M2tech hiface - found the bug :))

[Charles]

Hi there,

there's a bug in the hiface's data stream when playing from a mac.

 

Don't panic, I think this can be corrected very easily. For your reference I have just copied the email I sent to Marco of M2tech.

 

-------

Hey Marco,,

 

I've being bit-hunting the whole sunday, I finally found the Hiface bug, LOL

 

Apart from the clicks, I knew after about 5 seconds of listening to the Hiface through my Attraction DAC that something very basically was wrong ...

 

I have tested on OSX 10.4.11 and on OSX Snow Leopard. Clicks are gone in Leopard using the latest driver (1.03.3) from your website, but the sound issue was still there.

 

The sound was overly analytical with exaggeration of third harmonics. Some kind of a high-frequency disturbance and annoyance.

 

Looking at the clock-precision of the hiface's S/PDIF stream, and

considering the Attraction DAC's jitter immunity, I thought that this would definitely not be a jitter issue, but would have to do with the actual data being sent.

 

So I captured the data, sent from the hiface with a ESI Juli@ card and compared the ESI Juli@ recording with the original wav file.

 

(I should notice that the ESI Juli@ is 100% bit correct, maybe something for your test rig too ...)

 

Analyzing the data, I first appreceated that the samples were in their correct spots, so there is no asynchronouse sample-rate conversion happening, very good

 

When looking at the sample values, it took a while until I realized that the negative half-waves were represented 100% correctly, while on every positive half-wave, all samples are -1 bit (one bit subtracted).

 

For example, the hiface would send 25978 but the original value would be 25979.

 

With all negative samples being sent correctly, this means that the hiface is generating a zero crossing distortion (very simliar to that of a transistor amplifier), as every zero crossing of the signal will create an offset of one single LSB.

 

Now you can call me crazy, but I felt that there was an issue with the data right from the start, just by listening to the Hiface for a couple of seconds through my Attraction.

 

But to be honest, I did not think that it was just one single LSB, the very Atom of the digital sprectrum.

 

The good news, I also tested bit-accuracy on a Windows XP machine (the 2nd part of my sunday) and voila, your Hiface sends all samples 100% correct :-)

 

But on the Mac, you got this 1-bit zero-crossing bug.

 

However, XP sending correctly means, that it is only a tiny driver issue and you can correct it very easily, so that all your hiface-fans running apple can also benefit from 100% bit-correct playback.

 

Marco, congratulations, you are 99.99% there, now it is very easy to make it 100%

 

Charles

 

 

--

ALTMANN MICRO MACHINES

 

http://www.jitter.de

http://www.altmann.haan.de

http://www.mother-of-tone.com

 

 

 

 

[davidR]

Charles this is very good to hear and I'm sure will be more than appreciated by all the Hiface users out there. An interesting experiment you ran and even better results. Great work!

 

[Purite Audio]

Fantastic news, well done indeed.

Keith.

 

[Mr.C]

I have to say, I haven't noticed a degraded sound, but if it can get better, I'm all for it.

 

[digger945]

Bravo Charles.

Looks like your much more than an ordinary listener.

 

Perhaps you have missed your calling as a computer audio interface/dac designer/builder (if you're not one already that is). ;-)

 

 

 

[Lizard_King]

I am using the Hi-Face without any noise or problems!

 

[pj]

But as you've noted in the other hiface thread this is on your IBM laptop, right? The bug effects the OSX drivers only.

 

Paul

 

[pj]

Perhaps you have missed your calling as a computer audio interface/dac designer/builder (if you're not one already that is). ;-)

 

I'm sure that is a knowing wink

 

http://www.jitter.de/

http://www.mother-of-tone.com/attraction.htm

 

[MartinC700]

I just recieved an email from Marco related to a minor bug with the Mac driver (an occasional stutter start), and he indicated that a new higher performance Mac driver was being worked on, so perhaps the fix is on the way!

 

Like others here, I haven't noticed any real audio problems with the current driver, but any improvement will of course be welcome!

 

Regards

 

[Charles]

Yes, they are working on new Mac drivers since then.

