Measurements & Preview of Holo Audio Spring 3 DAC / wPreamp (Prototype)

[JohnYang1997]

the post from Topping referred to my post #76 and said "pure measurement errors"

Ok. If you insist on this.
Using music to measurements have large variance as you do not know when the music starts and when the music ends. And you do not apply window functions on your measurements.
Do frequency response at different set levels also measure linearity at different frequencies and see if there's difference. If there's this large of variance in DACs then the dac must be crap(but this is not the case). There has been countless times incompetent people finding differences in measurements without knowing how anything works.

 

[Le Concombre]

Ok. If you insist on this.
Using music to measurements have large variance as you do not know when the music starts and when the music ends. And you do not apply window functions on your measurements.
Do frequency response at different set levels also measure linearity at different frequencies and see if there's difference. If there's this large of variance in DACs then the dac must be crap(but this is not the case). There has been countless times incompetent people finding differences in measurements without knowing how anything works.

Using music to measurements have large variance as you do not know when the music starts and when the music ends.
Use a full track

And you do not apply window functions on your measurements.
I confess I don't see how this is relevant : I'm not measuring my room

Do frequency response at different set levels also measure linearity at different frequencies and see if there's difference.
Levels were matched as appears in all the points where the traces are common

If there's this large of variance in DACs then the dac must be crap(but this is not the case). There has been countless times incompetent people finding differences in measurements without knowing how anything works
Every time results don't verify your marketing credo? I offer you a terrific argument : get a Spring and compare to a Topping in conditions where certified independent third parties can certify honesty and show the world that there's actually no difference in reproduced music in a room when spending 3 times more

 

[Le Concombre]

I'd rather be surfing and it's time I go do so but I made another comp with solo piano 1536 vs 256, 256 is green

 

[Miska]

To each, what does the perspective formats sound like to you, that makes you prefer it.
DSD vs pcm

To me, PCM through R2R ladders sounds "flat", lacking body/weight/authority. Like drums for example sounding more like "flap-flap" instead of "thump-thump". And the overall scene sounding "congested" or "busy" with less detail separation.

Note this is not about the format itself, but about the overall conversion process. On SDM DACs performance largely depends on the particular modulator implementation and conversion process. But the above "flavor" behavior doesn't apply to SDM DACs, which have different kind of differences.

 

[pkane]

Got my Spring 3 L2. I invite you to perform a simple experiment with at least 2 or your DACs you claim are sounding identical : feed them the same signal, whatever, provided it's decently loud and lively (some chamber music peaking around 86 dB at LP in my exemple), record in your room the cumulative spectrograms yield by a whole track and show us the results. For whatever reason, call it distorsion if you wish, all traces are different in my experiment. Oddly, the BB 1795 TEAC 501 fed DSD 128 and the R2R S3L2 fed PCM 1536 are close while S3L2 fed DSD 128 and 256 yield quite different spectrograms, most notably they are the highest traces @140 Hz

Cucumber, I've done this type of comparison on hundreds of DACs, and have yet to find the huge differences that you posted. As John said, either something is misconfigured or broken, or your measurement technique is faulty. Maybe you can describe in more detail how you captured these?

 

[Music1969]

It is also one of the discrete DSD DACs. Of course there are number of others too (T+A, Playback Designs, Nagra, Denafrips, EMM Labs, Meitner, Esoteric, TEAC, Marantz, etc).

I'm curious about novel discrete bit-perfect DSD DAC designs.

I think you made Ted S. shed a tear, leaving him out of this list

 

[Le Concombre]

Cucumber, I've done this type of comparison on hundreds of DACs, and have yet to find the huge differences that you posted. As John said, either something is misconfigured or broken, or your measurement technique is faulty. Maybe you can describe in more detail how you captured these?

ok here is a new one, freshly made. Both exemples showed before were made minutes after the S3 was powered on. Maybe a 500 hours burn in is BS but maybe 20 minutes after powering on a DAC for the first time makes a difference. Anyway the S3 has now been fed continuously for several hours. Differences between 1536 and 256 are lesser but still exist. Output was lowered 6dB in HQP when playing PCM. sinc S was the filter in both instances and parametric filters entered in HQP's pipeline as TXT were of course the same and yield the response in my avatar with the 39 Hz bump flattened, no correction above 500, after the slight low end rise response is flat til 2K and then natural roll off. The mike is fixed behind and slightly above the LP. Settings are attached. Track was all of 7:50 Liszt/Wagner from Barenboim's 24/96 "On my new piano" so my body did not move a lot during measurements and temperature humidity and pressure probably did not change much either over 15 minutes. 256 is blue trace

 

[pkane]

ok here is a new one, freshly made. Both exemples showed before were made minutes after the S3 was powered on. Maybe a 500 hours burn in is BS but maybe 20 minutes after powering on a DAC for the first time makes a difference. Anyway the S3 has now been fed continuously for several hours. Differences between 1536 and 256 are lesser but still exist. Output was lowered 6dB in HQP when playing PCM. sinc S was the filter in both instances and parametric filters entered in HQP's pipeline as TXT were of course the same and yield the response in my avatar with the 39 Hz bump flattened, no correction above 500, after the slight low end rise response is flat til 2K and then natural roll off. The mike is fixed behind and slightly above the LP. Settings are attached. Track was all of 7:50 Liszt/Wagner from Barenboim's 24/96 "On my new piano" so my body did not move a lot during measurements and temperature humidity and pressure probably did not change much either over 15 minutes. 256 is blue trace

I see. So you measured your DACs, room, speakers, amp, the microphone and the ADC all at once. How did you determine that the difference was not due to something other than the DSD rate?

