Can someone pls post the sample rate of an MP3 which plays back @ proper speed on your Fuze?

@donp wrote:

My clip plays 44.1 files of various formats (flac, mp3, ogg/vorbis) about 0.7% high in pitch.  I haven’t tested any other sample rate. 

Please explain how you estimated .7%    Could that be a perceived difference between speakers or headphones?

@niko_sama wrote:


@donp wrote:

My clip plays 44.1 files of various formats (flac, mp3, ogg/vorbis) about 0.7% high in pitch.  I haven’t tested any other sample rate. 


 

Please explain how you estimated .7%    Could that be a perceived difference between speakers or headphones?

 

I detailed this in the other thread.  I made a .wav file with just a 1000 Hz sine wave, then converted to flac, ogg/vorbis, mp3, and CD.  IF you are used to tuning instruments to each other, or have studied radio engineering, you know that when 2 sources are out of tune with each other they produce a beat frequency equal to the difference, and you retune one instrument until there is no beat.  I did the same thing with a bunch of players.

 The CD player, DVD player, and computer were all very close, like 2 beats per minute out of the 1000/second signal, no difference whether the computer was playing the original wave file or any of the compressed formats.  Error roughly 1 part in 30,000 or  0.0033%

My portable players, Sansa e270 and Neuros, were both about 1/3 of a Hz off, 3 seconds per beat. Error roughly 1 part in 3000 or  0.033%

 The Clip was 7 Hz off.  I checked that by playing a 1007 Hz tone on the computer that was in tune with the “1000 Hz” tone on the Clip, which also tells me the Clip is playing high/fast.  7/1000 is 0.7%

 

 

 

@donp wrote:


@niko_sama wrote:


@donp wrote:

My clip plays 44.1 files of various formats (flac, mp3, ogg/vorbis) about 0.7% high in pitch.  I haven’t tested any other sample rate. 


 

Please explain how you estimated .7%    Could that be a perceived difference between speakers or headphones?


 

I detailed this in the other thread.  I made a .wav file with just a 1000 Hz sine wave, then converted to flac, ogg/vorbis, mp3, and CD.  IF you are used to tuning instruments to each other, or have studied radio engineering, you know that when 2 sources are out of tune with each other they produce a beat frequency equal to the difference, and you retune one instrument until there is no beat.  I did the same thing with a bunch of players.

 The CD player, DVD player, and computer were all very close, like 2 beats per minute out of the 1000/second signal, no difference whether the computer was playing the original wave file or any of the compressed formats.  Error roughly 1 part in 30,000 or  0.0033%

My portable players, Sansa e270 and Neuros, were both about 1/3 of a Hz off, 3 seconds per beat. Error roughly 1 part in 3000 or  0.033%

 The Clip was 7 Hz off.  I checked that by playing a 1007 Hz tone on the computer that was in tune with the “1000 Hz” tone on the Clip, which also tells me the Clip is playing high/fast.  7/1000 is 0.7%

 

Awesome work,   I was just reading the other thread with interest

DonP your quantitative analysis is the best, thanks for putting in the work to help us all understand this problem better.

Much better than ppl simply saying “it’s not that bad, live with it” =P

With your quantitative analysis, now we know for sure that this issue is fully ridiculous, and inexcusable on Sandisk’s part, assuming that every Fuze performs similarly in this respect.

Not every one does. It is starting too look at least to me like the issue is hardware related, and since very few have reported this issue it leads me to believe that there may have been a bad batch. Max what version is your fuze? (The first number in the firmware is either 01 or 02)

@maxplanck wrote:

DonP your quantitative analysis is the best, thanks for putting in the work to help us all understand this problem better.

 

Much better than ppl simply saying “it’s not that bad, live with it” =P

 

With your quantitative analysis, now we know for sure that this issue is fully ridiculous, and inexcusable on Sandisk’s part, assuming that every Fuze performs similarly in this respect.

1)  It isn’t that bad, and all his analysis proves is that he as WAY too much free time, and has no clue how to do research.  His sample of players is like trying to decide a national election with 1% of the results of one state.  Anyone that makes this big a deal about so insignificant a difference is totally missing the enjoyment of the music itself.  I really feel sorry for him that he can’t just listen.  Adrian Monk would be proud.  I suppose now he’ll begin his “analysis” with other brands of players, and posting his “results” on their forums as well.  Maybe he call the radio stations and demand they calibrate their equipment as well.  Then he can attend every live concert to verify that each performance is tuned to his very own 1000Hz tone.

