Mytek Stereo 192

[Miska]

Lowering Pure Music's gain by a few db cured all ills. Can't be done in DSD though.

 

I guess that one depends on the player though...

[ted_b]

I guess that one depends on the player though...

 

Show off!!

[orgel]

However, i would not simply assume that, after turning down the gain and hearing no apparent distortion, that just because the DSD leds show red that it is proof of clipping. I'm not 100% sure that is the case. (Once again i will ask Michal ) I see red meters a lot and do not hear anything close to clipping...

 

I'm pretty sure Michal mentioned somewhere (which of course I can't find right now) that although v1.6 of the firmware enabled the meters for DSD, the meters weren't calibrated yet for DSD and that this would be addressed in a future firmware update.

 

I too see a lot of red LEDs when playing DSD, but no clipping ... since I moved to a pre-amp that could better handle the Mytek's output levels. Before acquiring the "good" pre-amp, though, using software attenuation, or either the digital or analog volume control on the Mytek addressed the clipping issue pretty effectively.

 

--David

[tailspn]

I too see a lot of red LEDs when playing DSD, but no clipping ... since I moved to a pre-amp that could better handle the Mytek's output levels. Before acquiring the "good" pre-amp, though, using software attenuation, or either the digital or analog volume control on the Mytek addressed the clipping issue pretty effectively.

 

--David

 

0dB DSD is 50% modulation. Many classical SACDs go routinely +3dB, and I've seen +6dB on several. If the new metering is simply scaled to 0dB PCM, you'll see the red LED allot. The Mytek Stereo 192 has an output stage similar to the 8X192, and easily has sufficient headroom to accommodate +6dB DSD.

 

The downstream consumer preamp/amp may be another story.

[indistorm]

Ok, There is still clipping for me. In PCM mode it never reaches the red light unless there is deliberate clipping by the producer (I have made my own sample file with a melody clipping and can confirm this). So in PCM, if the music has clipping issues, it will clip. Now the SACD isos I load, whenever there is a loud passage it results in the red led light flashing with distortion. I have Queen's "A night at the opera" and in Bohemian Rhapsody, especially the middle part, there is clipping non-stop with distortion.. How can I solve this?? I cannot adjust the output volume in foobar, and my volume ctrl is at bypass (set at analog or digital with lowered output does not resolve clipping, only reduces the volume of already present distortion).

[ted_b]

Sorry, my edit time expired. Matt (J River) and I sent Michal emails regarding my findings. He immediately responded and has his folks (Chebon especially) on them. FYI.

 

Tonight I re-reinstalled USBPAL 1.33.10 and now all is well. DoP works with J River on either USB or firewire; ASIO USB works too, and supports both DSD64 and DSD128. Sorry for the firedrill.

[tailspn]

Ok, There is still clipping for me... Now the SACD isos I load, whenever there is a loud passage it results in the red led light flashing with distortion.

 

What does it do when you play a DSD download, as opposed to your ripped SACD?

[mytek]

0dB DSD is 50% modulation. Many classical SACDs go routinely +3dB, and I've seen +6dB on several. If the new metering is simply scaled to 0dB PCM, you'll see the red LED allot. The Mytek Stereo 192 has an output stage similar to the 8X192, and easily has sufficient headroom to accommodate +6dB DSD.

 

The downstream consumer preamp/amp may be another story.

 

Thx tailspn, these definitions are accurate. In our own experience we have not come across the overload issue yet, but maybe because we just assumed SACDs are made within the SACD standard that mandates +3dB max. and we did not have a louder SACDs. If someone can suggest a title that goes over +3dB, we'll try it asap. Another thing that we are trying to address is the actual DSD meter calibration. Since the top LED is 0dB , now level can go to red without distortion as +3dB is still ok. maybe we should consider an optional user calibration choice that will bring the meter 3dB down.... I have to think about how to best address this, need to experiment with bunch of recordings and test signals for couple of weeks. Any suggestions are welcome. Thanks a lot for valuable feedback!

