tracking levels for DSD

[springer]

OK - so help me out here cuz I just got my levels proper for PCM 24/44.1 tracking at a whopping -16db (my AD converters are loving life now).
I heard that you have to push 16/44.1 a little harder - maybe to -12db or -6db. My mixes are sounding much nicer at lower levels.
Now with my new MR1000, what is the general consensus (rule) with:
1) tracking levels on DSD
2) peak tracking levels on DSD
3) mixing levels
Thanks in advance!

[Westmalle]

I exported a programme of 1 hour DSD that overloaded a couple of times to Pcm without gain, and it was distorted.
The DSD itself however was distortion free.
Then I tried converting with -6 gain and -9 dB gain.

This worked.

[MBishopSFX]

Never, ever, NEVER go into clipping in DSD. It's really ugly!

[zenmastering]

The recommended maximum level for DSD is -6dBfs, also known as '0dB SACD'

The maximum level that you can get away with printing on an SACD is +3.1dB SACD, also known as -2.9dBfs. Current recommendations (Derk Reefman, Irwin Jansen paper) strongly suggest not to go above 0dB SACD for any audibly significant time.

The default setting in several DSD to PCM converters is to add 6dB of gain to make up for this 'offset', but if you've gone over '0dB SACD' then you'll get clipping in the resultant PCM file when this default gain is used.

It certainly helps to have the ability to meter the DSD levels properly, either in software or hardware. I use the Sonic Nexstage 'AFC' software. The Korg Audiogate software is sonically a pleasant surprise, but lacks decent metering.

Best Regards,

Graemme

Originally Posted by Westmalle

I exported a programme of 1 hour DSD that overloaded a couple of times to Pcm without gain, and it was distorted.
The DSD itself however was distortion free.
Then I tried converting with -6 gain and -9 dB gain.

This worked.

[RickZ]

Westmalle,

Did you use Audiogate to convert from DSD to PCM ?

I see there is a Gain control, but it doesn't respond. I too have DSD that went over, but had earlier converted to 96/32BitIEEEFloat, and the over wasn't really over, and applying gain reduction on the 96/32 worked fine, no audible problems. It would be convenient to dial in a gain reduction in Audiogate, and have the PCM file under 0 dB, but I can't figure out how to do that.

TIA,
Rick Z

[Westmalle]

Originally Posted by RickZ

Westmalle,

Did you use Audiogate to convert from DSD to PCM ?

I see there is a Gain control, but it doesn't respond. I too have DSD that went over, but had earlier converted to 96/32BitIEEEFloat, and the over wasn't really over, and applying gain reduction on the 96/32 worked fine, no audible problems. It would be convenient to dial in a gain reduction in Audiogate, and have the PCM file under 0 dB, but I can't figure out how to do that.

TIA,
Rick Z

I did one test, and it worked. without gain reduction it had a max level of 0,0dBFS,
afterwards, it was -6 or so.

I did a second recording yesterday.
I'll see if the gain knob still works, and let you know.

[springer]

Thanks for all the good info - but the question remains on the MR1000 specifically (is the MR1000 scale SACD or PCM) what is best recommendation for peaks and tracking levels?

[Westmalle]

I agree that there is better metering than the KORG MR-1000,
but you can put the preferences so that it holds peaks for 0, 4, 10 or ∞ seconds.

[d_fu]

Originally Posted by springer

the question remains on the MR1000 specifically (is the MR1000 scale SACD or PCM) what is best recommendation for peaks and tracking levels?

Lower mids will sound better around -13.67 dBFS, while highs beyond 5k are much better at -9 dBFS (only in Jazz - for classical music, the opposite applies, unless there is a Soprano, which will not sound good at higher levels than -11.3 dBFS)? Also mind that Steinway pianos must not be tracked any higher than - 17, while Bosendorfer easily can handle -10 (unless manufactured in 1987). And the range around -25 dBFs has that strange lack of bass on the MR1000...








Do you seriously think that digital levels will make a significant difference in sound quality (not referring to clipping here)? If so, what next? Sound differences between different hard disks or flash media? Does Vista sound better than XP? Do different driver versions of audio cards sound different? Does a Pentium IV sound better than an Athlon? At which temperature will a DualCore CPU sound best?



