Jump to content

Recommended Posts

3 hours ago, praezis said:

The amplifier is made from a LM386, an older headphone amp IC.

I wouldn't exactly call it a headphone amplifier. It's a rather nice low-power audio amplifier.  I know it does work really well with piezo pickups if for instance you're making an amplifier for people doing clock repair.

Then for those that are curious I'm attaching the tech sheet.

 

lm386.pdf

Link to comment
Share on other sites

4 hours ago, katolik said:

I identified this board, this is a ready-made amplifier based on LM 386, the green connector has been removed from it, on Aliexpress they are sold for US $ 0.32LM386-200-5-12-10K.thumb.jpg.e6ef6f90bd1ab846714c62dbdf8a2b19.jpg

I have a bunch of these.  They work but a little quirky.  As I recall, I had to modify the circuit to keep the poles out of the right have plane!  I will try to find that work and post it.

Link to comment
Share on other sites

8 hours ago, katolik said:

I identified this board, this is a ready-made amplifier based on LM 386, the green connector has been removed from it, on Aliexpress they are sold for US $ 0.32

That is the board.  If one merely wants the op amp chip, it can be purchased in full reel quantity of 2,500 for only $0.57 each.  How is it the complete board costs half as much as one chip on it at quantity?  Because that is not a real TI LM386 of course.

This board is supposed to be able to run down to 5V, which means it could in theory be directly run from USB with no boost voltage regulator.   But of course the power from USB will be much noisier than a battery.  I'm interested to see what difference it makes.

Edited by xyzzy
Link to comment
Share on other sites

2 hours ago, xyzzy said:

But of course the power from USB will be much noisier than a battery.  I'm interested to see what difference it makes.

It may depend upon your amplifier but I'm running mine off the USB Power and don't seem to be having an issue.

Link to comment
Share on other sites

13 hours ago, JohnR725 said:

I know it does work really well with piezo pickups if for instance you're making an amplifier for people doing clock repair.

That's what I will use this mic for, if at all: a headphone amplifier. But to make it work, I had to change the internal wiring of the jack.

Frank

Link to comment
Share on other sites

6 hours ago, JohnR725 said:

It may depend upon your amplifier but I'm running mine off the USB Power and don't seem to be having an issue.

It seems to work fine.  I'm not able to measure more noise, though I'm not sure what the best way to measure it would be.  I would think that the ratio of the power of the sound we are interested in vs the power of the background noise is the most useful.

The circuit in the microphone switches ground, not battery +.  This means that once I connected the USB V+ the microphone turned on, regardless of the switch, and without USB ground, as the audio ground and the power ground are connected in the microphone and the mic-in ground and the USB ground are apparently not isolated in the computer either.

Here is an example of the switching on the battery powered amplifier.  One can see the noise it creates vs the noise from the computer and audio cable alone. 

image.thumb.png.2b6c4641c3d6e2b8f58ef832376de950.png

And then here we have the noise when USB powered.  Doesn't seem significantly different.  If there was a watch, it would be around -10.

image.thumb.png.3f7e2894bbe33a1c1760066b73a5b041.png

Link to comment
Share on other sites

  • 4 weeks later...
On 12/2/2021 at 8:27 AM, praezis said:

But the chinese mic needs a much louder watch because of its worse signal/noise ratio.
For testing mics, the watch is substituted by a piezo disc driven by a tic noise voltage.

Test results
Needed for full recording level (volume = max, +0 dB):
  CN mic: 0.75 Vp / Z1 mic: 0.3 Vp @ stimulating piezo disc

Sensitivity for "amplitude measurement possible":
  CN mic: 45 mVp / Z1 mic: 9 mVp @ stimulating piezo disc

I am curious, how do you calculate the signal to noise ratio?  I don't think I see a SNR measurement here.

