Mechanical Spectrum Analyser

[mark_in_manc]

Hi Folks

I thought one or two people might be interested in this - a mechanical spectrum analyser. I didn't know such things existed, before I came by this one - not knowing what it was, but thinking it was interesting. Once I'd assembled it, slowly I worked out what it did.

The idea dates from the 1890s - considerably earlier than the first Fourier narrow-band analyers which appeared in my field (noise and vibration) in the late 1970s!

If people find it interesting, and it is not too off-topic on an electronics forum, I'll try to explain how it works.

cheers
Mark

[dinkydi]

Mark, the floor is yours. I, for one, am very interested.

Peter

[6AL5W-Martin]

amazing !

Compliment, this is a greatful unit.

greetings
Martin

[mark_in_manc]

You're very kind, Peter.

This is one of the most exciting skip finds I ever came across - all the more exciting because frequency analysis is a big part of what I do.

The device assumes you can plot one period of the waveform (time-history) you want to analyse, on a 360mm timebase - 1mm for each degree of phase of the fundamental. Presumably capture, if not analysis, of audio was a little tricky (!) - but if your data was for example tide levels (and Lord Kelvin invented one of these a whole lot bigger, for just that purpose, a little earlier) then data capture, with a clock and a ruler, might not be such a problem.

The device has two planes of motion. Left-right, a pointer moves along a carriage along the front of the instrument, and as it moves, a taught wire communicates this motion to each of the 5 pulleys on the top. Since they're different sizes, the biggest undergoes one full revolution, the next biggest, two...and so on.

Each rotating device (there are 5 on this one - others were apparently made with one, or three, of these 'ball and wheel integrators') records the amplitude of one harmonic of the wave. One of the photos shows how there are two 'readers' on each one, placed at 90 deg to each other. These record the amplitude of sin and cos components for each harmonic - and since the sin and cos readouts can be positive and negative, then the vector representing each harmonic can be 'steered' right around the argand diagram. This is just like the way Fourier analysis is normally taught in its 'simple' (trig) form - the sum of a_n [sin(nwt)] + b_n [cos(nwt)] - for those with that kind of background.

As the pointer moves left-right, the phase angle made by the readers to the ball on which they operates changes, but they don't actually read anything - like spinning an old ball-based PC mouse on its axis, but not 'pushing' it anywhere. The 'pushing' occurs as the device is bodily pushed up and down (in the direction of its shorter side - once you've taken the brake off!) when it rolls on knurled wheels over your chart/table. This spins a long bar under the whole thing onto which the 5 white wheels which can be seen in the pictures, are pressed. These make each of the 5 glass balls, rotate - *just* in the plane of the short side of the device.

So - as it rolls in the amplitude axis of the graph, the readers read up and down. As it moves LR, the phase made between the readers and the glass balls changes, implying that up-down motion will be picked up 'differently' in each sine and cos component.

As even I am now bored with the sound of my own typed voice, perhaps we could make this a game - using the easiest example of a square wave with 0 DC offset and an even mark-space ratio, can anyone work out how the device reads sin components for odd harmonics only? Hint - sine readers start out facing you as you look at the front of the instrument, when time = 0...

I can see how it works in my head, amazing. I presume one can change the 'frequencies' measured by scaling the curve traced by the pointer in the X direction. It's a super demonstration of parallel processing.

[Andrew2]

Mark, that's absolutely astonishing. I never knew anything like that existed! And look at the beauty of the engineering

[elanman99]

Its beautiful!

An incredible skip find, I'm going to start looking, where is the skip?

Its made me really curious, I presume it was not commercially sold, maybe it was made for, or by, a university. Are there any identification marks etc. It looks in the pictures to be about 40cm wide but how big is it?

Will it be on display anywhere? I would love to see it.

Ian

[peter_scott]

Hi Mark,

What a fabulous find! Thanks for posting and the explanation. I think you really must make a YouTube video of it traversing your square wave example and showing the results.

Peter

[Radio Wrangler]

Hi Mark,
When I read the title and clicked on the thread I was expecting an old spectrum analyser with a motor-spun butterfly capacitor.

Then I saw the date.

That machine is staggering. A 5th order analogue computer implemented mechanically. The whizziest spectrum analysers these days are optical jobs and they use a motor-positioned diffraction grating, so they qualify as opto-electro-mechanical machinery and would be great on a triple word score.

Peter Scott, Edinburgh? Unless there's another one you look a bit different since the last time we met Do you remember IMM's efforts with the HP9825 calculator doing plots to explain the Fourier Transform to his kids?

