MZ-RH10 & Mp3 Playback

[Qwakrz]

I have just recieved my NH10. I have a vast number of my CD collection already in MP3 formats as I use an in-car MP3 player.

I have encoded most of the tracks with LAME using my own preset of Compression Quality 2 and Encode Quality 2. The difference between the original and the MP3 is negligable when using my headphones.

I put these MP3's onto the NH10 and......

Talk about having no high frequency responce.

It seems that if the unit plays back MP3's it uses a roll off of about 1 or 2db per octave from 8KHz upwards. I tested this by encoding some music at 320Kbps and using that as a compair. Still *censored*

I then used sonic stage to convert the MP3 (same as on unit) into Atrac3+ (256 & 64) and both sounded LOADS better than the MP3, had better frequency responce and more life.

Why did sony decide to cause a MAJOR frequency roll off when using MP3. This could also explain why some people cant tell the difference between MP3's at different bitrates when playing back on this unit.

SONY YOU HAVE FAILED YET AGAIN

Edited May 24, 2005 by Qwakrz

[Michael1980]

What eq options are there for mp3 playback?

[jonny mac]

They can't make MP3s sound good - they're hoping people will realise they sound awful and re-rip it all to ATRAC. Then Sony will own them

It would also undermine the higher quality claims of ATRAC, something I believe but people should be allowed to make up their own minds

[Christopher]

Surprising it's like that even at 320kbps..what headphones do you have btw?

[Qwakrz]

You can overcome the poor frequency responce by turning the trebble up (Only after the hack). It does not seem to make it sound bad, infact it sounds like it should have done.

-edit- I have a set of sony's that clip over the ear, dont know the model as had them ages and they have worn abit.

I was also supprised that VBR at an average of 160 & 320 sound the same. I would think that it may even be that 64kbps sounds ok but personally that would just sound bad.

Edited May 24, 2005 by Qwakrz

[Christopher]

Are they Q55's? Not the best to test with..

[bug80]

Qwakrz, you're not the first one to report this. It makes me worry and it looks like Sony has either made an error or degraded the decoder on purpose, in favor of their own ATRAC format. The latter case would be a really bad situation.

link to topic where rubadubdub mentions bad MP3 playback quality

link to central topic about MP3 vs ATRAC playback on 2nd gen units.

Edited May 24, 2005 by bug80

[arias]

Are they Q55's? Not the best to test with..

←

Are the Q55s THAT bad? That hurts

I have no freaking idea where to find a pair of good earphones in stores in Toronto, and I have no credit card to get stuff online. Argh.

[atrain]

if you have a standard debit style bank account you can still use paypal.

[Peel]

It seems that if the unit plays back MP3's it uses a roll off of about 1 or 2db per octave from 8KHz upwards. I tested this by encoding some music at 320Kbps and using that as a compair. Still *censored*

I then used sonic stage to convert the MP3 (same as on unit) into Atrac3+ (256 & 64) and both sounded LOADS better than the MP3, had better frequency responce and more life.

Why did sony decide to cause a MAJOR frequency roll off when using MP3. This could also explain why some people cant tell the difference between MP3's at different bitrates when playing back on this unit.

SONY YOU HAVE FAILED YET AGAIN

←

It looks like this is true.

I generated 10 seconds of white noise as a test, saved it in different formats (PCM, 256&320k MP3s and 256k ATRAC3plus) and transfered them to my RH10. Then I played it back on the RH10 and recorded the analogue signal (EQ off, max volume) and compared the frequency responses of the original files to the analogue recordings. Relevant results:

(as you can see, the analogue signal was recorded at a lower level than the original, but that's not important)

red line: the original PCM (it's mostly covered by the yellow line; look on the right side of the graph). Flat frequency response, of course

green line: analogue recording of the PCM file -- there's a bit of high-frequency loss above 20kHz but not an extraordinary amount. Otherwise the frequency response conforms very closely to the original digital file. This is what I would expect.

yellow line: the orignal 256k MP3 file. As is normal, there is a sharp frequency cutoff above 19kHz or so; otherwise maintains the frequency response of the PCM

blue line: analogue recording of the MP3 file. Unlike with the PCMs, the frequency response here is significantly different compared to the original MP3.

Here's a logrithmic view:

So as you can see, on MP3 playback there seems to be a high shelf filter starting around the lower treble area.

By comparison, although these aren't shown in the above graph, the filter seemed to be the same with the 320k MP3, and it was NOT present in the ATRAC recording.