 

Although I got no click and pops under Leopard, I had this 1LSB crossover distortion, which does not happen with the hiface on a Windows machine. It's only on the Mac, and it gives a nasty sound.

 

Marco emailed me, that they identified the crossover bug to be caused by a piece of software that was recommended by apple itself as part of the software development kit.

 

They are now using a different routine, and if I understand Marco's last email correctly, they are still struggling with clicks and pops ...

 

As soon as the next mac update is available, I will test the bits over again, and post results.

 

We have put so much effort in ripping our CDs, so that they are bit-correct representations of the original, and many are putting a lot of energy in making great hi-rez vinyl recordings.

 

So what we really do need is a 100% bit-correct playback, else all our efforts will be in vain.

 

I consider the M2tech hiface a really great product, and I'm eager to get the new driver, to have correct playback.

 

If some hiface user here has a mac & XP computer, he may be tempted to listen to the sonic difference between the bit-correct XP output (driver 1.02) and the mac output with the LSB crossover bug (driver 1.03.3)

 

Charles

 

[BobH]

If some hiface user here has a mac & XP computer, he may be tempted to listen to the sonic difference between the bit-correct XP output (driver 1.02) and the mac output with the LSB crossover bug (driver 1.03.3)

 

Did this a while ago using my XP machine and my son's Mac and, though we didn't have either the tools or the skill to identify the cause of the problem, there was no doubt at all that a problem existed. I agree entirely with your assessment of the HiFace under the two OS's and look forward to hearing that the Mac side is fixed and achieving the same standard of playback as is currently available under Windows.

 

I've never owned a Mac but, it has to be said, gadget lust has hit overdrive with the dawning of the age of the iPad - maybe a Macbook Pro to go with it is on the horizon!

 

[MartinC700]

Hi Charles,

 

Thank you for the update (and for discovering the Mac bug!). I will try the Windows setup tonight. My first attempt was not sucessful, using Foobar into the HiFace on XP from a HP notebook, the sound sounded inferior to the Mac. I believe that my setup was not correct, so I will try again.

Regards

 

Martin

 

[Lizard_King]

Media Monkey with the Kernel Streaming Plug In (steve Monk) sound better than Foobar!

 

[MartinC700]

I just performed a quick comparison of a PC (HP Laptop) / Foobar 2000 / Hiface setup with a Macbook Pro / iTunes / Hiface: it was interesting.

 

The PC setup was to my ears smoother and more mellow (analog sounding?) than the Macbook. The Mac did seem more detailed but slightly more agressive in presentation. The ideal sound (to me) would combine the strong points of both, i.e. the detail of the Mac setup and the slighly melow PC presentation. I am hopeful that the new hiface drivers may do just that !

 

 

 

[Charles]

The detail and the aggressiveness that you have perceived (hiface-mac), are not separable, in this case, they go together.

 

What you perceived as detail, was a function of the distortion caused by the data error causing the zero-crossing distortion.

 

The third harmonics caused by the uneven zero-crossing give you the sensation of detail, however it is not detail which is on the recording as you cannot create detail by modifying the data of the original.

 

On the other hand, the more analog sounding representation from the hiface connected to your PC, was just correct data.

 

Because it sounds softer (compared to the bugged mac playback) you can listen deeper in the long run, much deeper.

 

And full detail can only be revealed in a setup that sounds so good, that it enables you to go deeper and deeper into the music.

 

If its really good, there seems to be no end to this, LOL

 

Charles

 

[MartinC700]

Just had a very interesting email from Marco at M2tech - (what a great company - they actually take the time to answer stupid questions from customers!) - I am sure he won't mind me summarising it.

 

He first mentioned that drivers that fix an audio bug (presumably the one on this thread) will be appearing in a few days :-)

 

Perhaps more interestingly he mentioned that some of the issues between Mac and PC are that with OSX, regardless of software settings, two data conversions are carried out by the system, apparently comprimising bit perfect output. If this is the case (and I certainly have no reason to doubt Marco), then perhaps this gives us all a clue as to why software like Amarra might actually be beneficial if it can bypass the OS issues. Marco mentioned that for this reason M2Tech prefer PCs.

