Member postings for S K

No one is suggesting that Q should be a constant. But it changing by factors of as high as 2 over relatively short time periods does not sound physically realistic to me, and it rather sounds like noise. But maybe I'm wrong?

Edited By S K on 10/08/2023 18:58:28

Duncan, yes, higher pressure and (probably) humidity should reduce Q.

But if I'm reading the graph right, a 3.5% atmospheric pressure change is plotted against a 100% change in Q.

SOD, I know you believe you are measuring Q, and you are getting "numbers," but I don't believe you are actually measuring Q. The old "garbage in, garbage out" problem, to my eye. If your data was Gaussian and well behaved, and if the obtained value of Q was stable, maybe, but at this point none of that seems evident.

Also, I am not sure you are using thousands of samples. Sure, you are collecting thousands, but per Q measurement you are throwing nearly all of them out after selecting only a few (I don't know the details, however, so I could be wrong on this point).

In addition, you are subtracting two numbers that are very close together from each other, getting a very small number (i.e., 6 or so orders of magnitude smaller), and then dividing that very small number into a comparatively large one again, tempting the gods of mathematical fate. In this scenario, minute deviations can cause huge impacts on the end result, as it seems you are seeing.

Try using the decay method to check. It can't be that hard, others have done it, and the decay method is much more intuitively related to the loss of energy per swing anyway. And also, the value of Q obtained this way should be quite stable from trial to trial (as I believe one would expect from a macroscopic pendulum of this sort).

Edited By S K on 09/08/2023 23:02:19

SOD: Following the discussion regarding Q in a different recent thread: If you are using this data to compute it, caution would be advised. The distribution is not remotely Gaussian, it's bicameral (or even tricameral?). Therefore, while a calculation of standard deviation is still mathematically possible, I do not think it will be particularly meaningful in the context of a calculation of Q for a system where a normal distribution is expected.

My advice would be to verify the accuracy of your calculations of Q using the un-impulsed decay method before torturing yourself trying to understand how it wanders.

I also suggest your focus should be on why the distribution is so dramatically non-Gaussian, especially why there are two huge lobes some distance apart. I'd bet you would feel it a breakthrough if you can figure out why.

All of the different swing-counting methods, e.g. "count swings until X% then multiply by constant Y" should yield the same results (I haven't checked, though).

It wasn't said, but I presume SOD's measurement was in a partial vacuum? I doubt that a pendulum's Q should change by 50% just due to temperature. I'd think Q certainly could change that much if the air pressure was changed, though. Barring that, I'd rather suspect that instantaneous measurements of Q done that way will be noisy. By comparison, the counting methods have averaging of many hundreds of swings built-in.

A concern about measured opto-interrupter performance is that one can't quite be sure it's just the opto-interrupter being noisy, or how much more potential there may be with an improved one. Unfortunately, the Sharp module does not provide access to the photodiode, but only to the digital output.

So, I've ordered a silicon photodiode and a laser module with which to do a few tests. The photodiode is quite small, only 1mm across, and it has a low capacitance and hence a high speed . I'll probably reverse-bias the diode and just look at the signal across a load resistor on a scope.

It only cost about 100 times that of a Sharp opto. 🙄

Anyway, the parts should be here in a few days, and I can use my gravity pendulum as a test subject.

I'm irrationally excited! 😄

Count the swings to 36.8% ... and then multiply that by 2*Pi. (Sorry, I left that out somehow.)

ChatGPT and its cousins are "large language models" and are not AI's. The "intelligence" they exhibit is purely emergent; an amazing side-effect of training them to write.

Still, judging their "intelligence" is amusing, given how well they can do. As one to write a college essay on a popular topic (one with a lot of text about it in its training set), let's say the American Civil War, and it would likely pass. Indeed, it may likely do considerably better than the average college student. Certainly its English and grammar, punctuation, etc., will be better than average!

But ask it to write an essay about something that is new or obscure, say "write an essay about how LK-99 works" (the new, highly controversial "room temperature superconductor" ) and it will likely fail because it had little or no training data on that subject.