 

best Regards, Michal at Mytek

[dmccombs]

Michal,

What sort of sound change should we expect from this?

What happens if a USB cable that doesn't have the power cut is used? I have a cut power usb cord but I could someday accidentally use a regular USB while testing something.

 

There are also jumpers that allow you to bypass the capacitors

in the analog input signal path. This will extend the lowfrequency

response to 0 Hz. It will also allow DC to pass, so

only bypass the capacitors if you are sure that there is no DC in

your signal, as DC can be harmful. The jumpers are located

behind the output XLR's.

[orgel]

Another thing that we are trying to address is the actual DSD meter calibration. Since the top LED is 0dB , now level can go to red without distortion as +3dB is still ok. maybe we should consider an optional user calibration choice that will bring the meter 3dB down.... I have to think about how to best address this, need to experiment with bunch of recordings and test signals for couple of weeks. Any suggestions are welcome.

 

Would it be possible to have the metering for PCM stay as it is, and have the metering for DSD go to red when the level exceeds +3dB? Maybe I'm alone in this, but that's how I imagined it would work once this is addressed.

 

--David

[Miska]

Would it be possible to have the metering for PCM stay as it is, and have the metering for DSD go to red when the level exceeds +3dB? Maybe I'm alone in this, but that's how I imagined it would work once this is addressed.

 

I think it would be quite inconsistent if the LED that says "0" on top would mean +3 dBFS... Anything exceeding the DSD 0 dBFS level anyway violates the spec.

[joelha]

Hi Ted,

 

Is there any advantage to using DoP vs. ASIO?

 

Joel

[ted_b]

Hi Ted,

 

Is there any advantage to using DoP vs. ASIO?

 

Joel

 

I guess purists would say ASIO is more direct, but no..I see no advantages other than features like the current DSD128 capability.

[indistorm]

Got something to report back on the issue of clipping/distortion on DSD playback. After a bit of research, it wasn't the fault of the DAC but the mastering engineer that remasterd "A night at the opera" by Queen (in 2002 I believe). I mentioned before that Bohemian Rhapsody has said distortion/clipping.. and guess what? It wasn't mastered properly. I can't believe they ruined such an amazing song, oh well, I still got PCM version of it. Another thing I wanted to point out about the Mytek DAC, and this time this is a positive. I am a music producer and have used a few DACs from my DAW to the amplifer and I have to say the clinical nature of Mytek DAC has shown me many flaws whilst producing music. I mean clipping is SO evident, unlike my previous DACs, when there's clipping the Mytek DAC is going to immediately demonstrate it to you. My previous DAC seemed to filter out clipping and static from clipping itself, and I prefer it the way Mytek DAC is showing music: exactly as it was meant to be.

[tailspn]

I think it would be quite inconsistent if the LED that says "0" on top would mean +3 dBFS... Anything exceeding the DSD 0 dBFS level anyway violates the spec.

 

You're quite right. But as a practical matter, many, especially wide dynamic range classical orchestra SACDs exceed 0dB at peaks. There's a minimum amount of increased distortion, even up to +6dB, and many labels/mastering engineers allow the peaks to go over, as opposed to applying limiting.

 

The problem is that many customers equate overall volume level with lack of quality. So the trade off is to master with a minimum amount of compression, to bring the quiet parts up, and let the peaks go over 0dB.

 

Pro gear, like the Mytek is designed for this, and has adequate headroom. But the Mytek Stereo 192 metering was designed for PCM, which by definition, cannot exceed 0dB

[Miska]

There's a minimum amount of increased distortion, even up to +6dB, and many labels/mastering engineers allow the peaks to go over, as opposed to applying limiting.

 

They should just turn down the overall volume instead of limiting or letting it go over the specced maximum. DSD works better at lower levels, sound quality just suffers when levels exceed maximum.