2)  So now that you “know”, what are you going to do about it?  Answer:  Nothing.  Unless you count the incessant whining.

 DonP found a way to conclusively prove the issue exists and to what degree on his Fuze…props to him for that. However, not everyone has the knowledge,inclination and/or equipment to do what he did. 

 I’m inclined to agree with Conversion Box that it doesn’t seem to be ALL Fuzes, but that a certain batch of them has the problem. All I can say is that if the problem exists on my device, I can’t detect it . And can it really be that bad?Aren’t there better things to worry about, like, will we still have our jobs tomorrow?

LOL.   I just love these discussions :P   let the geeks have some fun.   :)   Those that don’t care about the pitch/tone/timing error don’t have to read :P    LOL

The test is extremely easy for anyone to perform. Download this MP3 file: 

http://www.mediacollege.com/audio/tone/files/1kHz\_44100Hz\_16bit\_05sec.mp3

Copy the file to your Fuze. Play the file on your Fuze, and record the audio output from the Fuze to your computer. Upload that recording somewhere, and post a link to it here. I will do the rest.

jkj, DonP’s quantitative analysis has provided us with one data point for one unit. Of course we need more data points in order to determine if other units perform similarly. If we really wanted to do this properly, we would take multiple data points from each unit under different operating conditions (i.e. different temperatures, levels of battery charge, etc.)

For you to criticize the fact that he employed a measurement technique and obtained a data point, on the basis that he hasn’t collected data points from other units, makes no sense. Of course he can only measure his own unit, that’s the only one he has access to. He’s posting his results here in hope that others will perform the test, and post their results to compare.

His comparison of his Fuze unit to his other players indicates that his Fuze is off pitch on a whole greater order of magnitude, compared to his other players (at least under his testing conditions, which I assume to be standard conditions under which the Fuze would ordinarily be used.) So in his case, the problem IS quite bad. 0.7% pitch deviation is significant for musicians’ applications, as many have noted in these forums. 

Now that I know, I may refrain from purchasing the Fuze. I don’t own one yet. I will reserve judgement until we can determine how widespread this problem is, and its average magnitude.

Message Edited by maxplanck on 02-15-2009 08:48 PM

Message Edited by maxplanck on 02-15-2009 08:48 PM

Max listen very Carefully. So far no V1 Fuzes have this issue. So we know its constrained to V2. Now since not every person with a V2 fuze has complained or returned it you can safely assume not every V2 fuze has this problem. This means that there is a batch of V2 fuzes out there where the hardware is defective or calibrated wrong. We can safely assume all of this based purely on the two threads about this issue. What we need to do at this point is find out how to check whether or not the manufacturing batch is the same on the players where this issue occours.

Since many users may not ever notice the problem if it occurs w/ their Fuze, I’m not 100% convinced that the problem is limited to a small batch.  I prefer to judge by quantitative analysis, not by people’s ears.  Can you run that file through your Fuze and post it?  If the pitch deviation is << 0.7% I will be more convinced.

Nope. I already ran my fuze. I work in the music industry with the most perfect instruments on the market. When this first happened I tested a wave length that is used to calabrate FM Radio, and Shure IEMS (its the same) by playing it htru my fuze into a dummy head and It was spot on. I would check it again but my fuze is full of music I have to listen to tonight, I cant take anything off.

Edit: I just thought of this if thedifference is 1 or 2 cents off, and as you say most people wont hear it, is it a probelm at all. Not IMO its just an inconveniance.

Message Edited by Conversionbox on 02-16-2009 12:20 AM

So your procedure was to transfer the test signal file to the Fuze, play the file from the Fuze through a pair of IEMs while recording the IEMs’ output through a pair of microphones, then compare that recording with the original test signal file using your ears.  Right?

The problem with using your ears is that we have no way of knowing that your hearing is precise enough to distinguish a 0.7% difference.  That difference is only ~1/10th of one semitone.  If you recorded the Fuze’s output through the dummy head, why not upload it for us?

