WatchUSeek Watch Forums banner

81 - 100 of 451 Posts

·
Registered
Joined
·
920 Posts
In my opinion there is a bit to much noise in your recording when you are using the amp (I made the same experience several times).
TG is solving this issue with it's crazy good filtering algorythm. But in your first video one can see as well the noise in the graph.
Biburo is not able to detect if it's a signal or if it's noise and has to surrender.

Here some points how I get rid of the noise, depending where it's based:
1. power supply: I first tryed a 5V wall power supply which produced a lot of 50Hz noise. So I switched over to a 9V battery (very effective)
2. grounded shield: as shown in the first picture of the thread, about 15 centimetre length taken from an old coax-cable around the pickup-cable and as an extension the red cable which I can easily grab or even more comfortable just step on with my foot :-d (very effective and you see the difference immediately grabbing it or let it hang loose)
3. amplification (computer): give it a try to play with the amplification of the internal soundcard. For me +12dB and about 30% volume for soundrecording is comfortable (if the battery is getting weak I switch to higher amplification), drawback: higher values will amp as well the noise (basic)
4. amplification (external amp): same as before, mine has two potentiometers (Conrad 197688
one for amplification the other for load voltage) trimming those two potentiometers is very important for a good signal to noise ratio (basic)

for both amplification methods I use biburo to adjust to a combination of good soundsignal and low noise level. Noise then is just a few millimetres (about 1/3 of the signalgraph see picture in post #19
or even less when testing loud vintage watches)

If you give it a try and at least modify 3 and 4 your biburo should work fine as well and you will be able to compare ratio, amplitude and repere/beat error of different timing programms.
Thanks.

I'm running it from 9V battery.
All cables are shielded, and the circuit is in a metal box, also connected to the shield.
I'm not using any extra gain on computer side, I can only lower it from 0dB. (it doesn't help)
Mine has 1 potentiometer, I've tried fine tuning with it, but didn't really help :( ( Building a Microphone Amplifier - Watch Timing System I've made this amp)
(I'm not a pro in, so I don't know how to tweak it, I could just make the soldering, and PCB...)

However it is sure, that TG has a really good noise filtering algorithm.
 

·
Registered
Joined
·
250 Posts

@contrate-wheel again many thanks for the TG !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! it is the greatest Christmas gift for the low-budget hobbyist
:-!
Cheers to this statement! I've purchased three movements to begin my hobby...a NOS ETA 6498 (most $), a "used" 6498 (much less $), and a Seagull ST36 (6497 clone...least $). I can tell straight away that the NOS ETA6498 and the Seagull 6497 clone are in good shape although could use some regulation. The "used" 6498 has about 30-40 degrees less amplitude than the new 6498 and is -20 s/day dial up and -55 s/day dial down and 4 ms beat error in all positions so I can already tell it will need a service. I stumbled upon this thread before I even have tools (Christmas present from wife hopefully..lol) and I'm glad I did. It will be interesting to see the results I get when I'm X months down the road and do a full service on the used 6498. A great way to test my new skills.

Anyway, thanks a million to contrate-wheel for the hard work! This is the PERFECT timing solution to the hobbyist looking to keep associated costs down. Tools are expensive enough without spending several hundred dollars on a timing machine. Merry Christmas / Happy Holidays to all!
 

·
Registered
Joined
·
87 Posts
Discussion Starter #83
Hi all, sorry for not being very active lately...

Amplifiers: I still have to experiment, so take what I write with a pinch of salt. In short, Klaus used the Conrad kit no. 197688, a schematic can be found here, and Wizard used the amplifier designed by Stefan Vorkoetter for Watch-O-Scope, instructions here. Judging only by the results and the schematics, so no first-hand experience for now, my advice is to go for the Conrad circuit.

Why? Let's look at a spectrogram taken from a recording of Klaus

spectrogram1.jpg

and compare it to one of Wizard's samples

spectrogram2.png

On the vertical axis we have frequencies, on the horizontal axis we have time. Color represents intensity: from high to low is red, blue, gray. The vertical red streaks are the sounds of the watch (I zoomed in so the window is about one tic-toc wide). As one can see, in Wizard's sample there is no sound content at all above 11KHz. The W-O-S amplifier, in fact, is DESIGNED to filter everything above 11KHz quite aggressively. The instructions page says "to reduce high frequency background noise", however I disagree on this choice. It's enough to look at Klaus's spectrogram to see that the upper half of the spectrum is actually the cleanest, and it contains some very good signal. Klaus's sample also has much less noise. On a side note, unless clipping occurs, digital filters beat most hardware hands down, so, unless you have a compelling reason to filter in hardware, don't do it.

