DSD Battle Royale!

[wisnon]

We mostly agree, mon ami. I think the distortion just shows the need for filtration clean up, just like when dialling in an FM signal ...its hashy when the dial is not centered dead on the correct frequency...same for analog satellite broadcasts...fuzzy until it snaps into focus when fully dialled in. In digital satellite broadcasts, its either there or not, ie total hash or perfect picture. This dialling in for me for me does not represent a conversion process, just stripping away the dross.

In my non-expert opinion (you are the network engineer, not me...LoL), I understand DSD to be way more bit efficient in that ALL the stream carries is the state CHANGES in each pulse (from the initial state value), while with PCM it carries an replica of the previous pulse state value PLUS the incremental change. That means that each pulse carries a lot of "redundant" information and so space is "wasted".

This is why I distinguish between resolution and information....

BTW, how are you? It been a long time since we chatted. You are in the NL, correct?

 

[Julf]

We mostly agree, mon ami.

But the devil is in the details

I think the distortion just shows the need for filtration clean up, just like when dialling in an FM signal ...its hashy when the dial is not centered dead on the correct frequency...same for analog satellite broadcasts...fuzzy until it snaps into focus when fully dialled in. In digital satellite broadcasts, its either there or not, ie total hash or perfect picture. This dialling in for me for me does not represent a conversion process, just stripping away the dross.

I see your point of view, but formally that filtering is an essential part of the conversion process, just as with PCM. Both with PWM/PDM and PCM, the result of the first stage of conversion is a waveform with "steps" - in the case of PCM, the step between two adjacent samples, and in the case of PWM/PDM the step of going "up" or "down" by one unit. To turn that into a continuous, analog waveform, you need a reconstruction filter that averages out the signal. The "dross" the reconstruction filter removes is the difference between the discrete, quantized values and the proper continuous analog waveform.

In my non-expert opinion (you are the network engineer, not me...LoL), I understand DSD to be way more bit efficient in that ALL the stream carries is the state CHANGES in each pulse (from the initial state value), while with PCM it carries an replica of the previous pulse state value PLUS the incremental change. That means that each pulse carries a lot of "redundant" information and so space is "wasted".

Using that analogy, when the waveform has to change a lot (as in a steep transient), DSD (PDM) needs a lot of "one unit of change" data points, while PCM communicates the whole change in one sample. What you have to look at is the actual data rate needed to achieve the specified level of bandwidth and resolution/signal-to-noise ratio.

BTW, how are you?

Fine - just busy. How about you?

It been a long time since we chatted. You are in the NL, correct?

Correct - Amsterdam, to be precise, but I am originally from Finland.

 

[wisnon]

I am well and glad the same is true for you.

I see what you are saying and understand a lot is semsantic, but I then have one obvious question. If what you say is true about the steep transients, how comes that DSD seems to have better impulse response which I think is related to ability to handle steep transients?

All this not about DSD vs PCM, as they BOTH can be made to sound wonderful and both have more capacity that is needed to create a near perfect sonic illusion. Its just about the BEST way to replay a piece of DSD recorded music, ie top notch conversion to very hiRez PCM in digital domain, SD manipulation with DSP to still native DSD (1 bit guys) all in digital domain then some kid of analog filtration, or pure analog filtration only (still cant see this as conversion) to analog directly.

 

[Julf]

I am well and glad the same is true for you.

Good to hear!

how comes that DSD seems to have better impulse response

I haven't seen any real evidence that DSD would have better impulse response - and a lot depends on how you define "better impulse response". Higher slew rate? Less ringing?

All this not about DSD vs PCM, as they BOTH can be made to sound wonderful and both have more capacity that is needed to create a near perfect sonic illusion.

I definitely agree.

Its just about the BEST way to replay a piece of DSD recorded music, ie top notch conversion to very hiRez PCM in digital domain, SD manipulation with DSP to still native DSD (1 bit guys) all in digital domain then some kid of analog filtration, or pure analog filtration only (still cant see this as conversion) to analog directly.

Theoretically it is of course best to minimize the processing/manipulation, so if you have a pure DSD recording, that has been kept at 1-bit DSD all through the recording and mastering process, then the obvious way to reproduce it would be a direct DSD-to-analog conversion (involving an averaging step, also referred to as "filtration" ). What muddles the waters is that pretty much any processing involves conversion to either a pure PCM format (DXD) or a hybrid 8-bit PCM format ("DSD-wide") and then back to 1-bit DSD, so does the additional processing in the DAC really matter?

 

[PhilipAC]

I agree totally with where this "conversation" has got to. And I'm not sure of the answer to the question at the end.

But I do say, long live Channel Classics' DSD recordings which involve no non-DSD elements, and I hope others follow their path.

 

[Julf]

But I do say, long live Channel Classics' DSD recordings which involve no non-DSD elements, and I hope others follow their path.

Just curious, but don't they do any EQ or processing at all? Or do you count "DSD-wide" and/or DXD as DSD?

 

[PhilipAC]

Well, my understanding is that they do no processing or EQ, but in the interests of full accuracy I will investigate further and report back!

I do know they are very purist. They are the one of the organisers of the forthcoming new DL site "nativeDSD.com".

 

[Julf]

Well, my understanding is that they do no processing or EQ, but in the interests of full accuracy I will investigate further and report back!