If I understand your first result correctly, you have are measuring your tic noise generator.  This gives you a controllable and repeatable vibration source to measure.  Quite a useful test instrument.   If you have this generate a 750 mV p-p signal, which produces a vibration of some unknown but repeatable magnitude, then you get an audio signal that is at +0 dBFS on the ADC you are recording with.  With the Z1 it only takes 300 mV p-p to generate an audio signal at +0 dBFS.  In both cases, the gain on the input to the ADC is set to the same value (i.e. "volume = max").

Is that correct?  Ignore the fact that the statement, "audio signal at +0 dBFS," is very vague and could mean many different things.  There is some meaningful measurement of the volume of the digitized audio and that measurement is the same.

I doesn't seem to me that this measurement is that useful.  I see two primary attributes that are being measured.  One is the sensitivity of the piezo to vibrations.  This seems important.  The other is how much the microphone amplification circuit amplifies the signal from the piezo.  As long as this is sufficient, it seems less important.   The ADC input will have a gain that can amplify or even attenuate the input signal.  If the microphone amplifies the piezo more, then the ADC gain will correspondingly amplify its input less, producing nearly an identical result. 

So the Z1 microphone might have a more sensitive piezo, which would be useful.  Or it might just have more amplification, which isn't useful.  There's no way to know if it's better or worse than the Chinese mic from those numbers.

It seems to me what one really needs to do is compare the signal to the noise.

Link to comment
Share on other sites

Quote

I am curious, how do you calculate the signal to noise ratio?  I don't think I see a SNR measurement here.

I did not calculate it, but it was visible in the scope views and in the figures. 

Voltages [Vp] are 1/2 of [V p-p], but that doesn't matter with ratios. I used to give the height of the last pulse in the tic noise. The 1st pulse is about 1/3 height and must be recognized for amplitude measurement.

Quote

...Is that correct? 

Yes, correct 🙂

You are right, measurement of needed mV for FS_ADC is not so useful for practical watch testing, I just gave it for comparison. In fact Z1 has less amplification than the CN mic. Both use a piezo disc sensor, but efficiency from watch case to piezo out is rather different.

But very useful is to know, if a quiet ticking watch can be tested or not. That is what 'Sensitivity for "amplitude measurement possible" ' tells us.

Frank

Link to comment
Share on other sites

8 hours ago, praezis said:

But very useful is to know, if a quiet ticking watch can be tested or not. That is what 'Sensitivity for "amplitude measurement possible" ' tells us.

But doesn't this depend also on the algorithm used to measure amplitude?  E.g., if we detect via crossing some fixed threshold, then we might be able to amplify an undetected signal so that it cross the threshold and becomes detected.

Of course this also amplifies the noise.  At some point the signal is so weak that its peak amplitude will be no greater than the peak of the noise and additional amplification will not help.

So, can we measure that?  How many dB above the noise is the signal?  If we were trying to measure some continuous pure tone, then there would be plenty of ways to do that.  I.e., measure the RMS amplitude with and without the tone present and look at their ratio.  But our signal is not continuous, so I don't think it's so simple to measure.  Could one just gate the measurement on when the tick should be present?

Link to comment
Share on other sites

14 hours ago, xyzzy said:

Of course this also amplifies the noise.

You understood the issues very well, and noise level is the key.

14 hours ago, xyzzy said:

Could one just gate the measurement on when the tick should be present?

All software TMs do that already by a hold-off time, but it will not avoid noise, which averages to some constant permanent level.

I don't know how other program developers handled it. My PCTM in auto mode uses the lowest possible threshold, i.e. just a bit over noise level. 

Quote

The signal must be 30 or more dB above the noise.

That is a very luxurious demand, Guido 🙂

Frank

Edited by praezis
  • Like 1
Link to comment
Share on other sites

8 hours ago, praezis said:

very luxurious demand

Frank, you have a lot of experience with that. What I proposed is an empirical observation. But if you can share information from your experience it would be very valuable. Thanks for sharing.

Edited by guidovelasquez
error redaction
Link to comment
Share on other sites

11 hours ago, praezis said:
On 12/29/2021 at 2:58 PM, xyzzy said:

Could one just gate the measurement on when the tick should be present?

