Towcester, one of my old stomping grounds. I grew up in wood burcote and still have family in Grimscote and Roade 🙂
We live just at the bottom of the lane in Towcester and I often pass through Wood Burcote on my way to work (at Silverstone). It's a nice area to live in.
Regards,
Regards,
PM sent, but I found the reference to the maths and it works fine, post #5 of this thread.
There is a good reference thread on PinkFishMedia site:
Turntable speed analysis - Page 46 - pink fish media
Post #679 gives the formula for a demodulator. It also requires a Hilbert transformer and a differentiator, but it is all pretty well explained in the book listed in the thread: Understanding Digital Signal Processing (Third Edition) by Richard G. Lyons, starting on page 758. (You need the 3rd edition or later, earlier versions do not have a chapter on instantaneous frequency or phase computations.
Since this doesn't have to compute a solution in real time, processing power and speed should not be an issue. I was planning on visiting this as a winter project, but DSP is not my long suit. This would be an interesting exercise for the folks here who are fluent with DSP. I've already looked at the file formatting for .wav files, and it is very straight forward to extract the sampling data. I'm told Python has very good graphing capabilities for the output.
Turntable speed analysis - Page 46 - pink fish media
Post #679 gives the formula for a demodulator. It also requires a Hilbert transformer and a differentiator, but it is all pretty well explained in the book listed in the thread: Understanding Digital Signal Processing (Third Edition) by Richard G. Lyons, starting on page 758. (You need the 3rd edition or later, earlier versions do not have a chapter on instantaneous frequency or phase computations.
Since this doesn't have to compute a solution in real time, processing power and speed should not be an issue. I was planning on visiting this as a winter project, but DSP is not my long suit. This would be an interesting exercise for the folks here who are fluent with DSP. I've already looked at the file formatting for .wav files, and it is very straight forward to extract the sampling data. I'm told Python has very good graphing capabilities for the output.
Although a real time analysis (well two rotations delayed) would be an awesome diagnostic tool and I'd be willing to chip in to a crowdsource to fund dev of such.
Having said that as I am (like many here) of an age where you left computers to chunk away at problems for hours on end I can plan accordingly 🙂
Having said that as I am (like many here) of an age where you left computers to chunk away at problems for hours on end I can plan accordingly 🙂
Good grief, Scott! There is so much going on in the Boasash paper how'd you find what you need? And you say it works? You've run the math on a wave file?
No, I said LD says it's there somewhere and it works. As a start there are several versions of Matlab code for doing the basic Hilbert transform method of extracting the instantaneous frequency on the web. I would take a data set and compare the two results, I suspect the standard method has more noise and/or sensitivity to AM, etc. than LD's trick. This is the kind of problem I love but don't have the time right now to totally immerse in the details.
I pays to keep in mind that this is a problem of narrow scope as opposed to the totally general approaches in the literature. Usually this leads to some tricks that work in the special case.
I pays to keep in mind that this is a problem of narrow scope as opposed to the totally general approaches in the literature. Usually this leads to some tricks that work in the special case.
I am experimenting with a hardware engineers approach. I have made a LTSpiceXVIII simulation of a 3kHz centre PLL and use the test record 3kHz wav file as input. The demodulated output is looking good.
I can save that to a file and make a polar plot.
Using Audacity to cleanup the wav file first with a 2kHz 24dB HPF and 5kHz 24dB LPF makes the input wav look a lot cleaner
I can save that to a file and make a polar plot.
Using Audacity to cleanup the wav file first with a 2kHz 24dB HPF and 5kHz 24dB LPF makes the input wav look a lot cleaner
davidsrb,
Could you scale a version that can be switchable to use a 1kHz tone? Of all the test records I have, a 1kHz tone is much more common.
Ray K
This is a link to the LTSpice file
http://www.cix.co.uk/~dayojah/3kpll_LFtest.asc
Change the input file filename to suit on V1.
The output file name is set by the .wavefil declaration
No reason why it wouldn't work at 1kHz, just the filtering gets a bit tight between the 1kHz carrier and the cogging harmonics. Just change the MODULATOR mark and space to 1.1K and 0.9K
http://www.cix.co.uk/~dayojah/3kpll_LFtest.asc
Change the input file filename to suit on V1.
The output file name is set by the .wavefil declaration
No reason why it wouldn't work at 1kHz, just the filtering gets a bit tight between the 1kHz carrier and the cogging harmonics. Just change the MODULATOR mark and space to 1.1K and 0.9K
Last edited:
All good stuff, will be interesting to see how it performs as to AM rejection, broad spectrum noise and variation in carrier level. At least in spice one can build a perfect circuit in the sense it will perform as intended, and develop it.
I still haven't got round to doing the diligence on my SW demodulator/detector, just in case it's novel...... As pointed out, all required theory is disclosed in the text Scott linked, for example. Just a matter of whether anybody has previously combined factors in the way I came up with, and it does yield remarkable results. Potentially applicable to all manner of situations - imagine it upscaled in frequency......so it's probably already known, but hey one never knows and several standard methods definitely lack the extra step that makes all the difference.
Other polar plot SW solutions I've seen online seem to suffer from uncertain calibration, and relatively poor noise rejection IIRC. So I think it's worth focussing on those aspects.
Interesting to see how the spice hardware emulation performs against real data - if you can post a link to files I'll run calibration plots for comparison?
LD
I still haven't got round to doing the diligence on my SW demodulator/detector, just in case it's novel...... As pointed out, all required theory is disclosed in the text Scott linked, for example. Just a matter of whether anybody has previously combined factors in the way I came up with, and it does yield remarkable results. Potentially applicable to all manner of situations - imagine it upscaled in frequency......so it's probably already known, but hey one never knows and several standard methods definitely lack the extra step that makes all the difference.
Other polar plot SW solutions I've seen online seem to suffer from uncertain calibration, and relatively poor noise rejection IIRC. So I think it's worth focussing on those aspects.
Interesting to see how the spice hardware emulation performs against real data - if you can post a link to files I'll run calibration plots for comparison?
LD
One I made with a STD305S and the HFS75 test record, 3 kHz tone
http://www.cix.co.uk/~dayojah/3khzhfs75.wav
I suspect rumble could mess up my method, as it phase modulates the zero crossings, which is why I band limit to the range 2kHz up to 4.5kHz
http://www.cix.co.uk/~dayojah/3khzhfs75.wav
I suspect rumble could mess up my method, as it phase modulates the zero crossings, which is why I band limit to the range 2kHz up to 4.5kHz
Last edited:
that would be an asymmetrical limitation of the modulation pass band within the area of valid data. Remember Shannon, valid data is plus and minus half the frequency of the carrier tone. To address rumble, you should only need a simple HPF around 200hz or so, not critical.
Alan
Alan
The LTSpice simulation is already painfully slow.
As I am capturing with Audacity anyway, using it to cleanup first is quicker