Consensus about upsampling to 512 DSD

[mansr]

3 hours ago, GUTB said:

Any DS-type DAC would benifit from upsampling if for no other reason than to skip the SRC.

That depends. The TI/BB DSD1793 performs better with 192 kHz input and SRC enabled than with 384 kHz and SRC bypassed.

[mansr]

4 minutes ago, christopher3393 said:

Hi Mans. Can you say what you mean by this?

With inputs up to 192 kHz, the DSD1793 does an 8x upsampling, the 1536 kHz output of which is sent to the sigma-delta stage. In filter bypass mode, the sigma-delta modulator is fed directly with the (max) 384 kHz input.

 

This is the output of the DAC chip before analogue filtering with 384 kHz input in filter bypass mode. The input signal is a simple ramp consisting of the four steps easily visible in the scope image.

 

This is the output from a "ramp" of two steps at 192 kHz. The 8x interpolation has expanded this to 16 smaller steps. This makes the job of the analogue filter much easier.

[mansr]

6 minutes ago, Miska said:

Well it is not so clear, because the digital filter has quite poor stop-band attenuation and high pass-band ripple compared to better ones.

Some attenuation is better than none.

[mansr]

5 minutes ago, GUTB said:

This is an audiophile topic: non-audiophiles may not get much value from it.

Everybody out!

[mansr]

12 minutes ago, barrows said:

Let's take for example the PS Audio DS, and let's say we oversample everything to DSD 128 first.  Now if we do that, yes the DS will still oversample again up to its very high running rate, but the fact is that the filters necessary for going from DSD 128 up should not produce any audible artifacts, as all of their artifacts will be so far out of band to be entirely inaudible.  

It is also very possible that one can run a much better filter algorithm and modulator in software than what the DS can do in its fairly limited FPGA

If you do any processing of a DSD stream, the result has to be remodulated, so any limitations of the FPGA still apply. Given that DSD128 still has quite a bit of noise below 100 kHz, you'll probably get better results using the highest PCM rate of 352.8 kHz (384 kHz seems unsupported, weird).

[mansr]

1 hour ago, Allan F said:

Dunno why some people feel the need to use the term "audiophile" in a pejorative sense.

Probably the same reason some audiophiles use the term "hi-fi" in a pejorative sense.

[mansr]

1 minute ago, Ralf11 said:

distinguish audiophiles from audiophools

There's a third category: GUTB.

[mansr]

31 minutes ago, Ralf11 said:

he seems to be under-clocked

He's actually made a lot more sense the last few days. Unclear whether the kidney stone or the pain meds are responsible.

[mansr]

2 minutes ago, Le Concombre Masqué said:

Some might accuse me to jump on a "hot" topic but...

 

any experience with laptop cooling mats ? 

 

recommendation ? are they sonorous or noisy, interfere with SQ as being plugged on a USB port ?

I'm not familiar with that flavour of snake oil. Could you give a link to an example or two?

[mansr]

10 minutes ago, Le Concombre Masqué said:

snake oil ? https://www.amazon.fr/s/ref=nb_sb_ss_c_1_14?__mk_fr_FR=ÅMÅŽÕÑ&url=search-alias%3Daps&field-keywords=laptop+cooling+pad&sprefix=laptop+cooling%2Caps%2C324&crid=3RLQOYGANXUWN

Oh, not an audio thing. I always assume audio related accessories are snake oil unless there's reason to believe otherwise.

 

If your laptop is overheating, I suppose increasing the airflow around it might help. Many CPUs have thermal throttling, meaning they slow down if they run too hot. This could lead to stuttering when running a demanding upsampling. If that's not happening, the only thing one of these coolers will do soundwise is add fan noise.

[mansr]

1 hour ago, barrows said:

I find a very nice benefit oversampling to DSD 128 and DSD 256 using Audirvana Plus, on ESS 9018 and ESS 9038 based DACs, via Ethernet streaming to the Sonore Signature Rendu SE.  My current filter settings are:

 

Steep:     31.0

Length:   500000

Cut:         .92

AA:          100

Pre-Ring: .30

 

And I am using the higher order modulator.

Which modulator do you use? How would you characterise the differences between them?

[mansr]

Just now, barrows said:

i have not yet listened tested between them, too many things to test all the time!  I am using "B", 8th order.  The SoX DSM is your work yes?  Which one would you recommend from a technical perspective?

Yes, I wrote the modulator code. I would generally recommend 7th or 8th order. The lower order ones are somewhat faster, so they can be useful for that reason. Ultimately, whatever sounds best to you on your DAC is what you should use.

[mansr]

All analogue electronics have a signal to noise ratio that rarely exceeds 120 dB. A few DAC chips claim as much as 132 dB, but finished products are generally in the 110-115 dB range. The Benchmark DAC3 is an outlier with a specified SNR of 126 dB. This means 24-bit PCM covers the full dynamic range of any DAC or preamp with bits to spare. A digital volume control with proper dither does not lose any dynamic range compared to an analogue attenuator. Obviously, any attenuation, analogue or digital, of a signal in the presence of noise reduces the effective dynamic range. If music is attenuated by a large amount, some of the softer details will be lost. This is inevitable.

 

Also consider the listening environment. Music played at a high volume has peaks around 100 dB SPL. A very quiet room has an ambient noise level of 20 dB SPL. There is thus, in practice, at best 80 dB of useful dynamic range, a bit more with headphones.

[mansr]

20 minutes ago, semente said:

Making use of both measurements and listening assessments seems likely to provide better results than tasting alone.

Tasting is fine, but it can't hurt to check arsenic levels first.

[mansr]

3 minutes ago, GUTB said:

Care to name an example of a DAC and amp that won’t benifit from a pre?

Benchmark DAC3 + AHB2 amp. Probably not expensive enough for you though.