 

You can also always ask what Cookie Marenco thinks about DSD levels...

 

The problem is that many customers equate overall volume level with lack of quality. So the trade off is to master with a minimum amount of compression, to bring the quiet parts up, and let the peaks go over 0dB.

 

That's bad habit and reason why so much compression is used, there's a volume control in the amp to deal with volume. Let CD have all that compression, SACD and DSD should be left uncompressed and shouldn't exceed the max spec level.

 

Pro gear, like the Mytek is designed for this, and has adequate headroom. But the Mytek Stereo 192 metering was designed for PCM, which by definition, cannot exceed 0dB

 

I don't think it matters if those badly mastered DSD files keep the 0 dB LED glowing bright. It's a good warning sign anyway, especially for those mastering engineers for whom it has been designed for. If needed, Mytek could add +3/+6 dB LEDS to some future version instead of spoiling the meter scale for no real reason.

 

(it could be that the current meter calibration is a bit too hot, but I haven't got time to verify it)

[orgel]

(it could be that the current meter calibration is a bit too hot, but I haven't got time to verify it)

 

I think that may well be the case. I see red LEDs on most DSD "albums" -- mostly jazz and classical, including stuff from MFSL, Channel Classics, and other outfits that I wouldn't expect to be introducing clipping-related distortion.

 

While I think Miska has a valid point about 0dB really meaning 0dB on the meters, I still feel I could make the mental adjustment with respect to DSD playback, as long as the "alternative" peak level (+3, +6, or whatever Mytek deems most appropriate) was well documented.

 

--David

[tailspn]

I think that may well be the case. I see red LEDs on most DSD "albums" -- mostly jazz and classical, including stuff from MFSL, Channel Classics, and other outfits that I wouldn't expect to be introducing clipping-related distortion.

 

--David

 

Agreed, and I wouldn't expect them to be "introducing clipping-related distortion" either, and they don't.

 

My suggestion is if people would like to discuss DSD, then it behooves them to learn more than anecdotal knowledge about it before making judgements. Admittedly, it's now difficult to get the supportive specifications today, since so little continuing development is going on. But to add to our understanding, this overview has some merit.

 

http://www.sonicstudio.com/pdf/dsd/SACD_FormatOverview.pdf

 

Page 27 and beyond address the point you were making.

 

I believe the "spec" Miska alludes to for signal levels is the Scarlet Book. But that is for SACD production/authoring, not a spec on its interchange format, DSD. 0dB for DSD is set at an arbitrary 50% modulation. The measured distortion is entirely a product of residual noise, and the accuracy of the modulator, which in turn is a product of the ratio of sampling rate to signal frequency. In fact, lower levels of modulation produce more distortion due to the base line noise becoming a larger percentage of the signal. Up to 100% modulation, level does not enter the picture. Up to +6dB, there is no clipping. It's completely unlike PCM, which has a hard limit at 00000… and 111111… at 0dB.

 

I personally would never advocate for using compression or limiting, or any of the other enhancements used to polish up a less than optimal microphone selection and placement. I'm not a mastering engineer. But before I was to tell one how to do his/her job, something they do day in and day out for many years, I'd learn a bit about their craft.

 

Tom

[Esprit]

What happens if a USB cable that doesn't have the power cut is used?

What about firewire cable w/out power?

[Miska]

I believe the "spec" Miska alludes to for signal levels is the Scarlet Book. But that is for SACD production/authoring, not a spec on its interchange format, DSD. 0dB for DSD is set at an arbitrary 50% modulation. The measured distortion is entirely a product of residual noise, and the accuracy of the modulator, which in turn is a product of the ratio of sampling rate to signal frequency. In fact, lower levels of modulation produce more distortion due to the base line noise becoming a larger percentage of the signal. Up to 100% modulation, level does not enter the picture.

 

Given the experience I've gained while developing the delta-sigma modulator for HQPlayer, I would say it is not arbitrary limit, but a well chosen figure with proper reasoning behind the choice.

