Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?

[antcollinet]

True. If you look at the results of Amir's tests, I'd guess everything from top to bottom exceeds the limits of my (or anyone's) hearing. So, why bother to rank at all? If, in the final analysis, it's all limited by our physiology and biases why does anyone ever buy anything above a PROZOR 192kHz DAC ($13.99 on Amazon). It's spec's put it well beyond human limits. Silly statement? OK, then exactly where between the PROZOR and the $115,000 MSB SELECT DAC do audible differences end?

Just in case you are still reading. It’s easy - when the measurements say they do.

And that can happen at any price point. We have examples of transparent (or close to) measuring DACs at around $10 (apple dongle) and other examples of very high priced items that measure appallingly.

For me it is easy - shortlist from the blue/green sections of the Sinad chart - all of which are going to be fine audibly - and select from them based on criteria other than sound. Features, quality, after sales, price, aesthetics etc etc.

 

[DMill]

Maybe he was the engineer on the Modi? But I highly doubt it. The folks at Schitt have always been really great to work with and the Modi is a fine little DAC by my impressions. I’m sure a properly functioning Topping is also wonderful.

 

[AnalogSteph]

Audiometric testing from long ago (using headphones) found the threshold of hearing to be -8dBSPL at 3kHz.

Yes, for a sine signal. The detection threshold for narrowband signals like that is always going to be lower due to what is effectively FFT gain (though the bin spacing for human hearing is a bit wonky). Doesn't mean you need a wideband noise level quite that low - after all you can still spot signals down to -130ish dBFS in a N=16384 FFT of a properly-dithered 16-bit signal, too. Human hearing effectively seems to be around N=8.

170dB DAC dynamic range is not "insane". All the dynamic range improvement is on the low-side, the noise-side. There is nothing insane about lowering the noise-floor of an audio system. The more the better, I say. Do the math. A DAC with 40nVrms of broadband / unweighted self-noise (-146dBu) and a micamp / delivery system of +29dBu is 175dB dynamic range.

-146 dBu in 20 kHz bandwidth would be a voltage noise density of about 0.275 nV/√(Hz). That's absolute top-notch MC prepre terrain. It's the equivalent of a five freaking ohm resistor's thermal noise. You'll be looking at discrete input stages using super low rbb' transistors and actual power stages / power buffers (*) to drive the substantial wattage resistors in your feedback network if you want to implement a typical 6-14 dB amplifier (basically a full-grown headphone amplifier just for a line stage).

I'm not saying it's impossible but man is it ever going to be a headache, even with Horowitz / Hill and Douglas Self at hand. Especially if you have super low distortion targets as well, at which point I would expect you can upgrade your input stages to the bootstrapped (folded) cascode variety. And don't even ask about effective input protection...

The best balanced mic inputs you're going to find out there might have a EIN of -129 to -130 dBu for a 150 ohm source, 20 kHz unweighted. That's about -133 to -137 dBu from a dead short. Ones that can take full line level input like +18 dBu don't generally go quite that low. And newsflash, those are the figures at or near maximum gain, of course they won't do lowest noise and maximum input level at the same time. Analog stages capable of anywhere near that are most likely going to be found in the input stages of equipment with composite ADCs, like a Neumann MT 48.

About the most instantaneous dynamic range you can expect from a line stage in this day and age is somewhere in the low 150s dB. Some headphone amps are in the mid-140s, that's already really good for a shipping product. I guess you can expect about 3 dB more balanced. Still, not anywhere near 170. It's plenty for domestic playback purposes either way.

*) I was going to say Power Rangers as well but no, maybe not them...

 

[signalpath]

Yes, for a sine signal. The detection threshold for narrowband signals like that is always going to be lower due to what is effectively FFT gain (though the bin spacing for human hearing is a bit wonky). Doesn't mean you need a wideband noise level quite that low - after all you can still spot signals down to -130ish dBFS in a N=16384 FFT of a properly-dithered 16-bit signal, too. Human hearing effectively seems to be around N=8.

-146 dBu in 20 kHz bandwidth would be a voltage noise density of about 0.275 nV/√(Hz). That's absolute top-notch MC prepre terrain. It's the equivalent of a five freaking ohm resistor's thermal noise. You'll be looking at discrete input stages using super low rbb' transistors and actual power stages / power buffers (*) to drive the substantial wattage resistors in your feedback network if you want to implement a typical 6-14 dB amplifier (basically a full-grown headphone amplifier just for a line stage).

I'm not saying it's impossible but man is it ever going to be a headache, even with Horowitz / Hill and Douglas Self at hand. Especially if you have super low distortion targets as well, at which point I would expect you can upgrade your input stages to the bootstrapped (folded) cascode variety. And don't even ask about effective input protection...

The best balanced mic inputs you're going to find out there might have a EIN of -129 to -130 dBu for a 150 ohm source, 20 kHz unweighted. That's about -133 to -137 dBu from a dead short. Ones that can take full line level input like +18 dBu don't generally go quite that low. And newsflash, those are the figures at or near maximum gain, of course they won't do lowest noise and maximum input level at the same time. Analog stages capable of anywhere near that are most likely going to be found in the input stages of equipment with composite ADCs, like a Neumann MT 48.