If specific units had recommened levels like you suggest, how would one ever be able to record music with any considerable dynamic range, if a lot of the music were to fall outside that recommended ideal level range...? Would you seriously want to use a recorder if only mezzoforte were to sound good?

Don't clip it, keep a good safety margin, and worry about your mics and placement more than anything else...

[springer]

d_fu - you could just say "I don't know whether the MR1000 uses SACD, dbFS, PCM, dbVU, etc... on it's metering system" rather than sarcasm.

And to folly your sarcasm - YES - depending on what the AD converters can handle, things do sound different at different levels. I trust my ears to tell me this. I trust the reading I do from people like Bob Katz, etc.. on here that know what they are talking about. I just want to figure it out for this piece of equipment as I have done for other pieces that don't label or say in the manual.

[d_fu]

Originally Posted by springer

YES - depending on what the AD converters can handle, things do sound different at different levels.

Please provide links to credible statements of that nature. Or provide some sort of even semi-scientific explanation. How would one converter be able to "handle" higher levels than another? Are all bits not created equal? If they are not, what about 1-bit converters? How would you explain such differences? And what exactly does "handle" refer to? What happens beyond the level a converter can "handle"? Distortion? Non-linearity with regard to frequency response? Some other inexplicable change of "sound quality"?
Would this phenomenon be one of the actual converter chip (as opposed to the analog circuitry that precedes conversion)? I.e. would an AK 5385 always sound best at a certain level, regardless of what device it is built into?
Are you referring to maximum levels? Apparently not, otherwise you would not be asking about "tracking levels". So please explain how a converter with a specific ideal "tracking level" range as per your idea will handle music with a dynamic range of 20, 30, or even 50 dB? It would probably sound best with a static sine wave at the ideal level...
How have you ascertained the differences you hear? Did you compare identical material at e.g. -10 and -20 dBFS (peak or RMS?), played at identical SPL (which would involve analog gain or attenuation)?

[Plush]

Earlier Springer wrote "And to folly your sarcasm - YES - depending on what the AD converters can handle, things do sound different at different levels. I trust my ears to tell me this. I trust the reading I do from people like Bob Katz, etc.. on here that know what they are talking about. I just want to figure it out for this piece of equipment as I have done for other pieces that don't label or say in the manual."

Oh boy, sorry but all that you wrote above is wrong. I cannot decipher your english language when you say "handle." What do you mean to describe?

Bob Katz did not write that and either did the book you should read. It is called, "The Art of Digital Audio" by John Watkinson.

The converter is dumb and does not know the difference and it does not know how to make the sound different at different levels. The main think to adjust is the input level to your converter. Most converters allow you to adjust your input levels in relation to a known reference such as .775 volts= -18dBFS.
Perhaps what you mean above in your first posting is that you have adjusted 0VU to be -16 on your converter---is that correct? If so, it is a reasonable level and perfectly workable for 24 bit recording.

Please clarify what you are asking.

As far as the DSD levels on the 1000 are concerned, treat it conservatively just like
pcm recording. No need to jam the levels.

I'm sympathetic to your cause, but learn the proper terms of the discussion.

[d_fu]

Originally Posted by Plush

I'm sympathetic to your cause, but learn the proper terms of the discussion.

I don't think this is about terms at all... The original question was very clearly aimed at finding out at which levels the MR1000 would "sound best". IMHO, the idea of a "general consensus" on voodoo of that kind is absurd...

[springer]

Originally Posted by Plush

The main think to adjust is the input level to your converter. Most converters allow you to adjust your input levels in relation to a known reference such as .775 volts= -18dBFS.
Perhaps what you mean above in your first posting is that you have adjusted 0VU to be -16 on your converter---is that correct? If so, it is a reasonable level and perfectly workable for 24 bit recording.

I'm sympathetic to your cause, but learn the proper terms of the discussion.

OK - I have screwed up my use of terms and jargon. Here is what I mean.

