Melodeon.net Forums

Please login or register.

Login with username, password and session length
Advanced search  

News:

Welcome to the new melodeon.net forum

Pages: 1 [2] 3 4   Go Down

Author Topic: Electronic tuners  (Read 24175 times)

0 Members and 1 Guest are viewing this topic.

Martyn

  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 455
  • Hohnerohner
    • Martyn White Melodeon Tuning and Repairs
Re: Electronic tuners
« Reply #20 on: February 17, 2010, 01:59:06 AM »

I've been thinking about buying a new tuner. Regarding the Korg OT-120, is it possible to set it in some way when tuning the tremolo reeds to +20 cents at the bottom end tapering off at the top end? Or does the display tell you how many cents sharp or flat you are?

Martyn
« Last Edit: February 17, 2010, 02:00:58 AM by martyn »
Logged
'Anything from a small repair to a full restoration'.
www.melodeonrepairs.co.uk

Theo

  • Administrator
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 14475
  • Hohner Club Too
    • The Box Place
Re: Electronic tuners
« Reply #21 on: February 17, 2010, 09:03:17 AM »

The display tells you approximately how many cents you are off, but its not very precise, I think the smallest scale division is 10 cents.
Logged
Theo Gibb - Gateshead UK

Proprietor of The Box Place for melodeon and concertina sales and service.
See my YouTube channel  for videos on repair and tuning

TomB-R

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 598
Re: Electronic tuners
« Reply #22 on: February 17, 2010, 09:36:44 AM »

The display tells you approximately how many cents you are off, but its not very precise, I think the smallest scale division is 10 cents.

Eek! That's pretty vague! I have no complaints about my Seiko SAT500 which gets used for various things. It gives three levels of readout, coloured lights, a virtual needle, and most useful for reeds, a cents reading. At the moment I'm not desperately worried about less than 1 cent......
Logged

Martyn

  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 455
  • Hohnerohner
    • Martyn White Melodeon Tuning and Repairs
Re: Electronic tuners
« Reply #23 on: February 17, 2010, 10:26:24 AM »

I'm using a Seiko chromatic tuner at present and I find it ok for tuning the treble end but when I try to tune the low basses the needle waves about like a windscreen wiper. I can usually get around this by repositioning the microphone.
A good thing about it though is that it shows how many cents sharp or flat you are. Ideally I would like a tuner with this feature that can also handle the low basses. Any ideas?

Martyn
« Last Edit: February 17, 2010, 11:48:10 AM by martyn »
Logged
'Anything from a small repair to a full restoration'.
www.melodeonrepairs.co.uk

Bill the Farmer

  • Not quite so
  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 497
  • Hohner Club, Liliput...
    • Bill Farmer
Re: Electronic tuners
« Reply #24 on: February 17, 2010, 07:19:20 PM »

I've been working on a software tuner for some time, off and on. I would like to be able to do more than one reed at once, but I haven't worked out how it's done. It's a work in progress at the moment...

« Last Edit: February 17, 2010, 07:20:55 PM by Bill Farmer »
Logged
Bill of Old Radipole Village

Andy in Vermont

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 948
  • Mélodie
    • Melodeon Minutes (blog)
Re: Electronic tuners
« Reply #25 on: February 17, 2010, 07:24:23 PM »

I've been working on a software tuner for some time, off and on. I would like to be able to do more than one reed at once, but I haven't worked out how it's done. It's a work in progress at the moment...



Bill,
It looks great -- I like that it gives a readout in Hz as well as +/- cents --- this would be really useful for me (more useful than my Virtual Strobe in fact).  Sorry if this was covered in another thread or a previous statement in this thread, but which platform will the tuner work on?
-Andy

Steve_freereeder

  • Content Manager
  • Hero Member
  • ***
  • Offline Offline
  • Posts: 7511
  • MAD is inevitable. Keep Calm and Carry On
    • Lizzie Dripping
Re: Electronic tuners
« Reply #26 on: February 17, 2010, 08:59:50 PM »

Well done Bill!
I especially like the idea of the menu (presumably interactive) on the right hand side where it tells you what file to use, which other tools are required and a direct, live link through to Theo for when things go wrong. I've never seen that sort of innovation in a tuner before.