If you were going to test again, I would recommend ditching the dummy head and instead running the signal from the Fuze directly into your computer’s audio interface.

I would just do it myself, but I don’t have a Fuze (want to see the results of these tests before deciding whether or not to purchase).

Message Edited by maxplanck on 02-15-2009 10:12 PM

@conversionbox wrote:

Edit: I just thought of this if thedifference is 1 or 2 cents off, and as you say most people wont hear it, is it a probelm at all. Not IMO its just an inconveniance.

MOST people won’t hear it. Some people will hear it. Some people may not hear it, but are interested in training their ears by listening to music. If they train their ears using an off-pitch player, then they will be screwed.

Therefore it is a big problem. 

Message Edited by maxplanck on 02-15-2009 09:54 PM

No IEMS Direct line into the coputer which replecates the process. And I dont work at that studio anymore, so I cant get at that recording or I would gladly post it along with the results.

OK, thanks.  If your home computer’s motherboard even has a line in jack (usually the jack is red, and of the headphone type, located on the back of the computer case), you can record the output of your Fuze through this jack using free software such as Audacity if you don’t already have DAW software installed.  Would only take a few minutes  :stuck_out_tongue:  

Test tone file is conveniently located here:

http://www.mediacollege.com/audio/tone/files/1kHz\_44100Hz\_16bit\_05sec.mp3&nbsp;

Has anyone tested a 48KHz file? I have an e200v2 which I have been praying for rockbox to finally get working on… However in the time I have spent waiting I have been trying to chip in and learn about porting rockbox; as far as I know the e200v2 has the same chip as the Fuze does, the Austrian Microsystems AS3525.  I looked at the AS3525 site (http://www.austriamicrosystems.com/eng/Products/Mobile-Entertainment/High-Performance-Microcontrollers/AS3525) where they state “Sampling Frequency: 8-48kHz” for the Sigma Delta (sample rate [SRC]) converters.  Since the SRC (and the rest of the audio signal path; ex. DSP or any processing) probably operate at the highest sample rate supported by the chip I’d assume that anything at 44.1kHz is getting upsampled to 48kHz;  I would try and run a test at 48kHz; I would however I am at school (not living at home) and dont have any cables or anything like that to run the test, just some headphones :slight_smile:  

Until I can get rockbox running and fine tune the eq parametricly I plan to use foobar2000 to ‘convert’ all my music for headphone listening and while converting also process a custom EQ dsp (Electri-Q), a headphone (B2SP) and run SoX (very good sample rate converter) to convert to 48kHz sample rate… Another thing; what good is the eq on the e200 (presumably all the other sansa players) when all you can see is 5 bands; you dont know the frequency or slope of the eq lol…  anyways; hope this all helps or has atleast helped the discussion.

@mp3geek wrote:


 


I found my wife’s Chromatic tuner again and you guys sparked my interest so I made a few new test files using Audacity and tested them. The results are shown below.

  1. 440 Hz sine wave at 44100 sample rate encoded with LAME MP3 - 20 cents slow
  1. 880 Hz sine wave at 44100 sample rate encoded with LAME MP3 - 20 cents slow
  1. 440 Hz sine wave at 48000 sample rate encoded with LAME MP3 - dead on
  1. 440 Hz sine wave at 22050 sample rate encoded with OGG         - dead on
  1. 440 Hz sine wave at 96000 sample rate encoded with OGG         - dead on

 

There seems to be a pattern here. Does anyone want to comment.

 

Note that the real time clock is about 5 minutes slow - but I set it about a month ago.

 

p.s. I can hear a clear difference in pitch between these files.

 

http://forums.sandisk.com/sansa/board/message?board.id=sansafuse&message.id=17849#M17849 


@donp wrote:

 

 

 

 

So what level of precision does “consumer electronics” have?

 I made 1000 Hz  and 1002 Hz wave files in cool edit, converted to flac and ogg/vorbis for the players, and burned to 2 CD’s.  Playing 2 sources at the sime time will give a beat frequency equal to the difference in the tones’ frequencies (as anyone who’s tuned 2 instruments against each other knows).  The 1002 hz file was a sanity check to make sure I could hear the beat when 2 sources are known to be off… worked in all cases.