Using my AW1600, I used to track with peaks at (using the AW meters) just below 0. Then I would mix to 2 track, trying to keep the same levels. End product sounded very thin and I used to blame it on the preamps. I then read various posts about bringing down the tracking levels and mix levels until final mastering (from the Bob Katz site and followers here). So I experimented keeping tracking peaks to below -12 (once again on the AW meters). All of a sudden things started sounding like real instruments. I may be wrong in my explanations of this but I just want to know if i should do the same with my MR1000. I will just experiment and report my findings.
I apologize if I have offended anyone's digital knowledge rights here... just trying to gain some insight.

[d_fu]

Originally Posted by springer

I then read various posts about bringing down the tracking levels and mix levels until final mastering (from the Bob Katz site and followers here).

This might refer to the way different (multitrack) signals mix ITB at different levels - the idea (judging by a quick look through related threads) seems to be esp. not to compress indiviadual tracks, but to apply compression after the mix only.

This does not apply to the converters as such, really. You are not likely to be able to determine a difference.

You appeared to be a firm believer in voodoo of the digital-cables-sound-different variety - sorry if I misinterpreted that.

[David Rick]

Actually converters often do have sweet spots. It's usually related to the analog circuitry of the ADC input or the DAC output stages, but there are also digital causes. Here are some possible reasons:

• The designer has put 0 dBFS too close to the supply rails so that some opamp is in distress at high levels.
• The gain-bandwidth product of an opamp is reduced at high output swings. With less loop gain, it is less linear.
• Above a certain signal level, a discrete analog stage transitions from class A to class AB.
• Very low levels put the signal too close to noise floor.
• Some device has a fixed amount of crossover distortion, which is relatively less at higher signal levels.
• High signal levels modulate the supply rails, causing intermodulation distortion.
• A switchmode power supply's operating mode depends on the current demand. It may change operating frequency, skip cycles, or develop a subharmonic loop oscillation. The noise spectrum on the supply rails may change dramatically, and this may affect devices powered from those rails.
• The modulator of a high-order sigma-delta loop is prone to instability above a certain signal level.
• The decimation filter in a PCM converter overloads, even though all its input samples are below full scale.

Many of these causes can be reduced or eliminated by careful circuit design. Therefore one may expect that better converter sets may be less particular about signal level than garden-variety "prosumer" stuff.

In my own experience, my Lynx II cards prefer not to be driven especially hard. In this case, I think it mostly has to do with the opamp supply rails.

David L. Rick
Seventh String Recording

[David Spearritt]

Most of these are gross design faults indicating the machine is broken. One is common sense. In a well designed quality A/D modern recorder, sampling quality is independent of signal amplitude.

[sonare]

I would like to know how folks recording in DSD "manage" their levels? With 24-bit PCM it is simple to peak at -10 to -6 and adjust as needed later in post-production. With DSD any level adjustment after the fact and it's no longer "pure" DSD. Perhaps of little consequence when recording harpsichord but orchestra is a different matter.

Peak limiters on the master bus feeding the DSD device? Perhaps Michael Bishop could weight in on this.

Rich

[MBishopSFX]

Originally Posted by sonare

I would like to know how folks recording in DSD "manage" their levels? With 24-bit PCM it is simple to peak at -10 to -6 and adjust as needed later in post-production. With DSD any level adjustment after the fact and it's no longer "pure" DSD. Perhaps of little consequence when recording harpsichord but orchestra is a different matter.

Peak limiters on the master bus feeding the DSD device? Perhaps Michael Bishop could weight in on this.

Rich

For DSD peaks on Sonoma, I'm shooting for fast peaks to be no higher than +1 on the Max Peak meter. I use a small acoustic noise generator (sort of like pink noise) to set the preamp gain for each mic on-stage. That gets me really close to where I need to be. I'll ride levels somewhat on the board during the session, but avoid that to retain the natural dynamics of the piece. For safety, I sometimes print additional tracks (9-16) from another converter set that is aligned 3 dB down from the first set. If needed, we can edit a peak in from the -3 dB tracks if an "over" occurs. BTW, reference level for the main converter set is -16, 0vu = 1.228 vac.

.

[Larry Elliott]

Originally Posted by MBishopSFX

I use a small acoustic noise generator (sort of like pink noise) to set the preamp gain for each mic on-stage.
.

This sounds an excellent idea - like using a calibrator on measuring microphones. What level does the generator produce and how far away from the microphone do you place it.