PS will there be a Mac version please?
Logged
Steve
Sheffield, UK.
www.lizziedripping.org.uk

melodeon

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1732
Re: Electronic tuners
« Reply #27 on: February 17, 2010, 11:24:31 PM »

Looks good.
Logged

Theo

  • Administrator
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 14475
  • Hohner Club Too
    • The Box Place
Re: Electronic tuners
« Reply #28 on: February 17, 2010, 11:41:30 PM »

Have you worked out how to do one reed?
Logged
Theo Gibb - Gateshead UK

Proprietor of The Box Place for melodeon and concertina sales and service.
See my YouTube channel  for videos on repair and tuning

Anahata

  • This mind intentionally left blank
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 6767
  • Oakwood D/G, C/F Club, 1-rows in C,D,G
    • Treewind Music
Re: Electronic tuners
« Reply #29 on: February 18, 2010, 08:30:12 AM »

I especially like the idea of the menu ... where it tells you what file to use, which other tools are required and a direct, live link through to Theo

 ;D*Chortle* ;D

I would like to be able to do more than one reed at once, but I haven't worked out how it's done.

Fourier analysis, I'd guess. Find a FFT algorithm (Fast Fourier Transform) code is where I'd start looking.
Once you have a frequency plot, look for the peaks. A tremolo pair of reeds will be two peaks close to each other.
Logged
Слава Україні !
Music recording and web hosting: www.treewind.co.uk
Mary Humphreys and Anahata: www.maryanahata.co.uk
Ceilidh band: www.barleycoteband.co.uk

Bill the Farmer

  • Not quite so
  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 497
  • Hohner Club, Liliput...
    • Bill Farmer
Re: Electronic tuners
« Reply #30 on: February 18, 2010, 01:03:52 PM »

Have you worked out how to do one reed?

That's one of the problems. The lower reeds generate harmonics at a higher level than the fundamental, so it will tend to latch on to a harmonic, or jump between different harmonics. That's probably the cause of the jumping problem mentioned earlier.

Well done Bill!
I especially like the idea of the menu (presumably interactive) on the right hand side where it tells you what file to use, which other tools are required and a direct, live link through to Theo for when things go wrong. I've never seen that sort of innovation in a tuner before.

PS will there be a Mac version please?

Thanks. The menu is on the signal generator that I was using for testing.  The file link is a b*stard for fine adjustments.  :o
I hadn't though of a live support link. :Ph

If and when I get it working usefully, I will be able to do a Mac version relatively easily, because it will just reuse the same code.

I would like to be able to do more than one reed at once, but I haven't worked out how it's done.

Fourier analysis, I'd guess. Find a FFT algorithm (Fast Fourier Transform) code is where I'd start looking.
Once you have a frequency plot, look for the peaks. A tremolo pair of reeds will be two peaks close to each other.


It uses fourier analysis, together with an algorithm I got from http://www.dspdimension.com. Unfortunately, the frequencies from two or more reeds tuned only a few cents apart are too close together to be discriminated in that way, I think. I haven't investigated further yet. There may be an expert on digital signal processing out there who can point me in the right direction.
Logged
Bill of Old Radipole Village

Robin Harrison

  • Good talker
  • **
  • Offline Offline
  • Posts: 94
Re: Electronic tuners
« Reply #31 on: February 18, 2010, 01:37:43 PM »

I've used this tuner on concertina reeds...............it may be useful for accordions.
  TurboTuner...model ST-122
http://www.turbo-tuner.com/
   The major problem I've found is that for the really low reeds, even using an external mic, it has a hard time locking on to the note. It has sometimes been time consuming however it may be a function of needing more air than my tuning bellows can give.
 Nonetheless, I've found it otherwise easy to use  and accurate to +/- .02 cents.
   Hope this is helpful
               Robin
Logged

Anahata

  • This mind intentionally left blank
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 6767
  • Oakwood D/G, C/F Club, 1-rows in C,D,G
    • Treewind Music
Re: Electronic tuners
« Reply #32 on: February 18, 2010, 01:48:40 PM »

The lower reeds generate harmonics at a higher level than the fundamental.

