Pavel
Thank you for posting the identity of the two files and elaborating on their distortion signature.
Out of curiosity, on how many octaves span you find this technique works?
This is strange. Have you found out why?
I had a somehow similar issue (not with MAudio) but the difference was 9.17% (the percentage relation btn 44.1 & 48). The problem was with the PC settings of the SPDIF device.
George
Thank you for posting the identity of the two files and elaborating on their distortion signature.
Out of curiosity, on how many octaves span you find this technique works?
This is strange. Have you found out why?
I had a somehow similar issue (not with MAudio) but the difference was 9.17% (the percentage relation btn 44.1 & 48). The problem was with the PC settings of the SPDIF device.
George
For me not many, just a couple of octaves above A440 and same or less below. Lots of compromise on a smaller piano. I use an electronic tuner to set A440 and that's useful as a check higher and lower up the keyboard to. Bass notes are the hardest for me to get correct.
A really good book and practical on the subject is Jerry Cree Fishers "Piano Tuning, A Simple and Accurate Method For Amateurs". One I can thoroughly recommend.
Thank you Mooly
https://archive.org/details/pianotuning17571gut
(had I have a piano, I would spend my whole time on it)
George
https://archive.org/details/pianotuning17571gut
(had I have a piano, I would spend my whole time on it)
George
Hi John.
Despite confounders, Pavel's two files are subjectively distinctively different, in my experience.
In my testing, I got the same 'signatures' on the two test systems I used (PC speakers and my main system).
Despite some ABX results indicating there is no substantial difference, my ears tell me otherwise.
Mooly's 4way test (SY recording), despite confounders reveals naked opamp
distortion under abnormal load perfectly clearly IME.
I have trained my ears over long career to hear these kinds of distortions, I expect you have also.
Dan.
Yes, it dulls, deadens the sound on my m/c; if the key difference is the quality of the 'sparkle' of the sample then that aspect of the quality just disappears down the gurgler ...
Mooly was thinking of looking at this at one stage ...
The rate converter in Audacity is the one from the latest version of SoX - I've run samples up and down several times in rates on the converter, and finally taken a diff of the start and end files, no prob's at my end,
.
We have an upright here that needs a bit of TLC, and when I get a Round Tuit I intend to give it a go - years ago, a friend really got into this, probably using that very book.
I can cheat a bit, we have the Yamaha keyboard nearby - that should be vaguely accurate ...
Just did a quick re-listen to pick up where the differences between Bolton and Blomley are clear - and there is a section at the end where it is very easy to hear the difference: from 43 to 51 secs, put it on repeat, and switch between one and the other after a reasonable number of cycles for a single version - trivially apparent ...
No SPDIF involved, take an HP frequency standard set to 1kHz and record it at 44.1/24 and 48/24 the two files play back at 1.06kHz and 1kHz using same HP frequency standard. The expected file length problems are also there. The device uses ASIO via USB and I use Audition 1.5 and makes fine 24/96 recordings. Happens on very old XP machine and new laptop.
My conclusion is the clock divider is programmed wrong (really 44.1*1.06 or is that 44.1/1.06?).
EDIT Same thing playing Denon Audio Technical CD 1kHz reference tone - 1.06kHz out.
Last edited:
Having another look at what's contributing to the audible differences in the 43-51 sec segment of the two files, the spectrum "grass" of the two, though of very low value in number terms, is markedly different - do a screen capture of one version, do the spectrum of the other and overlay the two images to perfectly align; then use Alt Tab to flip between them - what astronomers use for picking up changes in star photos. To my eyes, enough variation to be audibly noticeable in the listening.
What is particularly interesting is that Blomley has very significantly less "grass" in the above 22,000Hz region, compared to Bolton - why?
What is particularly interesting is that Blomley has very significantly less "grass" in the above 22,000Hz region, compared to Bolton - why?
I ran into that type of issue on a Via computer interface. Its clock was not that far off and it still caused problems with SPDIF sync. The AES standard is 20 PPM I think, which is loose for a crystal.
A divider that will give both 44,1 and 48 clock chains from a single clock is not simple. You can get chips with it inside but otherwise its easier and cheaps for an audio product to use two crystals. Perhaps M Audio is using gate oscillator thats misbehaving? Actually its unlikely that the playback clock would be asynchronous to the record clock since the usual practice for USB is to use a common clock for them and async from the host computer locked to that clock.
You can get a more capable interface for around $100 on ebay that won't have those issues. There are a bunch of emu USB options which would be quite adequate.
I think its pretty naive to use unsynchronized digital record/play systems when trying to capture the effects of loading. I know its difficult when you start with 44.1 material. You could start with 176.4 or 192 content and pass it through the "DUT" which would move the highest frequency components much further from the nyquist limits.
I could make some tracks like this easily with the stuff on my bench if given some parameters of what you guys want to explore. I have enough hi-res content to work with.
A divider that will give both 44,1 and 48 clock chains from a single clock is not simple. You can get chips with it inside but otherwise its easier and cheaps for an audio product to use two crystals. Perhaps M Audio is using gate oscillator thats misbehaving? Actually its unlikely that the playback clock would be asynchronous to the record clock since the usual practice for USB is to use a common clock for them and async from the host computer locked to that clock.
You can get a more capable interface for around $100 on ebay that won't have those issues. There are a bunch of emu USB options which would be quite adequate.
I think its pretty naive to use unsynchronized digital record/play systems when trying to capture the effects of loading. I know its difficult when you start with 44.1 material. You could start with 176.4 or 192 content and pass it through the "DUT" which would move the highest frequency components much further from the nyquist limits.
I could make some tracks like this easily with the stuff on my bench if given some parameters of what you guys want to explore. I have enough hi-res content to work with.
I went to a piano store in Roseville, CA. Huge building.... with several concert grand types from yamaha, steinway et al. I went around and hit keys on each and only the low notes didnt sound on key to me. So, went to the 'Grands" and did same listening on the 'tuned' concert Grands and others (they have life performances inside this building)....... they all sounded different in the lows. perhaps they had different harmonic/overtone structures(?) and not pure... but the fundemental tuning freq seemed off. This was confirmed at the store --- they will all sound different... they are hard to get and stay in tune on those long low strings/notes..... I was told that for this reason very little has been written (other than F.Liszt) that use those low keys because of tune being off so easily and different sound. And, those keys get left behind in smaller models... which is no real loss.
THx-RNMarsh
Last edited:
when the wires have noticeable stiffness instead of being pure physics vibrating stings with only tension and mass you don't get pure harmonics, the series of overtones will be out of tune
http://www.afhalifax.ca/magazine/wp...aper-harmonie/newdoc/feynman_pt_dec_46_09.pdf
http://www.afhalifax.ca/magazine/wp...aper-harmonie/newdoc/feynman_pt_dec_46_09.pdf
I cannot confirm this. What data are you analyzing? Below links to the spectrum analysis of both files.
http://web.telecom.cz/macura/bolton.png
http://web.telecom.cz/macura/blomley.png
Remember the 50/60ohm divider (10 + 50) at the blomley output. It shifts the noise floor 1.58dB down. Input volume pot is used to keep same level. This introduces errors of order in 0.01 dBs between the test files. Thus perfect null is impossible. However, we are trying to find audible differences.
Last edited:
- Status
- Not open for further replies.