 

If you think in PWM terms you begin to have really long pulses when you exceed the said limit. (you get better linearity with shorter pulse widths)

 

As an example, ESS says in their Sabre paper, that they can achieve 90% modulation depth. With a note "All other things being equal, this 90% modulation depth alone delivers a 5.1dB increase in DNR." Pushing modulator to 100% doesn't make any sense and wouldn't allow any amount of DC offset in the source.

 

Baseline noise depends on the noise shaping curve and pulse levels. It is already way below thermal noise anyway even at 50% modulation.

[Miska]

I believe the "spec" Miska alludes to for signal levels is the Scarlet Book. But that is for SACD production/authoring, not a spec on its interchange format, DSD.

 

In any case, DSD was specifically used on SACD and lot of the guidance is thus written for "SACD authoring". For many (most?) cases DSD == SACD when it comes to content/mastering.

 

And if people talk about SACD rips exceeding this 50% modulation level, those SACDs are then violating the SACD specification.

[tailspn]

Two posts on Gearslutz from Bruce Brown of Pudget Sound (the mastering company that does the SACD rips for HDTracks) that may explain orgel's Mytek metering observations:

 

Clipping DSD ? - Gearslutz.com

 

 

Quote:

SACD levels are about +3dB hotter than regular CD's. Try ripping a hot SACD and then converting to PCM.. the whole album is clipped.

I had to have someone rewrite an EPROM to attenuate the DSD signal down to PCM levels so the conversion wouldn't clip.

 

Quote:

I have ripped approximately 2000 or so SACDs. When converted to PCM, all but about 25 had clipping. (PentaTone, Chandos, MDG, Harmonia Mundi, Naxos and others). I wasted another few months re-ripping.

We were using EMM Labs equipment and the clipping was discussed with Gus Skinas and Andreas Koch. Their recommendation was to attenuate the DSD signal by 3.0dB. So my blanket statement applies in this situation. The client does not want clipping and it was not unneeded. I'm not blowing smoke... as you say.

 

The DSD and PCM levels worlds are quite different, and when converting one to another, these differences must be accommodated. My guess is the Mytek DAC, when playing a DSD file lives in the DSD levels world, while the Mytek metering lives in the PCM.

[tailspn]

In any case, DSD was specifically used on SACD and lot of the guidance is thus written for "SACD authoring". For many (most?) cases DSD == SACD when it comes to content/mastering.

 

That's a very interesting thought you raise. To what degree (if any) does the SACD Scarlet Book spec diminish DSD. More importantly, to what degree do DSD downloads allow an improvement in sound quality over the restrictions of mastering for SACD?

 

Certainly DSD 128fs will have no such spec.

[Miska]

I think Grimm AD1 manual puts it best way in simple terms:

"DSD does not have a defined clipping level like PCM. 0dB DSD is defined as 50%

modulation index (100% modulation index means a constant ʻ1ʼ or ʻ0ʼ output sig-

nal. Hence, 50% modulation means a maximum of 75% or a minimum of 25% of

output samples are ʻ1ʼ). In practice the performance of all 1 bit sigma delta modu-

lators deteriorates very quickly with signals above 0 DSD. One problem is that the

high frequency noise increases severely. The standard for SACD production, ʻScarlet

Bookʼ, demands for a maximum HF noise. The AD1 will conform to Scarlet Book

with signals up to +1.8 dB DSD. Nevertheless we strongly discourage you to drive

the AD1 that hard. First, performance of any DSD AD deteriorates above 0 DSD."

 

 

I think same rules should be applied to DSD128 too. At least I am applying overload detection that way. It doesn't make any sense to have hundreds of only 1's or 0's in row, conversion precision and linearity will just suffer.

[tailspn]

Yes, and their DSD FAQ gives even a further explanation:

 

http://www.grimmaudio.com/whitepapers/dsd%20faq.pdf