Automatic level control that does it by adjusting the cutoff frequency of a sharp lowpass filter? (all digitally, of course).
You'll probably need a simple gain-controlled ALC first...

Unfortunately, the frequencies from two or more reeds tuned only a few cents apart are too close together to be discriminated in that way, I think. I haven't investigated further yet. There may be an expert on digital signal processing out there who can point me in the right direction.

I wonder if a first estimate of the approximate band would let you then switch to a fine tuned narrow band fourier transform covering just the range you need, with better discrimination. Not that I'm a DSP expert....
Logged
Слава Україні !
Music recording and web hosting: www.treewind.co.uk
Mary Humphreys and Anahata: www.maryanahata.co.uk
Ceilidh band: www.barleycoteband.co.uk

Theo

  • Administrator
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 14475
  • Hohner Club Too
    • The Box Place
Re: Electronic tuners
« Reply #33 on: February 18, 2010, 03:23:46 PM »

That's one of the problems. The lower reeds generate harmonics at a higher level than the fundamental, so it will tend to latch on to a harmonic, or jump between different harmonics. That's probably the cause of the jumping problem mentioned earlier.

That happens with many tuners, they tend to pick on the 5th.  I've found that happens a lot with the Strobosoft app on the iphone when I use it for tuning the fiddle.  It can happen with reeds too.

I did a little dsp  many moons ago, the only suggestion I can add to what's already been suggested is work out a typical pattern of harmonics for a free reed then look for this pattern in the signal, rather than just looking for a single peak.

Or taking a completely different strategy, I wonder if Dirk would be interested in collaborating with someone like you Bill,  if you could port his tuner to Mac, or Linux, and at the same time do something about the clunky user interface then we would all be winners, Dirk included.
Logged
Theo Gibb - Gateshead UK

Proprietor of The Box Place for melodeon and concertina sales and service.
See my YouTube channel  for videos on repair and tuning

Bill the Farmer

  • Not quite so
  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 497
  • Hohner Club, Liliput...
    • Bill Farmer
Re: Electronic tuners
« Reply #34 on: February 18, 2010, 06:33:55 PM »

The lower reeds generate harmonics at a higher level than the fundamental.

Automatic level control that does it by adjusting the cutoff frequency of a sharp lowpass filter? (all digitally, of course).
You'll probably need a simple gain-controlled ALC first...

I will have to explore that idea.

Unfortunately, the frequencies from two or more reeds tuned only a few cents apart are too close together to be discriminated in that way, I think. I haven't investigated further yet. There may be an expert on digital signal processing out there who can point me in the right direction.

I wonder if a first estimate of the approximate band would let you then switch to a fine tuned narrow band fourier transform covering just the range you need, with better discrimination. Not that I'm a DSP expert....


Thanks for that idea - it's pointed me in the right direction, I think. I found a relevant article here http://www.embedded.com.
Logged
Bill of Old Radipole Village

Lars

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 797
Re: Electronic tuners
« Reply #35 on: February 18, 2010, 07:25:03 PM »

I've used this one with good results on my old laptop, but still prefer a Peterson:

http://www.seventhstring.com/tuner/tuner.html

It's got all you need for DIY hobby use. It does pick up the 5th from time to time, but not very often at all when doing a single reed at a time. It allows hertz or cent presets, and will show very precise readings when the reed is within a few feet of the mike.

AND: It's free. It does, however, only show readings down to a single cent, and with my old laptop it sometimes need a five second break when the computer is 'thinking' too hard. As always, the ear is the most valuable tool by all means, but I found the tuner to work good enough for home use.
Logged

IanD

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1703
  • Too many melodeons...
Re: Electronic tuners
« Reply #36 on: February 20, 2010, 05:46:04 PM »

The lower reeds generate harmonics at a higher level than the fundamental.

Automatic level control that does it by adjusting the cutoff frequency of a sharp lowpass filter? (all digitally, of course).
You'll probably need a simple gain-controlled ALC first...

I will have to explore that idea.