I don't think there's any explanation other than that Sony is intentionally degrading the playback quality of MP3s. The (possible) argument that the filter is intended to improve perceived audio quality by reducing some of the artifacts that are present in the treble frequencies in MP3s is not believable for many reasons, including:

- the cutoff point of the filter appears to be constant, rather than varying according to the MP3's bitrate

- the type of treble artifacts I have in mind happen in the upper-mid treble, well above the cutoff point

- other decoders don't do this; it's the encoder's job to figure out how to handle the frequency curve for maximum quality when it's decoded straight.

- ATRAC (and other lossy formats) have the same sorts of artifacts but ATRACs aren't filtered by the RH10.

- There are (arguably) ways to slightly improve sound quality of decoded MP3s, but I don't think think they involve shelving cuts above 2-3kHz. If anything, it's more common to try to restore high frequencies (EG through excitment of harmonics - not through EQ)

and so on. In any case, nobody wants Sony deciding how their MP3s should sound. This doesn't really affect me, because I listen *only* to MP3s on my MD, and with my custom EQ setting I'm quite satisfied with the quality. But a mix of MP3s & other formats will make you nuts. And IMHO it's wrong on principle.

Anyway, in summary, if you want to check this out, it's actually very easy to confirm -- just take an audio file, convert to high-bitrate MP3. Listen on your computer; they should approximately (or maybe exactly) the same. Transfer to your MD. If you want, transfer an ATRAC version, too. The difference is undeniable; you *will* hear it even if you have crappy headphones & bad ears.

[Breepee2]

Wow, that's .... really Sony...

Could you add Atrac3+ HiSP as a comparison?

[bug80]

Peel, thanks for doing the test! I think it is now proven something is wrong with the MP3 decoder in 2nd gen units (or the RH10 at least), whether it is by accident or on purpose (although I can't imagine Sony introducing an approx. 3 dB/oct lo-pass filter by accident).

[jadeclaw]

And for THAT they took out SP/MDLP standalone recording?

Brilliant!

[MDfreak]

correcting myself

[bug80]

This is realy bullshit... ATRAC and mp3-codecs are not made to encode noise. They are made to encode music and use things as noise masking. With a totally unnatural piece of "music" is will be clear ANY codec will fail in some way.

←

That's not the point. Look at the difference between the original MP3 and the hardware decoded MP3.

[Qwakrz]

MDfreak, we are not compairing the codecs. This has been done to death.

We are however trying to point out that $ony has put a limit to the audio bandwidth that is available when playing back MP3's & only MP3's. This, we believe, is done to lure people across to Atrac.

Why does the same MP3 file sound better when converted to Atrac and the 2 compaired on the 2nd gen Hi-MD's? The proof is above.

[TeddyKGB]

It seems incredible !

If Sony has voluntary degraded the MP3 sounds, How can it despise its customers in a such a way !!!

[MDfreak]

(correcting myself) Did a comparison myself with a real piece of music and came up with the same sort of results (stupid me was looking at the wrong parts of the graph initially).

I also tried to correct the roll-off with the equaliser but sadly that only works partially because the EQ only has the 100, 250, 630 Hz bands that do not need any correction and then there are the 1.6, 4 & 10 kHz bands that can be corrected with 3, 6 or 10 dB.

Judging from the graph I made the following custom EQ for mp3's:

100, 250 & 630 Hz : 0 dB

1.6, 4 KHz: 3 dB

10 kHz: 6 dB

After testing this preset for some time it sounds good although not exactly the same off course, but maybe I like the sound of the mp3 WITH EQ-correction even more than the original sound of the RH10 (but that off-course is very personal).

The only stupid thing about this: people that buy an European RH10 CANNOT correct for this roll-off (luckily I hacked my RH10).

[xispe]

ahahah! another good one! ... damn! :|

[Damage]

Did anyone notice that's even with the MP3 playback, there's a variation on individual track level? That some tracks exhibit this roll-off and others do not?

Someone want to tackle this bit and see if that's true or am I just hearing things?

[Arto Suomi]

Judging from the graph I made the following custom EQ for mp3's:

100, 250 & 630 Hz : 0 dB

1.6, 4 KHz: 3 dB

10 kHz: 6 dB

In this thread:

Hat Trick For Mdcf?

, davidb says that he measured the following frequency response when he set the 10 KHz slider to the max (+10 dB):

1kHz 0dB

2kHz 0 dB

5 kHz +3 dB

10 kHz +7dB

15 kHz +9 dB

18 kHz +7 dB

20 kHz +2 dB

Looking at the graphs earlier in this thread, wouldn't setting just the 10 KHz to the max give a relatively flat response with MP3s? Could somebody measure what the response is with different kind of combinations of the EQ settings?