 

Regards

 

Martin

 

[Charles]

... where the bug actually is happening.

 

This OS-X is a complex thing.

 

However, my Mac mini's Toslink output does give correct data, so I just don't know whether this is a system issue ...

 

I do not think that additional software (like Amarra) will be needed to give bit-perfect output.

 

I am confident that the m2tech team will get it all working real soon, let's send them some good vibes,

 

Charles

 

 

 

[Bob Stern]

Marco at M2Tech seems confused regarding data loss caused by the two conversions in Core Audio.

 

I assume the two conversions he refers to are that Core Audio converts all audio to 32-bit floating point, then converts it back to whatever format is specified for the output device. This is a bit perfect process for 24-bit PCM data because 32-bit floating point consists of a 24-bit mantissa (including the sign bit) and an 8-bit exponent.

 

CA member "Ray W" and Kent Poon both proved that 24-bit playback from iTunes is bit perfect:

http://www.computeraudiophile.com/content/Playback-Engines#comment-22015

http://www.designwsound.com/dwsblog/?p=1591

 

On the other hand, it sounds like Marco found a bug in some sample code recommended by Apple or in a specific Core Audio framework.

 

[57gold]

Any word from Italia that the "bug", graciously found by man with the tech capabilities above, has been addressed?

 

Am a recent Apple convert. Great products. And would like to use the HiFace to BNC to the back of my Wadia 861i.

 

Have spent years getting the Wadia to sound pretty freaking good. Virtual Dynamics power cord, Mapleshade very thick maple base with huge brass footers...all with great effect, 3d imaging, smooth-extended highs, huge controlled bass, body and texture to music.

 

Heard an Ayre/Wilson system a couple of weeks ago, USB DAC with top pre-amp and mono amps ($30+K of cool stuff) through Wilson monitors...and my system more musical and resolving with Aerial 10Ts and McCormack DNA-500. Knopfler was less there and instrumental background more blurred, diffuse with all the new equipment. Have to say it was at a dealer prepping for an Ayre demo versus a tuned with cables (Yamamura and Mapleshade....).

 

HiFace or Legato or Wavelink or Weiss interface amy be ticket. Also may try TC Twin Impact. Cheapest entry is HiFace, if it's fixed for Macs.

 

[Charles]

I tested the OS10.6 and 10.4 versions of the actual drivers, and still got the same bug. Every positive sample is off one bit while all negative samples are correct.

 

I think there may be 2 ways to correct it.

 

1) find the piece of code where the wrong calculation is happening (that would be desirable).

 

or 2) just fix it afterwards somehow like:

if sample > 0 then sample = sample + 1

 

I am positive that we will get a bit-perfect MAC driver real soon, which would make the hiface a full recommendation, as it already is for Windows users.

 

Charles

 

[aubrey]

Marco appears to have cracked it with not one but 2 drivers...I have tried out both and both are stable on my 2008 mac book. blessed relief. He said he will not post it (the latest one) as yet (I assume its some kind of Beta and maybe he would like to be absolutely sure) because he wants to post it in serial order. Apparently this driver is number 2 down the line. But he has no objection to my sharing it.

 

If anyone wants a copy pl send me an email at aubrey[email protected]. I do not know how to attach it to this message.

 

Best regards

 

Aubrey

 

[Charles]

and found it to be bit-correct.

 

zero crossing bug gone !

 

Very good,

 

Charles

 

[MartinC700]

Oohh.... A new release of PureMusic and new hiface drivers (very soon anyway) - life is good!

 

 

 

[jonmarsh]

shines a little brighter, eh? I've been waiting for this news about the Mac driver, given how Marco and his team work, I believed it was just a matter of time. Easy to work on other things and wait until this is ready- now, once I have the driver in hand, time to install that and the newest Pure Vinyl 3.0 beta. I've also been finishing up some updates to my Mac Mini, though now I'm pondering upgrading to a Mac Pro by the end of the year, due to some other developments on the video front.

 

Still, it's great- feels like a number of things are just falling into place.