[mansr]

3 hours ago, Ralf11 said:

What I don't understand is why (or even whether) a $25,000 sounds better than a $5,000 DAC,

Compare the specs for the Benchmark DAC3 and the dCS Vivaldi. The latter is 10x the price and delivers worse (pretty average, in fact) performance. I haven't personally heard either, but John Siau seems like a no-nonsense kind of guy.

[mansr]

14 minutes ago, Ralf11 said:

Shoot out

 

meet you at the ok corral

I have no horse in this race. Yet.

[mansr]

4 minutes ago, Jud said:

I don’t know, but as I recall Archimago said some good things about the Audiventory results. I use Audiventory and like it.

Audiventory applies a very steep lowpass filter near 20 kHz (I don't recall the exact number). Those with a fear of filters should probably avoid it.

[mansr]

5 minutes ago, tailspn said:

I've been having a conversation with a CA participant about this specific topic. The question that's been raised is whether the sacd_extract he's using is slicing on the frame boundary. While the DSD bitstream is simply a stream of bits whose population (density) is proportional to the percent of modulation, and therefore level, it's actually packaged into frames for storage and handling capability. Slicing a DSD file into tracks must be done at a frame boundary, otherwise when rendered into tracks, the bits located between the arbitrary slice point and the previous or next frame boundary will be discarded. The problem there is the new rendered tracks ending and beginning frames, previous track end and next track beginning will be at different DC offset levels, and hence a click.

That must be a software issue. None of the common file formats have such a limitation.

 

5 minutes ago, tailspn said:

I slice DSD edited masters (very few anymore since Merging Album Publishing does all that from a DXD mtff format edited master) by manually choosing an appropriate track slice point in time on the Pyramix timeline, then backing off to the previous Frame Foot. That's a simple one keystroke command in Pyramix. The result after rendering is absolutely clickless track transitions when played as an album. However, any one track will click on  track start and stop when played independently.

Clicks can be minimised by cutting as close as possible to a zero crossing. Since the precise location depends on the lowpass filter used, it is, however, impossible for this to be exact, and some clicking may be still be present depending on the playback system.

[mansr]

7 hours ago, tailspn said:

Actually no, the structure of DSD bit and framing organization has nothing to do with which software is used to operate on it. That organization is in the original DSD Scarletbook, and later expanded upon with faster DSD bitrates.

 

Clicks CAN be minimized by selecting the point in time of zero crossing. I used to do that years ago at the cost of several hours per album. The problem is then rendering out the sliced tracks. That rendering process discards the bits within a incomplete frame where the slice was made. If you can suggest (prove) an application that allows arbitrary slicing at bit points within a frame, which then separates out the resulting tracks without discarding bits back to the frame foot, I'd be most interested 

 

I can only speak from my experience using Pyramix, which is the professional DAW of at least 80% plus of production studios world wide producing DSD content.

OK, I don't know the SACD requirements, and I don't have experience with Pyramix. I was talking about the file formats used for downloads. DSF is by far the most popular, and it supports any number of samples exactly. DSDIFF does indeed require a multiple of 8 samples (sorry, I misremembered). As you correctly point out, DSD data itself has no frame structure, so there is no inherent reason to restrict the cut points. SACD, being based on the DVD physical medium, may well require sector alignment. I know DVD does. What size "frames" does Pyramix use?

[mansr]

43 minutes ago, Summit said:

What I try to illustrate and explain is that it’s important to get a representative survey sample to assure that inferences and conclusions can reasonably extend from the sample to the population/audiophiles as a whole.

This is especially true if the responders are self-selected.

[mansr]

1 minute ago, tailspn said:

Yes, I believe a raw channel modulator output has no frame structure. But immediately thereafter it's packaged (formatted) into the DSDIFF format for storage, processing, and playing out. It's not a Pyramix invention or proprietary  mechanism, but inherent to the DSDIFF (DSD) format.

DSDIFF is just one of several formats. DSF, which is far more common for downloads, has no such limitation.

 

1 minute ago, tailspn said:

This DSDIFF 1.5 specification pdf may help to explain:

 

http://www.sonicstudio.com/pdf/dsd/DSDIFF_1.5_Spec.pdf

 

Detailed in it are DSD Frames made of data Chunks, which in turn are various support chunks including sectorized raw DSD data bits. To the best of my knowledge, frames/chunks are time based, so at different DSD bitrates, there will be a different number of bits contained in each frame/chunk.

 

The point remains however, whether dff, or dsf at any available bitrate, DSD data is organized into frames. Any DSD editor not separating a full album file into tracks (slicing) along those frame boundaries will corrupt that frame, and it in turn will be discarded. The effect then when playing will be a click at the track transition bridging two different DC offset levels in onee bit time.

DSDIFF packs 1-bit samples into 8-bit bytes. The resulting per-channel byte streams are then interleaved and written to disk. If the number of samples to write isn't a multiple of 8, the file has to be padded or the odd samples at the end discarded. DSF uses a different block structure, but more importantly, it records the exact number of samples in the file header. If the final block is incomplete, reading software will know this and can stop processing at the actual end. I know this because I have done it.

[mansr]

10 hours ago, jabbr said:

This one is curious. The spectrum suggests it has been converted from a CD quality source. The horizontal line around the 60% mark in the spectrogram could be the result clipping in the resampler or the sigma-delta modulator momentarily saturating. What do you know about the provenance of this file?

[mansr]

3 minutes ago, jabbr said:

The DSD was sourced from the PCM.

Why, oh why? Just release the damn PCM and be done with it.

[mansr]

1 minute ago, jabbr said:

Interesting ...

I just re-ripped the ISO from SACD and ran again ... a slight difference ...  this one is definitely Analogue Productions:

That's exactly the same image as before.