Or is this a know level source that you place on the stage and then adjust pre amp gains?

Thanks
Larry

[d_fu]

Here's a nice little tool...

[beyarecords]

Originally Posted by d_fu

Here's a nice little tool...

Thanks, d_fu.

[zenmastering]

Originally Posted by MBishopSFX

For safety, I sometimes print additional tracks (9-16) from another converter set that is aligned 3 dB down from the first set. If needed, we can edit a peak in from the -3 dB tracks if an "over" occurs. BTW, reference level for the main converter set is -16, 0vu = 1.228 vac.

This is exactly what we do with the IsoMike DSD recordings: the second set of four channels is a few dB lower than the main set and on occasion it's saved my butt during editing. Because the two sets of tracks are in perfect sync, the edit is instantaneous. This is how we can avoid going to PCM/DXD in the Pyramix.

We don't have the luxury of riding levels, but that's offset with the fact that the IsoMike recordings are all done in the same hall and we are now pretty good at 'guessing' where levels need to be.

Michael, your method of using an acoustic noise source is excellent advice, thanks!

Graemme

[MBishopSFX]

Originally Posted by Larry Elliott

This sounds an excellent idea - like using a calibrator on measuring microphones. What level does the generator produce and how far away from the microphone do you place it.

Or is this a know level source that you place on the stage and then adjust pre amp gains?

Thanks
Larry

It's a handheld pink noise-like generator with small transducer. There's a pencil taped to the side of it for placing the erasor end on the mic capsule grill. That keeps the distance equal on all the mics. For this generator and the level I need for the DSD converter, we shoot for -26.5 on the Sonoma audio range meters.

.

[Shadow_7]

I recently got the Korg. And between -30 and -12 it sounds best for me. It sounds good -12 to -6, but once in that -6 to 0 range things get a bit odd. Or at least things on the hot end of gain seem to distort to some degree. I like the Korg in that the detail makes my SDC's sound like LDC's in terms of hearing the room and distant conversations. Even hearing yourself after exerting yourself and panting. Where typically on other lesser gear you'd swear the same mics wouldn't pick up anything less than 100dB if it was more than 6' from the mic.

Not that I know what any of those "levels" mean. But my ear tells me the Korg runs a little hot. Too much gain. And it doesn't handle too much well. Not that I've engaged the limiter yet. And it may only be my perception of things. All I know is that piping my Korg MR-1000 to my studio monitors from .wav's generated from standard CD-ROMs yields some nice results. I wish it was my laptops soundcard. Even at 44.1kHz @ 16 bit it sounds awesome. And that's with a 6' RCA cable to 1/4" TS adapters. Which goes to my headphone preamp, then to my Studiophile BX8's. I have some old drum and bugle corps CDs and it's a completely different (and better) experience using the Korg to play back the audio.

They really, really, really need to make a firewire DSD device(and/or PCI). I'd buy it. Even if it was as much or more than the Korg MR-1000. My mom even notices the difference. Which is a non musically trained senior citizen. And she even notices that the difference is that there is better clarity and more detail. Even before I bothered telling her that that's what she should be hearing. And at that time I was just using a 1/4" stereo adpater to a 1/8" for cheap PC speakers from the headphones out.

First impressions though. I've only had the unit about 24 hours. And have done maybe 30 minutes of recording to date. Although that's a lot of data at 5GB of stereo DSD per hour,

[springer]

After testing the unit over the weekend last I found that I could easily record 3 hours on my AA batteries. I did a live mezzo soprano/piano thing this week with the AKG C480's that I normally use with (the now up for sale) Oade PMD660. I found that my tracking levels mainly between -30 and -12 had the preamps gained at about 9 o'clock. It's nice that I can now use my less sensitive Gefells and B&K's without worrying about gain!
This is using the L setting and no Limiter. The AKG's come in at about 20mV/Pa - the gefells and other high end mics are about 10-14mV/Pa.

I used 1bit/5.6 mode and was suprised momentarily by the splitting of my files if it goes over 1GB but then remembered that is the way it works...

Headphone amp is pretty good too and provided ample signal for my cans.

Next I am going to try it with ext. preamps - trusty BG2.
Thanks to all who have helped me understand this new and exciting unit.