All software TMs do that already by a hold-off time, but it will not avoid noise, which averages to some constant permanent level.

The point is not to avoid noise, but to find a way to measure the sensitivity of the microphone.  Compare the level when the signal is preset to that when it is not.  To come up with a useful signal to noise measurement.

Link to comment
Share on other sites

20 minutes ago, xyzzy said:

The point is not to avoid noise, but to find a way to measure the sensitivity of the microphone.  Compare the level when the signal is preset to that when it is not.  To come up with a useful signal to noise measurement.

Personally what I would think would be nice is to expand that thought slightly. It be nice to have a universal way of testing microphones, the interface electronics between the microphone and computer like the preamp and the computer interface itself. This is because unfortunately all three aspects are important if you want to have a decent clean signal for processing.

There isn't exactly a standard of audio input for computers it can vary quite a bit. Or if you decide to get around computer interfacing and purchase a USB interface that presents interesting challenges. They may look the same but unless they were purchased at the exact same time they may have different circuitry with one might work another might not.

Or basically be nice to have a way that you can run a program that would tell you that yes the computer interface is clean you can use this computer. You can plug in your amplifier and run it without the microphone is it still clean. Then the microphone with nothing attached. Then it be nice to have a standard test watch to verify that everything really is working.

 

Link to comment
Share on other sites

4 hours ago, guidovelasquez said:

What I proposed is an empirical observation. But if you can share information from your experience it would be very valuable.

So here is a tick from a USB earbud with a quiet movement.  We can see the loudest part of the sound, at about 2.011 seconds.   It's about 7 dB above the noise.  The first sound should be about 12.5 ms before, at about 1.998.  I certainly can't see it.  There's no way a simple threshold detection will find it and not the noise around it. The peak at 1.993 looks larger and that's just noise.

image.thumb.png.dd5e9a2db7cfc43e4292b448aba7da3c.png

But, some filtering and FFTs later, we can make this graph below.  The first sound can be detected.  It's the yellow bit of energy around 20kHz near the blue vertical dashed lines, which are the detected locations of the tick and tock average amplitude.  It no more than 20 dB above the surrounding noise.

image.png.2433551050c1aaad0ce70803faa48d67.png

Suppose this microphone had poorer high frequency response and dropped everything above 15kHz.  That sound at -12.5 ms would be pretty much all gone then.   If the software does thresholding, it would think both the microphones are equally bad, but the reality is that one that can detect sound at 15 kHz is more capable.

So I think if one measures the microphone against an algorithm that does nothing to reject noise, then it will be overly dependent on the microphone's own ability to reject noise.  But noise can be dealt with through software.  But the microphone's ability to capture a signal can not be augmented after the fact.

  • Thanks 1
Link to comment
Share on other sites

52 minutes ago, xyzzy said:

But noise can be dealt with through software.

Effectively, the noise can be treated. I use Peace EQ. The equalizer can eliminate unwanted noise.
Here I show how the sound of a microphone constructed from a piezo electric disc and using a transistor as a preamplifier can be handled.

1668467124_EspectrogramaconysinEQ.thumb.png.fb332e62b6fe03f067ca6f945f4c4672.png

Filtering with Peace EQ


However, IMHO, WOS alone can solve a very noisy signal situation. Apply a pre-amplication if necessary. And it also integrates filters that can be tested to obtain better results if the signal is not good.

525705352_NiveldeSealenconfiguracin.thumb.png.4f7be5e4fcd4e0e94fc27dc13df209c8.png

Example of poor signal in with 0 dB in WOS

965390339_WOSAplica37dBdeganancia..thumb.png.65f660e81d59be42951e3658b3d34bcf.png

When analyzing, WOS applies 37 dB and much better.  Increase the signal, also increase the noise, but the difference between signal and noise is clear.   EQ is not being used here. The Signal is being used as it is delivered by the transistor Mic, without a preamplifier injected directly to the pc.