[marcnet]

How about adding the following test to the listening tests & frequency response graphs:

Take the 256kbps MP3 and convert it to a .wav using winamp, foobar, LAME etc. Then convert the .wav to a HI-SP 256kbps ATRAC and a 132kbps ATRAC.

[Abbstrack]

hm..i wonder what would happen someone were to contact sony as a tech support issue b/c of the less than stellar mp3 playback. i mean if sony is advertising these units as native mp3 playback units, and at the same time degrading the playback sound of the mp3's, i would think as consumers there should be some sort of recourse, esp. if enough people call looking to exercise their warranties on these units.

or am i dreaming?

[TeddyKGB]

In spite of all the defaults of this 2nd Hi-MD generation (no line-out, Sonicstage restrictions,...) I was ready to buy one.

But with this new revelation about Sony degardations , it's too much !

I 'll buy a HD MP3.

[Damage]

In spite of all the defaults of this 2nd Hi-MD generation (no line-out, Sonicstage restrictions,...) I was ready to buy one.

But with this new revelation about Sony degardations , it's too much !

I 'll buy a HD MP3.

←

Mind you, not all MP3s are affected by this. At least with my RH910, some MP3s sounds fine and unaltered, and others exhibit this strange roll-off behavior. The only thing I can pin it on is the encoder used on each of the MP3s.

The only thing I can think of to track this down is to see which encoder (LAME, Fraunhoffer) does this. It should be easy enough. Take your favorite song from CD, encode one with Fraunhoffer, the other with LAME, in various bitrates and see if you can spot the ones that exhibit this behavior.

[MDfreak]

In this thread:

Hat Trick For Mdcf?

, davidb says that he measured the following frequency response when he set the 10 KHz slider to the max (+10 dB): 

1kHz 0dB

2kHz 0 dB

5 kHz +3 dB

10 kHz +7dB

15 kHz +9 dB

18 kHz +7 dB

20 kHz +2 dB

Looking at the graphs earlier in this thread, wouldn't setting just the 10 KHz to the max give a relatively flat response with MP3s? Could somebody measure what the response is with different kind of combinations of the EQ settings?

←

I did some freqencyanalysis of my own (including the +10 dB suggestion) and came up with the following result:

[attachmentid=322]

Here the green line is the original Hi-SP format so that is used as a reference. Then the blue line is the mp3-version that clearly undershoots the original (better sounding) version in the higher frequencies.

Then the purple line is the +10dB suggestion for the 10 kHz band. It matches pretty nice but it clearly overshoots at about 13 kHz and undershoots at about 8 and 18 KHz. Not totally satisfying to me. Therefor I tried another EQ-setting: +3 dB @ 4 kHz and +6 dB @ 10 kHz. You see the result in the red line. Clearly it does not match exactly but it follows the green line more evenly. Also to my opinion it sounds more natural because there is no peak at about 13 kHz.

I hope this is of some use for you people.

[MDfreak]

Mind you, not all MP3s are affected by this.  At least with my RH910, some MP3s sounds fine and unaltered, and others exhibit this strange roll-off behavior.  The only thing I can pin it on is the encoder used on each of the MP3s. 

The only thing I can think of to track this down is to see which encoder (LAME, Fraunhoffer) does this.  It should be easy enough.  Take your favorite song from CD, encode one with Fraunhoffer, the other with LAME, in various bitrates and see if you can spot the ones that exhibit this behavior.

←

I tried a random selection of my mp3's and had the problem with every one of them (I think). Off course it is worthwhile trying if it is an issue with particular encoders but I think that would be real strange.

Hmm, I have to get al list with well known mp3-encoders soon (or programs that use particular encoders).

P.s. maybe you do not hear the problem with some mp3's because of the characteristics of some mp3's. If there are e.g. less highs or less lows in a mp3 the inbalance is hard to hear.

[Qwakrz]

Well I can say that LAME encoded MP3's show the problems as thats all I use.

[MDfreak]

Well I can say that LAME encoded MP3's show the problems as thats all I use.

←

The tracks from which I know which codec is used are LAME to. The other mp3's come from the internet so no idea how they were created.

[Damage]

Quickly off top of my head:

LAME (various versions, but I think most uses 3.92)

Fraunhoffer (the Windows default encoder in part uses Fraunhoffer encoder, this would include things like SonicStage, WMP, etc.)