About mains hum, it's definitely there in Wizard's audio.

hum.png

And, what is worse, there are all odd-numbered harmonics of 50Hz up to 1050Hz. I don't know where it mght come from, but I don't see any obvious problem in the design of the amplifier. This noise should not affect much tg, because it filters out everything below 3KHz, however I used a rather bland filter. I believe that W-O-S filters this out too, don't know which kind of filter, but in the settings window you can configure the threshold. Biburo, unfortunately, might not have any band-pass filtering.

Again, don't take this post to seriously, because I don't have first-hand experience. I promise to experiment and share some info in the future.
 

·
Registered
Joined
·
920 Posts
contrate_wheel: Your analysis is not perfect, maybe my recording is to blame.
I've checked it with Audacity, and it is not cut at that much at 11K.
But yes, there is the hum, I don't know why. It is in a metal box, connected to shield, I'm using shielded cables :( (if you have any idea, how to kill it, let me know)

Maybe I'll try the conrad type, when I have time.

audacity1.jpg

audacity2.jpg
 

·
Registered
Joined
·
87 Posts
Discussion Starter #85
Right, probably the sharp drop was caused also by the encoding/decoding. Anyway there is quite a difference between the bottom half and the upper half of your spectrogram, and this is most definitely a consequence of the filters in the amplifier (there are three of them). Anyway it should be easy to disable the low pass filters: you just have to remove C2, C4, and C6. It should not compromise the stability of the circuit: if Stefan is reading, can you confirm? To stay on the safe side, maybe first leave C2 in place.

I have no idea about the hum. I would suspect a faulty ground connection, but I'm sure that you already tested this sort of things...

On the positive side, as long as you use tg and probably W-O-S, I don't think that the hum will affect your results by much.
 

·
Registered
Joined
·
920 Posts
Right, probably the sharp drop was caused also by the encoding/decoding. Anyway there is quite a difference between the bottom half and the upper half of your spectrogram, and this is most definitely a consequence of the filters in the amplifier (there are three of them). Anyway it should be easy to disable the low pass filters: you just have to remove C2, C4, and C6. It should not compromise the stability of the circuit: if Stefan is reading, can you confirm? To stay on the safe side, maybe first leave C2 in place.

I have no idea about the hum. I would suspect a faulty ground connection, but I'm sure that you already tested this sort of things...
I've removed C4, C6. It can be seen in Audacity, that it works, the is not as sharp drop, as there was before.
However there is a lot of noise in the higher frequencies, so in all it became worse, and even TG is not as stable as before, with this setup. :(
 

·
Registered
Speedmaster coaxial on Forstner Komfit
Joined
·
1,882 Posts
I'm trying to run the pre-built binary on windows7, and I get "Error opening audio input: Unanticipated host error". Just to see whether the system can find a microphone, I ran Skype, which works fine. How old are the prebuilts? The link does not give me any choices, and github doesn't seem to have them. I tried compatibility mode for XP SP2, but that doesn't help either. Trying to build under OS X just turned into a huge waste of time. (There seems to be something odd with the specification of the ffwt3 library.)

Update: just having a microphone connected is not enough, it also has to be the default audio source. Windows 7 seems to work. I'll try Halda's OS X build next. Thanks for the quick reply!
 

·
Registered
Joined
·
920 Posts
I've made a small video comparing TG and my Horometer.

 

·
Registered
Joined
·
315 Posts
Trying to build under OS X just turned into a huge waste of time. (There seems to be something odd with the specification of the ffwt3 library.)
Did you use the homebrew version of fftw?

If you install homebrew from Homebrew — The missing package manager for OS X then it should be as simple as a "brew install dmnc/horology/tg".

If you still have problems, let me know which OS version you are on. All working ok for me here on 10.11.2.
 

·
Registered
Speedmaster coaxial on Forstner Komfit
Joined
·
1,882 Posts
Actually, yesterday I tried following the recipe for using brew exclusively, but brew update fails with some dyld issue and Xcode complaint. I verified Xcode is installed with command line tools, but there must some interaction with software installed using other systems. Too bad these porting/installation helpers like fink or brew all insist on reinventing the wheel, instead of installing in a standard or at least compatible way.
I'm running 10.9. Is there a way to clean up the path just for brew (I.e. make it work without obliterating other installs)?

