one either. And it had the weather protection flap over the valve so deterioration of the leather was much reduced.
And they were Irish, and didn't care about it having shit bits. The traction pipe did not go all the way. It stopped about 500m from the end. Trains arriving at that end would coast from the end of the pipe to the end of the line. Trains departing had to have some poor sod push them by hand until they got to the pipe. It hardly needs saying that this was epically shit, but they never did anything about it. Making sure the flap valve re-sealed properly after the train passed was achieved by having some other poor sod walk along the line pushing it down, and they never implemented a less shit alternative to that either. Irish railways were well into doing things like this. So were English ones, especially in the beginning, but they almost invariably soon started fretting over it being shit and replaced it with something not shit as soon as they could get it together, whereas the Irish ones just didn't give a shit and left it like that for ever. I suppose if you have a large enough supply of poor sods you can claim anything works no matter how fucked it is.
Incidentally, it is possible that the Dalkey system may have been where the Devonian rat myth originated. It is reported that a director and chairman of the company in its post-atmospheric days, Laurence A Waldron, did cite rats eating greased leather seals as a real cause of failure at Dalkey - only it wasn't the longitudinal flap valve, it was the annular seal around the piston itself that they went for. So maybe that story got chinese-whispered over the course of time...
Back to Pigeon's Nest
Be kind to pigeons