80858334_ConfiguracindeAudio.png.00378b4120f38b75b7298da051ae7dec.png

This is the audio setup in the example above

And I share a video to see what is explained.

 

Sin Filtro digital.png

WOS Resuelve.png

Configuración EQ.png

Link to comment
Share on other sites

It is silly that WOS makes you manually adjust the gain.  It should do this automatically to normalize the signal to full scale.  I find it surprising it would not do that.  It seems most basic to me that anything that uses only time domain analysis (i.e., thresholding) that one would rectify the signal, remove any DC component, and then normalize to full scale.

Link to comment
Share on other sites

On 12/30/2021 at 8:57 PM, guidovelasquez said:

Effectively, the noise can be treated. I use Peace EQ. The equalizer can eliminate unwanted noise.

I haven't seen that application before, since I don't use Windows and it's not portable.  It looks like it lets one setup a biquadratic filter cascade but it doesn't actually do any audio processing.  It makes a config file for another application called Equalizer APO, which does the actual processing.

I wanted better filter control in tg, so went overboard and made a fancy system to create biquad filter cascades.  I've been working on it for quite some time and it looks what I've done is reimplement both Peace EQ and Equalizer APO put together, except it's portable and not Windows only.  The filters can be adjusted and audio reflects the changes immediately as the sliders are dragged.  The graph updates in real time too.

I also avoided some limitations.  It looks like PeaceEQ forces one to have exactly 13 filters in the cascade?  I allow any number.  And the equalizer limits filters as if the sampling rate is 44.1kHz, while I allow for the correct sampling rate, e.g. 48 kHz is required for most USB audio devices most PC hardware can do 192 kHz (I've found > 48 to be useless).   They do have some extra filters I haven't implemented.

This is what your chain would look like.  I turned off the 0 dB gain peak filters, since they do nothing to the audio, but do add numerical instability to graph, as it is generated from an equivalent single higher order polynomial rather than the cascade of quadratics.  The Q values are different because I used a sort of normalized bandwidth instead.  The idea is the notch or peak is a certain width in Hz for a given BW value, then you can drag the frequency around and the notch/peak moves with it and *stays the same width*.   Using Q like in PeaceEQ results in the width changing as the frequency changes.  I found this annoying when interactively adjusting the filters.

image.thumb.png.7e1ed0552012df3fe264edcd6823aebe.png

  • Thanks 1
Link to comment
Share on other sites

On 1/4/2022 at 8:33 PM, xyzzy said:

the filters.

Excellent essay. Quite technical. But very useful. I hope in the next two days to share some information about it.
Thanks for sharing.
It would be very helpful to share a waveform graph of the trial. And much better a recording of about 30 seconds in wav format.

Link to comment
Share on other sites

  • 7 months later...

Hello There!

I've been following the reportage on DIY time graphers with a great deal of interest.  Wonderful stuff!

Waiting time aside, personally speaking, I believe it's going to be very hard for  "non-techies" to use the units coming out of China, without proper pre- and post-sales support. 

I think using "commodity" parts (like the "chipamp") is a great idea, to keep costs low and parts availability high.

But where the Chinese stuff falls apart is (as usual) user experience, support and design. 

No surprises there, we beat this particular drum at POLYU all day long.  You might even say it's why SoD exists.  🙂

Here's my unit:

image.thumb.png.32a23bc792cc9aa5df0f9333feeb6ae9.png


I enjoyed the comments in this thread and wanted to thank you all for them....but I am also curious...what would you do to resolve the above considerations, the ones I have identified?  What would you add/modify/delete from the Chinese experience?  How would you "wrap" the hardware so as to make it easier for newbie Watchmaking Enthusiasts?

Personally, I'd like to:

- Remove the battery from the equation and have a simple USB plug that does everything

  - What does everyone think about that? 
  - How might that be done without injecting noise in the supply voltage?