 

[Le Concombre]

I see. So you measured your DACs, room, speakers, amp, the microphone and the ADC all at once. How did you determine that the difference was not due to something other than the DSD rate?

this is so blatantly fallacious argument ( the only variable remains the feed and thus the path inside the DAC) that I'm tempted to do a new comp. Maybe the matching low end in the previous one was due to long sustained chords on the left of the piano in the chosen track and thus maybe I can prove big differences on the whole spectrum : I'll be back

 

[sq225917]

You measure the dac output, not the room.

Facepalm

 

[pkane]

this is so blatantly fallacious argument ( the only variable remains the feed and thus the path inside the DAC) that I'm tempted to do a new comp. Maybe the matching low end in the previous one was due to long sustained chords on the left of the piano in the chosen track and thus maybe I can prove big differences on the whole spectrum : I'll be back

Really? Are you sure all these other variables don't matter?

 

[Le Concombre]

for the second time this new comp doesn't show low end discrepancies, suggesting that first comp might have shown early burn in effect, but yet global differences between 1536 and DSD256 feed

 

[Le Concombre]

Really? Are you sure all these other variables don't matter?

how ? please don't be shy, I have a PHD and been a research scientist, I think I could understand your explanation, how crazy it would have to be, of how changing the DAC feed, thus inside path, influences room, speakers, amp etc

 

[pkane]

how ? please don't be shy, I have a PHD and been a research scientist, I think I could understand your explanation, how crazy it would have to be, of how changing the DAC feed, thus inside path, influences room, speakers, amp etc

If you really are a research scientist, you should know better as to how control for variables in an experiment. The explanation is simple: you need to measure the actual effect you are trying to demonstrate in order to demonstrate it. As directly as possible, excluding as many variables as possible and quantifying and eliminating the ones that remain. If you involve multiple unknown and uncontrolled variables in the experiment, your result is at best open for interpretation, and more likely, simply faulty. This is not solved by waving hands and proclaiming that they don't matter. You already posted a number of variable results from exactly the same set up (claiming yet another variable -- break-in or warm-up as the explanation, again without any evidence).

 

[Le Concombre]

If you really are a research scientist, you should know better as to how control for variables in an experiment. The explanation is simple: you need to measure the actual effect you are trying to demonstrate in order to demonstrate it. As directly as possible, excluding as many variables as possible and quantifying and eliminating the ones that remain. If you involve multiple unknown and uncontrolled variables in the experiment, your result is at best open for interpretation, and more likely, simply faulty. This is not solved by waving hands and proclaiming that they don't matter. You already posted a number of variable results from exactly the same set up (claiming yet another variable -- break-in or warm-up as the explanation, again without any evidence).

I discussed the early burn-in hypothesis that had not been listed as possible cause of differences to rule it out at least for my 2 latest comps, that still show differences between PCM and SDM feed. I have a background in biophysics and biochemistry and room + active speakers + mike is just the equivalent of substrate with only one variable evaluated : the DAC feed thus inside path. This is kind of crazy that you doubt that 2 things different measure different. You're a fanatic believer in a credence ; sorry I proved you wrong

 

[VMAT4]

Does anyone make a DAC/Preamp with streaming capability and a headphone jack/amp? Would that be a big leap from this prototype?

 

[Kenmoreira85]

Does anyone make a DAC/Preamp with streaming capability and a headphone jack/amp? Would that be a big leap from this prototype?

https://apos.audio/products/matrix-mini-i-pro3-mini-i-3-wireless-bluetooth-dac-amp

 

[VMAT4]

https://apos.audio/products/matrix-mini-i-pro3-mini-i-3-wireless-bluetooth-dac-amp

Thanks! I'm waiting for Roon to work with Primephonic . I'm also waiting for Spotify to go lossless or Qobuz to become more reliable. Don't think I'll hunger strike for these things though. Change takes time.

 

[sq225917]

From cold the clock in a dac will settle in, maybe 30 minutes, but it won't have this affect. Maybe it's just as simple as the extra signal processing required for the higher data rate causes different thermal conditions in the dac? Ever run a gaming pc?

 

[Le Concombre]

From cold the clock in a dac will settle in, maybe 30 minutes, but it won't have this affect. Maybe it's just as simple as the extra signal processing required for the higher data rate causes different thermal conditions in the dac? Ever run a gaming pc?

processing was done outside the DAC on a Mini M1 running HQP. I'm OK with discarding the comps made minutes after first power on if warming-up/early burn in is considered as a potential extra variable. But last 2 comps were made in a row with the DAC fed continuously for hours so I think warming-up/early burn in may be totally ruled out as an extra variable