EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: cncjerry on January 09, 2020, 01:04:33 am
-
I picked-up an HP 3562a dynamic signal analyzer recently. I plan to use it for audio related work and possibly fix a Krell stereo amplifier I have that has developed intermittent distortion only in the high frequencies. In the meantime I wanted to see if I could use the 3562a for measuring pendulum clock timings. I've used a time interval counter(TIC) or frequency counter in the past in conjunction with an FFT in Excel. A friend has over 350 analog master clocks. Knowing what a good FFT looks like vs the clock under test allowed him to focus in on the specific gear ratio that wasn't performing correctly.
I would have thought the 3562a would have a function that would enable me to build histograms or FFT type plots based on the pendulum swing timing. I can't seem to figure it out as the 3562a is more oriented toward voltage over time (or frequency) and in this case, the voltage doesn't matter, just the variance in timing pulses. I'm still reading. I know there is a data capture function and an FFT command so I'm still looking at that. I'll post the plot I'm trying to get. A good way to think about the plot I want is if you had a quartz clock that pulsed once per second, exactly, you would have a peak right at one second. If you put that into RPM, and I know the 3562a can do RPM variances, you would have an exact 60rpm for the perfect clock pulsing once per second and variations would show in the histogram for any imperfections away from that number.
In the case of a pendulum attached to a gear train, because of the very slight variations in the meshing of the gears, the weight of the hands, etc, etc, you will see peaks of varying sizes for the gear ratios and other factors, of which there are many. The pendulums swing back and forth with one cycle of 2 seconds (e.g. 30hz or 30rpm). Of course I can just use the TICs I have, but I thought the 3562a approach would be interesting.
So that's the question, can the 3562a do this?
Thanks
Jerry
-
Interesting idea. So you want to have some sort of microphone or other sensor near or attached to the watch/clock and analyze the vibration. I have a friend who used to repair watches that had a test jjg that the watch fit into and it would measure the speed. I remember seeing the schematic for it, maybe from Germany. It had some crazy gained up transistor front end.
I would imagine the unit you have could do this. I would also imagine if you had a known good unit and compared another against it, you may even be able to detect other problems. If you don't mind, post what you come up here. I would like to follow along.
-
We've been using sensors to either break a light beam or a small hall effect transducer near a magnet that is mounted permanently on the pendulum. I was pretty impressed when my friend figured out the various peaks we were seeing in the datasets. We were collecting a lot of data though. Once we got the clocks pretty close so that they were keeping good time and the tick/tocks were equal (can't remember what that's called, in beat?), we would run the data collection over night. Then we used Excel, some code that I wrote, tlmelab or Stable32 to analyze the data. We were mostly looking at the FFTs though to see if we could figure out what was right or wrong with a clock. The challenge my friend had was he would put several of these large International Time Recorder (predecessor to IBM) clocks in an office lobby, restaurant, etc and then they would all have different times shown over a period of a few days or weeks for the better ones. People aren't used to clocks being wrong anymore. The one in my office keeps good time. It has mercury filled pendulum weights but that is the only compensation device. There are barometric compensation units available too. My clock is within a minute over 6 months but the temperature is really stable in my lab. I'll post a picture of what we are seeing once I get to my lab.
Jerry