Unfortunately, the frequencies from two or more reeds tuned only a few cents apart are too close together to be discriminated in that way, I think. I haven't investigated further yet. There may be an expert on digital signal processing out there who can point me in the right direction.

I wonder if a first estimate of the approximate band would let you then switch to a fine tuned narrow band fourier transform covering just the range you need, with better discrimination. Not that I'm a DSP expert....


Thanks for that idea - it's pointed me in the right direction, I think. I found a relevant article here http://www.embedded.com.

The problem with using FFTs is that if you want good frequency resolution the acquisition time span of the FFT has to be long -- to get 1Hz resolution you need a 1 second acquisition window, so if you're trying to tune a reed within a fraction of a beat per second the response time is very long. This is true whether you use a wideband or a narrowband transform, the delay times the resolution is constant. Overlapped FFTs can help by increasing the update rate but there's still a big delay if you want fine resolution.

A better method for tuning is to use an FFT for coarse frequency estimation and a digital phase-locked-loop (DPLL) for fine frequency measurement, this then breaks the link between frequency resolution and response time. If the FFT has enough resolution to separate out peaks then you can use digital filters and a separate DPLL for each reed and measure several at the same time, like Dirk's tuner does.
Logged
Oakwood Model 4 D/G, Castagnari Dony D/G/#, Castagnari Tommy G/C, Baffetti Binci D/G, Hohner Preciosa D/G, Melos Bb/Eb, Lightwave SL5 and Kala California fretless basses

Bill the Farmer

  • Not quite so
  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 497
  • Hohner Club, Liliput...
    • Bill Farmer
Re: Electronic tuners
« Reply #37 on: February 20, 2010, 07:36:14 PM »

The problem with using FFTs is that if you want good frequency resolution the acquisition time span of the FFT has to be long -- to get 1Hz resolution you need a 1 second acquisition window, so if you're trying to tune a reed within a fraction of a beat per second the response time is very long. This is true whether you use a wideband or a narrowband transform, the delay times the resolution is constant. Overlapped FFTs can help by increasing the update rate but there's still a big delay if you want fine resolution.

A better method for tuning is to use an FFT for coarse frequency estimation and a digital phase-locked-loop (DPLL) for fine frequency measurement, this then breaks the link between frequency resolution and response time. If the FFT has enough resolution to separate out peaks then you can use digital filters and a separate DPLL for each reed and measure several at the same time, like Dirk's tuner does.

The algorithm I have been using gets around the FFT size problem by using a medium sized overlapped FFT which gets updated 11025/1024 ~= 11 times a second. It then uses the phase difference between subsequent runs of the FFT to calculate the actual frequency. It appears to be about as accurate as the clock on the sound card. I hadn't thought of using a DPLL, I shall have to look that up. I wrote one more than 25 years ago, but I can't remember how it worked after all that time.
Logged
Bill of Old Radipole Village

IanD

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1703
  • Too many melodeons...
Re: Electronic tuners
« Reply #38 on: February 21, 2010, 01:07:02 PM »

The problem with using FFTs is that if you want good frequency resolution the acquisition time span of the FFT has to be long -- to get 1Hz resolution you need a 1 second acquisition window, so if you're trying to tune a reed within a fraction of a beat per second the response time is very long. This is true whether you use a wideband or a narrowband transform, the delay times the resolution is constant. Overlapped FFTs can help by increasing the update rate but there's still a big delay if you want fine resolution.

A better method for tuning is to use an FFT for coarse frequency estimation and a digital phase-locked-loop (DPLL) for fine frequency measurement, this then breaks the link between frequency resolution and response time. If the FFT has enough resolution to separate out peaks then you can use digital filters and a separate DPLL for each reed and measure several at the same time, like Dirk's tuner does.

The algorithm I have been using gets around the FFT size problem by using a medium sized overlapped FFT which gets updated 11025/1024 ~= 11 times a second. It then uses the phase difference between subsequent runs of the FFT to calculate the actual frequency. It appears to be about as accurate as the clock on the sound card. I hadn't thought of using a DPLL, I shall have to look that up. I wrote one more than 25 years ago, but I can't remember how it worked after all that time.