Asking them "trick" questions will also often fail, since they have no in-built reasoning skills at all, they just predict text. And we've all seen how confident in their own opinions they can be, even when blatantly wrong: their training data was full of very authoritative and confidently written text. The obvious holes in their abilities, such as a lack of math skills, are being patched by allowing them access to other tools.

All that said, these large language models do at least feel a little like general AI, at least when they aren't goofing up.

Edited By S K on 08/08/2023 19:13:12

I measured a Q of about 18,500 in my genuinely-free pendulum (no restoring power input). The pendulum was about 24" in length, had about a 1 lb bob, and it was rocking on knife edges:

Gravity pendulum link

I measured it by counting the number of swings until the amplitude decayed to 1/e (to 36.8%) of its original value. I used a video camera to capture the amplitude of the swings, with an engineer's rule behind the bottom tip of shaft.

In a vacuum, Q will go up dramatically, since air resistance becomes a non-issue and essentially the only remaining loss of energy is in the hinge. The value of achieving high Q has been debated, but I have faith that obtaining high Q promotes high performance.

What you want is a small FPGA with a compensated or controlled reference clock. You should be able to reach say 150-500MHz depending on the FPGA's specs. But FPGAs can be a pain to program.

I was thinking about the resolution of timing. If the typical noise on period measurements taken via an opto-interrupter is 5us RMS, then you have to average something like 2500 readings before you approximate the resolution of a 10 MHz system clock (5E-6/sqrt(2500)=100E-9).

Actually, I vaguely recall that the resolution from randomly sampling an interval t is t/sqrt(12). If that's the case, then you have to average 30,000 samples, i.e. over 8 hours.

Or something like that - I've forgotten more math than I've ever remembered. But if true, it seems that there's a lot of payoff to building a better opto system, especially if you are interested in short-term effects.

Edited By S K on 05/08/2023 23:21:17

Oh, yes: The cool kids are right!

How cool is it to look at your cell-phone and know what the time is straight from your free pendulum clock, instead of submitting like a blinkered slave to whatever Apple says it is!

Plus, you can obsessively monitor its performance from almost anywhere in the world.

A win-win all around! Cool!

So I have a timing system running now, albeit a sketchy one. At the moment I am using a mini 16MHz Arduino. I'm getting the usual ~62.5ns resolution from that, but I intend to replace it with something faster down the road (I have a 120 MHz stand-in on-hand).

I doubt that the Arduino keeps particularly good time, so I've attached an auxiliary RTC (the popular DS3231, temp. compensated to a decent 2 ppm from 0 to 40C) and a separate temperature sensor (MCP9808, typical 0.25C accuracy).

Anyway, my question: I can think of several very different ways to do timing, but people here have already tried many things and learned a lot in the process. What is the current wisdom on the best approach? Thank you.

Edited By S K on 04/08/2023 01:56:59

I mean that if you deliver period measurements, for example, instead of trying to send optointerrupter on/off signals, then latency should not be a bother. Of course, that means that you must be satisfied with the way those period measurements are made, and that may require another module if you want to do it in a particularly-accurate way, but many such modules are also postage-stamp sized. (BTW, I am not promoting ESP32's and have never used one; it's just an example.)

Just curious, though: What sort of case are you imagining that has so little room? I'm making some progress towards mine, but thus far I'm using only an aluminum profile as a "case," i.e., just a mounting scheme. So I'm also considering some sort of remote instrumentation.

I second the above about measuring the latency or temporal noise in the 433 schemes before use.

The cool kids all use Bluetooth or WiFi to transfer data to apps they create that run on their phones. Small WiFi-enabled microcontrollers such as ESP32 are barely bigger than those 433MHz devices. Do a minimal amount of math and there should be no latency-related issues.

But I can see the draw of a simple "binary" transceiver.

Joseph, et. al.,

Would you have any advice about winding a coil for this purpose? E.g., what aspect ratio, gauge, no. of windings, etc? Really, anything that might help?

I have a couple of 1/4" neodymium magnets and 4 ounces of 30 gauge magnet wire on hand for this purpose (bought without much thought. I can 3-D print bobbins in arbitrary shapes.

Thank you.