Update: it finally worked, after obliterating /opt/local and allowing brew to overwrite the links for portaudio and fftw. (I decided that it's unlikely I installed something important on this machine.) Couldn't find either fink nor macports on this machine, so the conflicting stuff must have been from manual installs.

Now of course it turns out none of the microphones I have provide a suitable input for reliable operation.
 

·
Registered
Joined
·
87 Posts
Discussion Starter #92
Hi all, sorry for deserting this thread so long, and happy new year to all!

@LCheapo: I'm happy that the recipe prepared by dmnc finally worked. It's a pity if none of your mics turns out to be good enough. If it so happens, you might consider sending me some sample of your best audio, this will help me in future improvements of the algorithm. Thanks, anyway, for considering my program!

@Wizard: Thanks for your video. I've tried to find the cause of the difference in the amplitude, but I failed at that. As we noticed already there is quite a bit of noise in the audio, plus some produced by the codec. This turns out to be too much to analyze the audio either manually or with Watch-O-Scope or Watch Escapement Analyzer. The waveform detected by tg seems to give little room for doubt, but it results from a moderately complicated process of filtering and integration, and, without comparison with any more direct method, it's hard to say anything. In your video with WOS, as you noticed I believe, the outputs of WOS are just swinging around haphazardly: unfortunately the audio is clearly too noisy for WOS to detect even the rate.

By the way, for any software or hardware, the order of how difficult it is to get these numbers is the same: from easiest to hardest (namely from less noise-sensitive to more noise-sensitive) they are rate, beat error, amplitude. So one should not trust one of these results unless the easier ones are stable and trustworthy.

The Horometer graph looks roughly consistent with the graphs determined by tg (I take this picture from your video)

horometer.jpg

However one pixel on the Horometer screen is 0.625ms wide, and we are looking here for a half-millisecond discrepancy. It can not be seen clearly just because it's less than one pixel wide. The Horometer is also telling us only half of it's story, because it draws a vertical marker where it detects the first pulse (the vertical line at 20ms), but it does not tell us where it detects the third pulse, so it's hard to understand what it does. All the interesting action in this graph takes place between the 20ms and the 30ms mark. If you could zoom in this section, it would be interesting to compare tg's waveforms with the Horometer's ones. Thanks again for your support!

@all: If anyone else experiences the same error in the amplitude, I am highly interested in sorting this problem out, so any feedback on this issue would be appreciated.
 

·
Registered
Joined
·
7 Posts
Firstly, I'd like to thank Contrate_wheel for this program which seems very interesting as it seems to work great according to some testimonials here. And I would like to help if I can providing examples.

But first I need it run on my OS X and also find a suitable microphone, latter which will pose no problem I guess.

I'd like to ask for some help to run/compile the program. So far I've followed the page 3 instructions: installed xcode, then homebrew, then installed: fftw, gtk+, portaudio (these come up via command: brew list). I've downloaded the full source code 'tg-master' folder to my 'users/username/downloads' folder. But when I do the command whilst being in directory 'Users/username': 'brew install downloads/tg', I get the following error:

Error: No similarly named formulae found.
==> Searching taps...
Error: No formulae found in taps.

Same with commands: 'brew install /users/username/downloads/tg-master' or 'brew install ~/downloads/tg'

When I do command: 'brew install downloads/tg-master/tg, I will get the following though:


==> Tapping downloads/tg-master
Cloning into '/usr/local/Library/Taps/downloads/homebrew-tg-master'...
Username for 'https://github.com':

But I guess that is not the desired route.


I am very much unfamiliar with commando's and such, and have tried to search for answers online, but after many hours looking and trying I just decided to post my problem as there might be others too who encountered this problem. Also, everyone here seems to be glad to help and some apparently very knowledgable around these technical matters :)

The precompiled version of Halda does work btw. But I'd like to be able to get future versions of contrite_wheel working, which also now have the reset and clear option extra I believe.

I'm running Yosemite 10.10.5. If there is any information needed to help troubleshoot, I will gladly supply it of course.

Many thanks in advance.
 

·
Registered
Joined
·
29 Posts
I am very much unfamiliar with commando's and such, and have tried to search for answers online, but after many hours looking and trying I just decided to post my problem as there might be others too who encountered this problem.
You're almost there! In Terminal, simply go to the directory where the downloaded source code is (by using the cd command). Then enter the command make.

This should invoke the compiler and produce the executable for you. Good luck!
 

·
Registered
Joined
·
7 Posts
You're almost there! In Terminal, simply go to the directory where the downloaded source code is (by using the cd command). Then enter the command make.