That would work as well. Instead of using phase differences (of which frequency bin?), a better method is to use windowing and measure not just the peak FFT bin level but also the ones on either side, then fit these to the shape of the window function to work out the signal frequency offset from the bin centre frequency.

e.g. if centre bin is at F1 and bins either side are equal level, frequency equals F1
       if bins F1 and F2 are same level, frequency is (F1+F2)/2

Ian
Logged
Oakwood Model 4 D/G, Castagnari Dony D/G/#, Castagnari Tommy G/C, Baffetti Binci D/G, Hohner Preciosa D/G, Melos Bb/Eb, Lightwave SL5 and Kala California fretless basses

Bill the Farmer

  • Not quite so
  • Respected Sage
  • ****
  • Offline Offline
  • Posts: 497
  • Hohner Club, Liliput...
    • Bill Farmer
Re: Electronic tuners
« Reply #39 on: February 21, 2010, 04:45:49 PM »

The problem with using FFTs is that if you want good frequency resolution the acquisition time span of the FFT has to be long -- to get 1Hz resolution you need a 1 second acquisition window, so if you're trying to tune a reed within a fraction of a beat per second the response time is very long. This is true whether you use a wideband or a narrowband transform, the delay times the resolution is constant. Overlapped FFTs can help by increasing the update rate but there's still a big delay if you want fine resolution.

A better method for tuning is to use an FFT for coarse frequency estimation and a digital phase-locked-loop (DPLL) for fine frequency measurement, this then breaks the link between frequency resolution and response time. If the FFT has enough resolution to separate out peaks then you can use digital filters and a separate DPLL for each reed and measure several at the same time, like Dirk's tuner does.

The algorithm I have been using gets around the FFT size problem by using a medium sized overlapped FFT which gets updated 11025/1024 ~= 11 times a second. It then uses the phase difference between subsequent runs of the FFT to calculate the actual frequency. It appears to be about as accurate as the clock on the sound card. I hadn't thought of using a DPLL, I shall have to look that up. I wrote one more than 25 years ago, but I can't remember how it worked after all that time.

That would work as well. Instead of using phase differences (of which frequency bin?), a better method is to use windowing and measure not just the peak FFT bin level but also the ones on either side, then fit these to the shape of the window function to work out the signal frequency offset from the bin centre frequency.

e.g. if centre bin is at F1 and bins either side are equal level, frequency equals F1
       if bins F1 and F2 are same level, frequency is (F1+F2)/2

Ian

That was what I tried when I first started working on the idea about 5 years ago. It wasn't very useful, I found. So I started searching on the web and found Stephan Bernsee's page on Pitch Shifting Using The Fourier Transform. I just use the bin frequency calculation part. It appears to work quite well, see the debug output below. The phase change calculation is done on all bins using the change in phase from each FFT run to the next as compared to the expected change for that bin. There's a very clear explanation on Stephan's page. I haven't yet found another algorithm that will give a fast accurate frequency measurement.

Code: [Select]
Input 1KHz:
Bin, Bin freq, Calculated f, Phase change, Bin level
370, 995.910645, 999.997736, 0.068878, 0.230102
371, 998.602295, 999.973269, 0.073584, 0.551781
372, 1001.293945, 999.985533, -3.071415, 2.542450
373, 1003.985596, 999.987880, 0.069438, 2.513233
374, 1006.677246, 1010.749854, -3.071271, 0.465892

Input 440.6Hz:
Bin, Bin freq, Calculated f, Phase change, Bin level
162, 436.047363, 440.533345, -1.671452, 0.026782
163, 438.739014, 440.560611, -1.663091, 0.168153
164, 441.430664, 440.552464, 1.478087, 0.568718
165, 444.122314, 440.549081, -1.662557, 0.452225
166, 446.813965, 451.315157, 1.486924, 0.052688
Logged
Bill of Old Radipole Village
Pages: 1 [2] 3 4   Go Up
 


Melodeon.net - (c) Theo Gibb; Clive Williams 2010. The access and use of this website and forum featuring these terms and conditions constitutes your acceptance of these terms and conditions.
SimplePortal 2.3.5 © 2008-2012, SimplePortal