Cheers
David

[gingpeakin]

Hi Mark

Your find reminds me of this:

http://www.tatjavanvark.nl/harmonium/harmonium.html

Ging

[mark_in_manc]

Thanks for your interest, folks - I'm glad you like it.

I'm not sure who made it, but if you go web searching for 'Henrici Coradi analyser' you can find out more about it. This one looks very, very similar to the originals made by Coradi, a Swiss - but the dials look like a kind of white plastic (they're not bone or ivory), which suggests perhaps a later date.

If anyone in the NW wants to come and see it, I'm sure we could set that up. Perhaps you could bring a favourite waveform for analysis...

It was a university skip. I strongly suspect some administrator told some poor porter to 'get that ****** office (or lab) cleared' ... and they did. The time for really interesting finds seems to be mostly over - all the 60s and 70s stuff has gone (teak benches and all), and now we're into melamine, chipboard and old PCs. The 'stuff' of science and engineering is not what it used to be, which is a shame since it was 'stuff' which got me into it. Indeed, this and a lack of 'internationally-leading thoughts' might well soon get me out of it...

I'm a bit crap with a video camera - it took an hour to get those 4 photos sized and up! I've leave the sq wave example up as a challenge, and try to get someone at work to help me with the video, which I'll reveal when someone gets the answer right.

[mark_in_manc]

Sorry to double post - Ging, I found that site a while back. Ms Vark seems like a seriously unusual character...

[turretslug]

Fascinating. And to think it was in a skip. Sacrlege!

[woodchips]

Amazing piece of engineering, out of a skip!

Ball resolvers on their own are rather interesting devices, much more so than the disc and ball variety.

If you are interested in the theory try Practical Geometry and Engineering Graphics by Abbott. When I first came across this, doing a Fourier transform using a pencil and paper, I couldn't believe it, thought you needed DSPs and similar. There are other books that cover the same principles, but this is a paper and pencil job, so simple, so obvious.

Bob

[Avonauris]

Wow! That's fantastic peace of equipment.

[mark_in_manc]

Originally Posted by woodchips

Ball resolvers on their own are rather interesting devices, much more so than the disc and ball variety.

Do you mean wheel-and-disk? There's one of these latter in the Manchester science museum, which I think formed part of an (airborne) WW2 bomb sight.

Originally Posted by woodchips

If you are interested in the theory try Practical Geometry and Engineering Graphics by Abbott. When I first came across this, doing a Fourier transform using a pencil and paper, I couldn't believe it, thought you needed DSPs and similar.

Sounds good. Sometimes I tell students all you need is a knowledge of the orthogonality test (int f(x)*g(x) dt = 0 unless f=g for orthogonal f and g), and some squared paper. They like the latter straight away, and the former is OK once they can spell it...

[woodchips]

The disc type integrators used two stacked ball bearings between the disc and roller. The reason for two rather than one is to stop any scuffing when the ball unit is moved across the disc. These were commonly used in all sorts of navigation equipment, look at the various air and ground position indicators. Problem with the disc type integrator is that at the extreme ends of movement the multiplication of speeds gets too great and it ends up slipping. These old navigation units weren't usable above about latitude 70-75 degrees. The ball integrators solved this problem and would work all over the ball's surface, interesting things. Bob

[mark_in_manc]

Than ks for that - it's interestng. I don't know anything about mechanical navigation systems, though being interested in vibration measurement I was interested to read about attempts to measure the acceleration of (say) an aeroplane (using, for example, piezo-electric accelerometers) and then electrically double-integrate it to get displacement.

With vibration problems, one can happiply do this (once for u, or twice for x)and roll off at half a Hertz or so, or perhaps even higher. But for navigation, if you want to know where you are, I guess it has got to go down to DC, reliably. Problems with drift must have been incredible.

My device skids a bit, but I think the wheels on the readers could be slipping because my balls are a bit dusty through lack-of-use. I'll have to get them out and hit them with some Solvol

[Dekatron]

Fantastic machine, I am happy you shared it!

I almost thought it was a bogus/fake machine somewhat like the Swedish Fatilary Calculus which was presented as doctor's thesis in 1955. Many people believed it to be true but it was fake from the beginning to the end.

[mark_in_manc]

Martin - that sounds like something worth a google! The original paper on Cepstral analysis also sounds like a p*** take - quefrencies, liftering etc etc - but isn't (though students often think I'm taking the p*** if I try to teach it to them, and sometimes I'm inclined to agree).