This should invoke the compiler and produce the executable for you. Good luck!
Halda, many thanks. I now have an executable and working program on OS X! Regret the work I have put in getting a newer Bash shell working, resulting temporarily in a non-working Terminal. All is well now and have the newer Bash shell available if needed.

As soon as I have a mic, I'll start playing around with it and will see if I can get some useful examples to post here. Shame I threw my broken earphones recently which had a mic I never used, oh well...

Keep on the fine work!!
 

·
Registered
Joined
·
29 Posts
If you don't have a microphone yet but still want to try out tg on your Mac there are tools that will allow you to route the audio from other apps into a "virtual microphone" so you can for example play a YouTube clip of a ticking watch and have tg analyze that audio as if it was coming from a microphone.

There's an old free app called Soundflower that allows you to do that, and also a brand new one (but not free) called Loopback.
 

·
Registered
Joined
·
87 Posts
Discussion Starter #97 (Edited)
@Roli-poli: I'm happy to hear that you managed to compile tg.

For other OS X user, it is better to point out that there is a little confusion in the operations that Roli-poli attempted. To use the homebrew formula prepared by dmnc, as far as I can understand not being a mac user, the correct sequence of operations is

step 1 - install homebrew as indicated in [url]http://brew.sh[/URL]

step 2 - run brew doctor and do everything it asks of you

step 3 - run the magic command: brew install dmnc/horology/tg --HEAD

And this is it. No other step whatsoever is required.

Edit: corrected the list of steps as indicated by dmnc.
 

·
Registered
Joined
·
7 Posts
@Roli-poli: I'm happy to hear that you managed to compile tg.

For other OS X user, it is better to point out that there is a little confusion in the operations that Roli-poli attempted. To use the homebrew formula prepared by dmnc, as far as I can understand not being a mac user, the correct sequence of operations is

step 1 - install homebrew as indicated in Homebrew — The missing package manager for OS X

step 2 - run the magic command: brew install dmnc/horology/tg --HEAD

And this is it. No other step whatsoever is required. Please correct me if I am wrong...

Well step 1 should be installing Xcode (if not already) via Macstore>developer tools, or through commando. Next indeed homebrew, but I understood also some dependencies (if that's the correct term), namely: gkt+, ffth, portaudio. This can be done through 'homebrew', through commando in terminal: brew install gtk+ portaudio fftw
Then, or earlier, you have to download of course your full source code files which come in the folder: tg-master
Instead of the 'brew install..'-method, the following method worked for me with, where you use the path to the tg-master-folder as the path for the for the 1st of the following 2 commando's:
1: cd /users/username/downloads/tg-master (note that path is an example, just as dmnc/horology/tg was)
2: make
Now, there should be 2 new files in folder tg-master.

I have also tried 'brew install .../tg --HEAD', as also
without the '--HEAD', but then homebrew couldn't find anything. Probably I have done something wrong, as it worked for others, but the above mentioned method worked for me.

Also, I don't really know what Homebrew does, or what Xcode exactly brings to table, and if they are indeed needed, or just more practical. Not asking for this information either by the by
b-)
 

·
Registered
Joined
·
7 Posts
If you don't have a microphone yet but still want to try out tg on your Mac there are tools that will allow you to route the audio from other apps into a "virtual microphone" so you can for example play a YouTube clip of a ticking watch and have tg analyze that audio as if it was coming from a microphone.

There's an old free app called Soundflower that allows you to do that, and also a brand new one (but not free) called Loopback.
Thanks for the tips, but am planning to get some 'real world examples' :). Cheap clip on mic will be ordered tomorrow. I did however tried it already for just a quick test on the integrated mic on my laptop and, believe it or not, through the mic on a toy Sega I-Dog, but this was done on Windows Vista machine. Too much noise in both quick tests I gathered quickly (was noisy).
 

·
Registered
Joined
·
7 Posts
Thanks for the tips, but am planning to get some 'real world examples' :). Cheap clip on mic will be ordered tomorrow. I did however tried it already for just a quick test on the integrated mic on my laptop and, believe it or not, through the mic on a toy Sega I-Dog, but this was done on Windows Vista machine. Too much noise in both quick tests I gathered quickly (was noisy).
Funnily enough I just found out that I've tried using the light sensor on the toy as a mic, and not the actual mic which is 10cm below the light sensor. So, even despite a noisy laptop in vicinity and the clock not being near the mic, it locked on for a moment :)
 
81 - 100 of 451 Posts
Top