Open source timing software.

To install from source is pretty straightforward on a mac - using intel (x86) hardware. I can't speak to M silicon, but it should be the same. I'm on Ventura.
Git way:
Assuming you have git, fftw installed find a place to install tg-timer. I have a folder for git stuff. go to your folder and ...
you can also download the zip file from github, unzip to a folder, then open the folder in your terminal and start from ./autogen.sh
I hope that helps.

Ok, so I have a questions or two about tg-timer and it's a bit meandering, so forgive me
I have a problem with TG. The beat error and amplitude seem to jump all over the place. I'm using one of these guys as my microphone
might not be the exact thing, but close enough. It's the USB-C version, not the version with analog out (annoyingly)

I thought this might be the issue, so I listened to the output directly, and A) it's super loud, B) there's a 1khz (ish) tone that breaks in intermittently, like microphone feedback.

So, my questions are... Is this mic just fubared? Can I trust the results of TG with this noise? Does the high output of the piezo mic mess up TG?
A feature request for TG would also be: can we add a noise gate if that's an issue? Seems like an FIR noise gate is possible, but beyond my capabilities. http://www.jatit.org/volumes/Vol100No15/6Vol100No15.pdf

Ok, now to get a little more meandering ...
Here's a couple screenshots of TG

These two screenshots were taken within a couple seconds of each other, you can see how much the beat error and amplitude jump around.
What could cause that? Is that my mic being too loud or that weird 1khz noise?
Follow up, so I understand what TG is doing ...
How is it calculating the amplitude and beat error? I'm assuming these waveforms

Show the"tic" and the "toc" of the escapement.
What is the blue line showing? How is it setting what's at 0 and what's at the blue line?
this software is really cool, but it would really help me better use it if I understood what it was doing.

Thanks!
jrubins

jrubins said:

might not be the exact thing, but close enough. It's the USB-C version, not the version with analog out (annoyingly)

I thought this might be the issue, so I listened to the output directly, and A) it's super loud, B) there's a 1khz (ish) tone that breaks in intermittently, like microphone feedback.

Click to expand...

From what I've seen, the usb audio implementation in these is not the best. The circuit is noisy and the PCB unreliable. And it doesn't use USB audio async mode. The analog output is better, assuming one has good enough computer sound.

So, my questions are... Is this mic just fubared? Can I trust the results of TG with this noise? Does the high output of the piezo mic mess up TG?
A feature request for TG would also be: can we add a noise gate if that's an issue? Seems like an FIR noise gate is possible, but beyond my capabilities.

Click to expand...

Yes, high output and noise are bad. But you can filter noise. Check the latest branch for the feature. I describe it here, showing how to remove a background noise caused by the circuitry in one of these microphones. I have also added a feature to check signal levels.
I read that paper, thought maybe there would be something interesting. They implemented an 8-tap low-pass FIR filter in VHDL. That's like an undergraduate intro to signal processing lab. The signal processing already done is well beyond that.

These two screenshots were taken within a couple seconds of each other, you can see how much the beat error and amplitude jump around.
What could cause that? Is that my mic being too loud or that weird 1khz noise?

Click to expand...

Those are very odd looking waveforms. Yes, it could be the background noise. Loud noises at a constant frequency are bad. But are easy to filter. The waveform lacks a clear unlock, impulse, lock and this makes amplitude detection hard. The post I linked to above describes now just filtering out noise, but also how to identify and amplify the portions of the tick sound that most clearly show the operation of the escapement for better amplitude measurement.

How is it calculating the amplitude and beat error? I'm assuming these waveforms
View attachment 17341701
Show the"tic" and the "toc" of the escapement.
What is the blue line showing? How is it setting what's at 0 and what's at the blue line?

Click to expand...

There are sections from timegrapher manuals that often get posted here. They explain how to use a timegrapher, how amplitude is calculated, what beat error is, and so on.

The blue line is the estimated position of the unlock. "0" is the estimated position of the lock. The scale on the bottom shows how much time is between them in milliseconds, while the scale on the top shows this in degrees of amplitude. The higher the amplitude the faster the balance is moving as it passes through pallet fork. The faster it moves the less time it takes to get from the unlock to the lock. Which is the lift angle, sort of. The escape tooth hits the pallet stone about the same time the fork hits the banking pin. So this is why the top scale gets larger as the beat gets shorter. You'll see adjusting the lift-angle will also adjust the scale.

tpiepho said:

Yes, high output and noise are bad. But you can filter noise. Check the latest branch for the feature. I describe it here, showing how to remove a background noise caused by the circuitry in one of these microphones. I have also added a feature to check signal levels.

Click to expand...

Which branch? You mention a couple in that post, and I tried json-dump and color-coded-tics. I didn't see any new features in the interface after switching branches and recompiling (which i may have messed up. I'm not a huge git user). The filters and spectrogram look badass, exactly what I'm looking for!

I read that paper, thought maybe there would be something interesting. They implemented an 8-tap low-pass FIR filter in VHDL. That's like an undergraduate intro to signal processing lab. The signal processing already done is well beyond that.

Click to expand...

Oh yeah, super basic, that's the point. A simple FIR noise gate could be a solution for removing low level noise. From what I've seen from the original TG, the noise rejection was all on the high side, not the low level side. I'm excited to try out your version (didn't know you had features beyond input selection and HP filter built in, so that's pretty sweet!).

Those are very odd looking waveforms. Yes, it could be the background noise. Loud noises at a constant frequency are bad. But are easy to filter. The waveform lacks a clear unlock, impulse, lock and this makes amplitude detection hard. The post I linked to above describes now just filtering out noise, but also how to identify and amplify the portions of the tick sound that most clearly show the operation of the escapement for better amplitude measurement.

Click to expand...

Yes, that looks like a handy guide. I will refer back to it once I get the best version installed and running.

There are sections from timegrapher manuals that often get posted here. They explain how to use a timegrapher, how amplitude is calculated, what beat error is, and so on.

The blue line is the estimated position of the unlock. "0" is the estimated position of the lock. The scale on the bottom shows how much time is between them in milliseconds, while the scale on the top shows this in degrees of amplitude. The higher the amplitude the faster the balance is moving as it passes through pallet fork. The faster it moves the less time it takes to get from the unlock to the lock. Which is the lift angle, sort of. The escape tooth hits the pallet stone about the same time the fork hits the banking pin. So this is why the top scale gets larger as the beat gets shorter. You'll see adjusting the lift-angle will also adjust the scale.

Click to expand...

I'm familiar with the basics of the timegrapher, not with the specifics of TG-timer. I guess what I'm looking to understand is how it decides where the lock and unlock are happening.
Once i get a grasp on the tg software, I can fiddle with the piezo mic hardware a bit. Failing that, I have other mics even a cheapo ($1) trrs phone headset is better right now (with a few modifications). I just want to use my new toy properly, y'know!
Same watch with a dollar store headset mic.


EDIT: PS i'm still just learning, so happy to look at any materials that are useful. I'll say my background in audio reproduction helps me understand the ins and outs of signal processing, but totally tip my hat to your expertise. 🤠