There are MP3 programs that can identify the codec used to form the track as well.

Those are the two I can think of off top of my head that's used in most of the mp3s out there, with some stragglers still using either XING or BladeEnc.

Hmm, now that's most of it I believe. Easiest way to confirm that it exists is to record a level signal (or what not) across various bitrates (inc. VBR), and do the analysis as done previously. All it would take is time and patience.

Good to know that my hearing is becoming more tone deaf then ever.

[pata2001]

First of all, Lame will cut off the high frequency unless you specify it not to do it. -V1, -preset standard, will do a lowpass of 19000Hz, which is actually correct per those graphs. Preset extreme has a lowpass of 19500Hz. To disable this, you have to add -k after the presets on the command line. This will disable any lowpass and lame will use the whole frequency band. Of course, this negates the purpose of music compression (to ignore/compress frequencies that are not hearable by most people), and you'll end up with a bigger file, or a lower quality MP3.

Oh, and you don't need any fancy equipment. Just decompress the MP3 to wav, and use EAC to do a spectrum analysis, It will display the same cut off on the 19000Hz frequency range.

Edited June 3, 2005 by pata2001

[bug80]

First of all, Lame will cut off the high frequency unless you specify it not to do it. -V1, -preset standard, will do a lowpass of 19000Hz, which is actually correct per those graphs. Preset extreme has a lowpass of 19500Hz. To disable this, you have to add -k after the presets on the command line. This will disable any lowpass and lame will use the whole frequency band. Of course, this negates the purpose of music compression (to ignore/compress frequencies that are not hearable by most people), and you'll end up with a bigger file, or a lower quality MP3.

Oh, and you don't need any fancy equipment. Just decompress the MP3 to wav, and use EAC to do a spectrum analysis, It will display the same cut off on the 19000Hz frequency range.

←

Of course, but that is not the issue here. The graphs show that the RH10 cuts off high frequencies that are present in the original MP3 file. A decoder should not do that.

Switching off the low pass in Lame is not recommended, by the way. The presets are optimized for giving the best overall results, giving a certain bitrate.

[dex Otaku]

lame and other encoders sometimes use [especially with <160kbps CBR rates and VBR] an ATH filter that basically drops everything above 15kHz. If you do a spectrogram of a song encoded this way, you'll see the high end wisp in and out as sounds exceed the ATH [absolute threshold of hearing] above 15kHz ..

This shouldn't affect things in the way the RH10 is, though.

[MDfreak]

Hi people,

Got myself the program "EncSpot" today to figure out which codecs where used for the mp3's I have. I came to the conclusion that about 90% was made with some version of the LAME codec. Luckily I also found some other encoders.

Now I tested the FhG (Fraunhofer) Fastenc @ low quality versus Hi-SP. It was an mp3 with a bitrate of 256 kbps and got these results when playing the original and Hi-SP version on my RH10:

[attachmentid=328]

Hi-SP: Green line

Mp3: Blue line

Clearly when playing mp3's with the FhG codec on a RH10 the roll-off is also there.

I will try to do some analysis of some other codecs soon.

Adding some results for the "Xing new" codec @ 192 kbps (according to EncSpot):

[attachmentid=330]

Still same result, the blue line is Hi-SP, the green the original mp3.

So I tend to conclude that it is a problem with ALL codecs but that it is not equally noticable with all mp3's because some mp3's have both highs and lows and some have more highs and some more lows.

[Qwakrz]

I will just re-make the point of our little discussion as a few people seem to not notice what we have proved.

We have taken an MP3 & coiped it onto a 2nd gen Hi-MD (e.g. RH10). We have also taken a Hi-SP copy of that MP3 (converted the MP3 track into Hi-SP using SonicStage) and transfered that onto the same Hi-MD, this is done so any problems in the MP3 should be carried across into the Hi-SP track as well. In the perfect world the 2 should sound as near as damn it the same, but......

What we are finding is that the frequency responce is limited when playing back the MP3 and not limited when playing back the Atrac file (remember the Atrac we are playing started off as that first MP3 file originally) so the unit must be falsely limiting the MP3 playback bandwidth to make Atrac look good.

Hope that clears things up.

All my home encoded MP3's are done in Lame with the -k switch unless I need to limit the size of the files. I also dont use presets as I prefer to specify the max & min bitrates for VBR as well as quality and speed settings.