 Playing Cool edit against foobar on the same PC, flac, wav, or ogg, no beat (everything consistantly in tune) 

CD player vs DVD player (both fairly cheap consumer models, different brands and about 20 years apart in age) - roughly 30 seconds per beat  (1/30 hz), for an difference of 1 part in 30,000 or 0.003% 

CD player or Cool Edit vs  Sansa E200 (rockbox playing flac)- beat of ~1/3 Hz, error 1 part in 3000, or 0.03%.

Cool Edit vs Neuros player (playing ogg file) - Also about 1/3 Hz, so 0.03%

 Cool Edit vs Clip (playing flac) - beat of 7 Hz, error about 1 part in 140 or 0.7%   I checked this one by generating a new wave of 1007 Hz, which was in tune with the Clip playing the “1000 Hz” file.

So the Clip’s pitch error (and presumably play speed error) is over 20x worse than my other portables, and 200x worse than the difference between my CD player and DVD player.

So the typical standard for consumer electronics (including an older Sansa model) really is a lot better than the current lot.

http://forums.sandisk.com/sansa/board/message?board.id=sansafuse&message.id=17849#M17849

From the test results posted so far, it looks like 44.1kHz files are the ones being played back too slow.  DonP’s measurement indicated a ~0.7% slowdown. MP3Geek’s measurement (probably cruder than DonP’s due to his methodology) indicated a ~1.15% slowdown (~20 cents).

@My initial suspicion was the same as yours, that 48kHz files were being played back @ 44.1kHz, thus resulting in slowdown. However, if that were the case, the slowdown would be 8.125%, which is far larger than the amount of slowdown indicated by the test results. This suspicion is also defeated by the fact that 48kHz files apparently play back @ normal speed, the problem only occurs when 44.1kHz files are played back (I’m still awaiting more solid confirmation of this though, we’ll know when we see more people post their test results).

The fact that MP3Geek’s Fuze can play some files back at the correct speed indicates that the internal clock mechanism is probably ok, so I think that this problem can probably be corrected via firmware (don’t know for sure though, since I don’t know the device’s internals). 

DonP, the sample rate of your test file was 44.1kHz, right? 

Message Edited by maxplanck on 02-16-2009 01:46 PM

my theory is/was that its trying to preform a sample rate conversion from 44.1khz to 48khz; its not trying to play it back at 48khz, something in the sample rate converter could be ‘messed up,’ since 44.1khz to 48khz is a ‘messy’ conversion (not evenly dividable as something like 22.05khz ->44.1khz is).  be it too little processing power, or some other reason the chip is not converting the sample rate fast enough or maybe 0.7% is the difference between an ‘even’ division in the chips clock cycles…

That is why I am using my computer to convert my tunes to 48khz since it seems to be the sample rate the chip is using; sample rate conversion (accurate sample rate conversion) takes alot of processing power if you want to do a good job and maintain quality (granted, since we are dealing with mp3 the quality will probably not show up, might as well not make it worse than necessary though :P)

http://src.infinitewave.ca/ 

See this site for a comparrison of software sample rate converters (as an example of what I mean in the quality difference) 

EDIT: another example reguarding to the chips SRC; have you ever watched video on a computer that is way to slow to play it? you can either watch the video choppy or you can watch ‘smoothly’ with less frames (frame rate goes down).  I have a feeling that there may be some sort of ‘waiting delay’ that the chip is producing 0.7% of the time; like if the dac is waiting for the processor… 

Message Edited by Chesteta on 02-16-2009 02:33 PM

It sure would be weird for the Fuze to be designed to play everything @ 48kHz, thus requiring a messy sample rate conversion from 44.1kHz for almost every file, since 44.1kHz is by far the most common sample rate for audio files (someone pls correct me if I’m wrong about that). I have a hard time swallowing the idea that even a drunk engineer would design the system that way.

But if there’s anything in this world that tops human thoughtlessness, it’s human thoughtlessness. With that in mind, those theories sound plausible to me. 

I may be missing something, I have no experience with hardware design.

In any case, let’s see what peoples test results are. That will shed some more light on the problem and hopefully reveal it to be a bad batch rather than a universal problem with the Fuze, and thus we can hopefully just avoid the bad batch and leave this problem for the guys @ Sandisk who are actually getting paid for their time to figure out.

Message Edited by maxplanck on 02-16-2009 03:26 PM