- Offer color choices
  - The current color (grey) is SO ugly.
  - Also, the current enclosure offers NO markings that lead the way to getting help

- Re-engineer the sliding part that captures the watch
  - It currently feels somehow "not right" (it is a little too  smooth, and therefore a bit slippery)

Since 2014, there's been over 50 pages of content concerning DIY time graphers.  It's obviously a topic of great interest to people, and the discussion often verges into the kind of product development debates I see in our Industrial Design stream.

But have any conclusions (or any consensus) been reached over the intervening 8 years since the discussion began?

g.
----

Link to comment
Share on other sites

2 hours ago, Gramham said:

Waiting time aside, personally speaking, I believe it's going to be very hard for  "non-techies" to use the units coming out of China, without proper pre- and post-sales support. 

yes that seems reasonable but were talking about the Chinese is it reasonable to them? For instance what about the 1000 or the 1900 machine very popular timing machines. Where do they come from who makes them warranty support after sales repair anyone have any ideas?

if you looking at the hardware-based machines the program for the 1000 and the 1900 are similar but there's variations especially in 1000 machine. So how many companies are making these how many companies have cloned these?

then for instance I'm currently looking at aliexpress.com for yet that looks similar to yours but not exactly the same. I snipped out an image of the description now I'm not going go over every single point I find humorous will be here all day.

But let's look at the key points designed for amateurs but it's a professional design. I do like the ceramic filter capacitors professional signal processing etc. etc. Somewhere in this discussion somebody had one of these and took it apart creative writing has occurred here. yes I'm really being nice when I say creative writing

or there's a reference to the analytical software does it come with software? There's even a reference to an android cell phone I did know these worked on android cell phone?

So yes somebody purchasing one of these will see the description of it looks impressive it's a great price and will probably show up in this discussion group asking how the heck do they use this thing. Because you not to get any support out of Chinese.

image.thumb.png.d725ff30b8afa1da46f3bafa9566de73.png

 

 

2 hours ago, Gramham said:

Remove the battery from the equation and have a simple USB plug that does everything

  - What does everyone think about that? 
  - How might that be done without injecting noise in the supply voltage?

despite what the description up there says of the electronics I'm assuming it's the same electronics we saw in the previous unit. It have to look at the lowest voltage the integrated circuit runs out and then you could tap in the USB power. Yes have to make sure that all the grounds are correct so don't end up with shorting something out or some other bizarre grounding issue.

then as far as actually implementing this I'm already doing it. I really hate batteries so for my amplifier design which is different than this. I'm using a USB audio device and I'm just taking the 5 V power off the power my circuit seems to work quite nice but mine is different circuits of you blow your computer up tapping into the power I'm just warning it may not work in your case. Just a disclaimer here.

 

2 hours ago, Gramham said:

Offer color choices
  - The current color (grey) is SO ugly.
  - Also, the current enclosure offers NO markings that lead the way to getting help

- Re-engineer the sliding part that captures the watch
  - It currently feels somehow "not right" (it is a little too  smooth, and therefore a bit slippery)

did you notice how this is 3-D printed? 

a lot of times with 3-D printed stuff it comes from an open source somewhere. A lot of parts of the planet will have open source C are allowed to use it for a variety of reasons you can even modify it it's how things get improved and made better because people are modifying whatever exists. But typically the Chinese don't seem to operate that way because if we had the source files then we can look at improving the design.

Oh and then there is one other problem. For those of us that understands or at least think we understand timing machines their incredibly useful devices but sometimes interpreting the results can be interesting. Any new person using a timing machine will find things confusing. Doesn't really matter whether it's a Chinese 1000 machine or a witschi several thousand dollar machine producing similar results will be confusing. Then sometimes interpreting the results of a timing machine isn't much different than trying to read a crystal ball and predict the future. in other words timing machine reading isn't always an exact science. 

 

 

 

 

 

 

Chinese DIY timing machine microphone.JPG

  • Like 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.




×
×
  • Create New...