Maybe it's easier to do compensation than I first thought.

I could put a spring-steel pin through the bottom of the rod, and support the bob (which will be approx a 3" diameter cylinder that is 3" high) by about 0.55" of the same 360 brass (2.05"-3"/2). I could ream both the support and bob 0.001" large, which should be a decent sliding fit that doesn't wobble. That sounds easy to do accurately.

Edited By S K on 24/07/2023 03:41:43

I looked at whether it's worth adding temperature compensation to an Invar rod. (This is after I ordered a brass tube to use as the compensator.)

The temperature coefficient of 360 brass is 20.5 ppm, while that of Invar is 1.2 ppm. So, given a 35" net rod, about 2.049" of brass (acting opposite the Invar) can be added to reduce the net TC by 100x.

But here's the problem: The compensation is so sensitive that, if you are off that mark by a rather small amount, say 0.1", you are only going to make it worse. It's not that cutting a tube to an exact length is terribly hard, but it's the net assembly that may easily incorporate an error that small. Using a lower TC compensator can help with this sensitivity, but that would require a longer rod, or you will be pushing the bob up much higher.

I don't think it's worth the risk, at least unless it can be adjusted finely, and then only with a lot of trial and error measurement over very long time periods, too.

Edited By S K on 23/07/2023 22:21:17

John: Perhaps you mean "consistent?"

I propose using a hollow (coreless) coil. I have an old relay somewhere whose coil I can likely scavenge, or else I'll 3D print a bobbin for one.

I've just bought two 1/4" by 1/4" cylindrical grade 52 Neodymium magnets for this purpose, either for use in a pair or to have an extra. (I also splurged for brass rather than cast iron for the bob.)

The picture above is approximately what I'm thinking of, with the magnet on a short stick so there's space for it to continue through the hollow coil without the rod hitting it. I've seen this approach used somewhere else, I think using a horn shaped piece of steel rather than a magnet, but I can't find it now.

My question is where the magnet should be relative to the coil at the point of the impulse (approximately). I was thinking inside it, at the center of the coil, but now I see that's wrong. I suppose needs to be while near the outside, e.g. at the entrance to the coil.

In the alternative, I don't like the idea of a "flat coil under the bob," but I do like John's Helmholtz coil approach. If the latter, I think I could use the two magnets facing out from the rod with their poles in the same direction (i.e. NS-NS), as I don't have the opportunity to put one inside the rod as John did. Maybe that's a more elegant solution.

... or maybe I'll do electromechanical. As I said, I tend to cut first and design later. 😉

Edited By S K on 23/07/2023 20:02:15

Dave, yes, I will try for the CP, assuming I can find it or calculate it well enough. But I'll want to dodge the bob, so if it's too near that, I'll move it away from it a little.

As for positioning, I'm thinking of mounting everything on a long, e.g. 40", T-slotted aluminum profile. That way, I can do the gross positioning of components along it pretty easily. But the impulse mechanism might need its own adjustment for the swing, the extent of which would vary depending on its location, etc. A little thought would need to go into that.

Also, while the bottom of the pendulum is the ideal spot for an opto-interrupter, I'd like a clean design, so I might integrate it with the impulse mechanism too, e.g. impulse coil on one side and opto on the other.

Yes, I had wondered how much the position of the impulse mechanism had to do with practical matters concerning the mechanism itself. Gravity escapements also typically impulse higher up than the CP would be.

I had a chuckle at the anachronistic mention of an "IBM PC" (IBM got out of that business eons ago, and who calls them that as a generic name anymore?). And that "frame coil" is a bit odd, too. But the article serves as a caution about how sensitive the pendulum can be to the timing of the impulse - a few ms off is too much, apparently.

If going with electromagnetic impulsing, I had imagined a small magnet at the end of a short rod, with an open coil able to accept the length of the rod past the point of the impulse (I've seen this elsewhere). A question about this: Is there a clear optimal position of impulsing the magnet relative to the coil? My first guess is that the magnet should be mid-coil at the time of the impulse, but as the lines of force should be roughly parallel within the coil, I'm not sure it should matter much once it's past the entrance.