Edited June 3, 2005 by Qwakrz

[Tea&Crumpets]

Hi everyone,

Shame about Sony not being perfect

I was thinking, maybe the whole idea with cutting off the top-end freqs for MP3 is an attempt to make the compression less apparent... you know how high compression can make cymbals sound bad etc. well, if you make it quieter then it is less noticeable. Might be a good thing! Anyway, I suppose we should at least have a choice in the matter.

Sorry if this has been suggested already.

T&C

[Arto Suomi]

I'm still not 100% convinced that Sony did this on purpose. If that were the case, it would be logical that also NW-HD3 and NW-HD5 had cripled MP3 playback, but they don't. This could also be a bug in the firmware, e.g., the roll-of is meant for low-bit rate MP3's but the code is missing the check for the bit-rate and erronously applies filtering to all MP3s. But I guess we won't know for sure before the next generation of HiMD's come out (if its still there, it's on purpose for sure).

By the way, has anyone tried this on any other 2nd gen devices besides the RH10?

If the filtering is done on purpose, there might be a similar setting as for the Euro output power limiting that could be switched on and of in the maintenance mode.

[bug80]

Hi everyone,

Shame about Sony not being perfect

I was thinking, maybe the whole idea with cutting off the top-end freqs for MP3 is an attempt to make the compression less apparent... you know how high compression can make cymbals sound bad etc.  well, if you make it quieter then it is less noticeable.  Might be a good thing!  Anyway, I suppose we should at least have a choice in the matter. 

Sorry if this has been suggested already. 

T&C

←

You´re right that lower bitrates need more filtering of high frequencies, but this should be done by the encoder, not the decoder. Even at very low bitrates, it is not up to the player to do any filtering. This is also not done with ATRAC, so why would a player do it with MP3 files?

This could also be a bug in the firmware, e.g., the roll-of is meant for low-bit rate MP3's but the code is missing the check for the bit-rate and erronously applies filtering to all MP3s

See the comment above

Edited June 3, 2005 by bug80

[The Low Volta]

I'm still not 100% convinced that Sony did this on purpose. If that were the case, it would be logical that also NW-HD3 and NW-HD5 had cripled MP3 playback, but they don't.

←

you'd think so, but then why wouldn't they include an equalizer that also works for MP3 on the HD5? as stated here (go to 'View all specs'):

    NWHD5HS.EU8

HDD Network WALKMAN NW-HD5HS - 30 GB - Exclusive to Sony Style

  Direct MP3 compatibility - with 40 hours of battery life!     

330,00 € VAT incl. 

    Order now

  Sony Style Exclusive Offer! The brand new NW-HD5H is an ultra-compact 30 GB HDD player, combining great looks with cutting-edge design and an incredibly long battery life. We have thought of everything: not only does the NW-HD5H offer 40 hours of battery life, the battery is removable meaning you can carry a spare or pick up a new one while you're out. The NW-HD5H is extremely user-friendly: sporting an exceptionally large and easy-to-navigate 7-line LED screen, the attention to detail in the design of the player is obvious. Its technologically-advanced 'follow turn' feature means the display is synchronised to your hand movements, flipping from horizontal to vertical depending on your every move. 

30 GB internal Hard Disk Drive memory

Fast music and data transfer via Hi-Speed USB connection

Stores audio, visual and data files (Word, PowerPoint, ATRAC, MP3, JPEG, MPEG etc.)

SonicStage® 3.1 software for easy music management and transfers

Sony's unique G-sensor technology protects the internal HDD even if it falls from a height of 1 m

Direct ATRAC / MP3 compatibility

Extremely long playback time of up to 40 hours with the built-in rechargeable battery

Large 7-line (1.5 inch) LCD with various backlight colours

Ultra-compact size weighing just 125 grams

Supplied accessories: headphones, rechargeable battery, USB cable, AC adaptor, carrying pouch, SonicStage® 3.1

            Specifications 

Warranty 1 year

Colour (audio) Silver

Compatible format ATRAC3plus, ATRAC3, MP3, WMA, WAV

Group WALKMAN

Name HDD Network WALKMAN

External battery case No

Group function Yes

Hold Yes

Line out Headphones, speakers

Remaining battery indicator Yes

Weight (g) 130

6 Band Graphic Equalizer Yes (in ATRAC mode)

...

and Sony's still advertising Atrac3+ @64bps as better than MP3 @132bps...

so I wouldn't be surprised at all if it definitely was intended (just not intended to be discovered )

Greetings, Volta

PS: also note the Sony advertising-crap like "only 125g" and in the specifications "130g"...5g is no big deal, the fact that they lie about it is

[xispe]

unbeleivable ...