Thanks! Really curious to hear/see what their recording/mastering chain looks like.

I do know they are very purist. They are the one of the organisers of the forthcoming new DL site "nativeDSD.com".

Wish there was a "native PCM" site too (as in recordings that are true hi-res recordings, and not hi-res re-recordings of lower-res (analog or digital) media)).

 

[wisnon]

"What muddles the waters is that pretty much any processing involves conversion to either a pure PCM format (DXD) or a hybrid 8-bit PCM format ("DSD-wide") and then back to 1-bit DSD, so does the additional processing in the DAC really matter?"

Julf,

As far as I understand there is no such conversion in the Lampi DSD module, just active and passive SS filtration (digital and analog) followed by active tube filtration at the output stage.

"THE LAMPIZATOR DSD DAC HAS USB PORT BUILT IN, SOLID STATE DIGITAL FILTER, PASSIVE DISCRETE ANALOG FILTER AND ACTIVE DISCRETE TUBE FILTER. IT DOES NOT MANIPULATE THE DATA IN ANY SHAPE OR FORM, IT DOES NOT CONVERT IT, UPSAMPLE, RECLOCK OR DOWNSAMPLE. IT DOES NOT GO THROUGH PCM PROCESS EITHER. IT IS PUREST NATIVE DSD WE KNOW OF. WE PRACTICALLY ONLY GENTLY REMOVE THE CARRIER FREQUENCY FROM THE RAW DATA AS IT COMES FROM HARD DRIVE. NOTHING MORE NOTHING LESS."

 

[Julf]

As far as I understand there is no such conversion in the Lampi DSD module, just active and passive SS filtration (digital and analog) followed by active tube filtration at the output stage.

I agree - when I was writing about was processing during recording, mixing and mastering, and my point was "so does the additional processing in the DAC really matter?" - if there is already processing happening at the recording stage, does additional processing at the DAC (if done) make any difference. Thus, while one has to admire the single-mided purism of the Lampi, the question is if it is the optimal solution for most recordings?

 

[PhilipAC]

Julf

If you go to www.channelclassics.com>General Info>About recording equipment, it gives quite a bit of detail. I have asked them about EQ/processing and will report back when I get an answer.

Philip

 

[PhilipAC]

I agree - when I was writing about was processing during recording, mixing and mastering, and my point was "so does the additional processing in the DAC really matter?" - if there is already processing happening at the recording stage, does additional processing at the DAC (if done) make any difference. Thus, while one has to admire the single-mided purism of the Lampi, the question is if it is the optimal solution for most recordings?

I understand your point, but if one is playing a DSD file I can't see why the Lampi approach would ever be SUB-optimal. Surely, the less additional processing the better?

 

[Julf]

Surely, the less additional processing the better?

Nothing wrong with processing as long as the effects aren't audibly harmful. Analog/passive filters might not be the most optimal ones.

 

[Julf]

Philip,

If you go to www.channelclassics.com>General Info>About recording equipment, it gives quite a bit of detail.

It gives details about the chain until the ADC, but not beyond that. They could get away with their all-analog-until-ADC chain as long as they record everything live, without any multitracking, editing or processing - but that severely limits the kind of music they can do.

 

[wisnon]

I agree - when I was writing about was processing during recording, mixing and mastering, and my point was "so does the additional processing in the DAC really matter?" - if there is already processing happening at the recording stage, does additional processing at the DAC (if done) make any difference. Thus, while one has to admire the single-mided purism of the Lampi, the question is if it is the optimal solution for most recordings?

Well my DSD experience is limited to Mytek (twice), PBD 3 and 5 in a marathon listening session, my Chord Qute HD and the Lampi DSD-only Dac. The Lampi is subjectively best to my ears. I am waiting on an invitation to go back to chez PBD and do a direct comparison. I will also get to hear the CH Precision Dac in late January, but that is so far only a DSD to 768khz PCM converter board.

I predict that by next Summer there will be a bunch od DSD filter boxes/modules on the market.

YES, IMHO, the filtration approach is the best one currently for DSD, followed by the 1 bit converters. However, a quality DSD to PCM conversion can be fabulous too.

 

[wisnon]

Decimation is always harmful, though not necessarily audible.

 

[Julf]

Decimation is always harmful, though not necessarily audible.

How is it harmful if it is not audible?

I prefer to reserve the word decimation for outright downsampling - a DSD-to-PCM conversion is a resampling instead of a straightforward downsampling.

 

[Julf]

IMHO, the filtration approach is the best one currently for DSD, followed by the 1 bit converters.

What is the actual difference between the "filtration approach" and a pure-DSD 1-bit converter?

 

[wisnon]

I think Cirrus makes a 1 bit chip. Not sure what goes on inside that chip. PBD and EMM upsample everything to DSD128 and use DSP as far as I know, staying in 1 bit all the way. At least that is what I understand. If you know better, please bring it on. I welcome more knowledge.

 

[Julf]

I think Cirrus makes a 1 bit chip. Not sure what goes on inside that chip. PBD and EMM upsample everything to DSD128 and use DSP as far as I know, staying in 1 bit all the way. At least that is what I understand. If you know better, please bring it on. I welcome more knowledge.

No, I don't have much more information on the chips either - I can see the upsampling to DSD128 requires DSP, but not sure as to why they feel the need to do it (apart from being able to support DSD128).