About the most instantaneous dynamic range you can expect from a line stage in this day and age is somewhere in the low 150s dB. Some headphone amps are in the mid-140s, that's already really good for a shipping product. I guess you can expect about 3 dB more balanced. Still, not anywhere near 170. It's plenty for domestic playback purposes either way.

*) I was going to say Power Rangers as well but no, maybe not them...

You're thinking single-path architecture. The quiescent noise floor of today's best single-path DACs is around -116dBu BB/UW (assuming FS of around +18dBu). That's a dynamic range of around 134dB BB/UW, and that's pounding against the limit of low resistance design (DAC / IV / Buffer).

To achieve -146dBu quiescent DAC noise BB/UW requires a multi-path approach. It's the only way. Neither Horowitz, Hill, nor Self have ever written about it (to my knowledge, and I've devoured most of TAOE). An equivalent DAC stage resistance of 4.6 ohms was achieved in the low-path of a multiple-path topology. That's the lowest we've achieved due to hard tradeoffs below around 4.5 ohms. Unfortunately, no available test equipment can measure BB noise this low. Not even close. It's calculated using the Johnson-Nyquist noise equation. There's a discussion of this in AES paper #21106, if anyone's interested.

I'm familiar with microphone preamps. I think we've shipped more ultra-high-performance micamps than anyone (>50,000 channels). In fact, we're now working with NASA on the LISA project, with literal nitrogen-bath circuits. Lowest noise circuits I've ever seen. Actually applying audio design to laser testing. The lasers need to cover over 1M miles. We're having a lot of fun

We have an audacious goal to re-imagine every link in the audio chain (mic, pre, ADC, DAC, power amp) in multi-path architecture, both PDM and PCM circuits, with each link performing up to 170dB dynamic range. We know the theory. But making it all work together is a life project. Wish us luck.

 

[Freedom666]

Without reading all the thread.

I found often in CD Players (and one more expensive DAC like Benchmark for 1000 Euros) that there were no Cs in between the legs of the LM78xx regulator.

You easily can hear the difference just adding whats described in the data sheet for these regulators (kerko Cs in between the legs).

Now it is questionable if you can easily measure the difference and

could the difference really be found out in a blind listening test?

 

[Blumlein 88]

Without reading all the thread.

I found often in CD Players (and one more expensive DAC like Benchmark for 1000 Euros) that there were no Cs in between the legs of the LM78xx regulator.

You easily can hear the difference just adding whats described in the data sheet for these regulators (kerko Cs in between the legs).

Now it is questionable if you can easily measure the difference and

could the difference really be found out in a blind listening test?

If you can easily hear the difference, it would easily be found out in a blind test. If it is easily heard you can measure the difference too.

 

[solderdude]

We have an audacious goal to re-imagine every link in the audio chain (mic, pre, ADC, DAC, power amp) in multi-path architecture, both PDM and PCM circuits, with each link performing up to 170dB dynamic range. We know the theory. But making it all work together is a life project. Wish us luck.

You could hire Rob Watts.
He claims he can hear differences -350dB down.

 

[ads_cft222]

Let’s say you have a pianist and a singer . You record it under the best possible microphone (s) 2m away . Will the recoding be able to be indistinguishable from reality if reproduced by a system ? If yes, what will be the characteristics of this system ? If the bottleneck is the recording itself , what are the characteristics of the system that will reproduce this recording with 100% accuracy? I mean technical accuracy not what everyone hears which is subjective

 

[Purité Audio]

A microphone ( any microphone set up) is not the same as your ears, and musical instruments in a room will propagate sound in a completely different way to a pair of loudspeakers.
All an audio reproduction system can do is to reproduce the recording as accurately as possible.
Keith

 

[Freedom666]

Let’s say you have a pianist and a singer . You record it under the best possible microphone (s) 2m away . Will the recoding be able to be indistinguishable from reality if reproduced by a system ? If yes, what will be the characteristics of this system ? If the bottleneck is the recording itself , what are the characteristics of the system that will reproduce this recording with 100% accuracy? I mean technical accuracy not what everyone hears which is subjective

You can check your loudspeakers:

Buy some instruments like triangle, metal xylophone, plastic flute (metal and plastics are independent of changes in humidity, that's different with wood and the like), record them well in the near field.

Play the record on your boxes and play the same recorded instruments simultaneously.

You can hear how well your boxes are and you know as a developer of equipment when your job is finished.

If source and reproduction are indistinguishable all is perfect.

I realized this with fullrange loudspeakers and use of dsp.

Like this one

How to make a loudspeaker sandwich cone

A good book in german on loudspeakers is Götz Schwamkrug "Lautsprecher - Dichtung und Wahrheit", its from 1986. PDF download here http://www.einklang-audio.com/buch/dichtung_und_wahrheit.pdf Concerning sandwich loudspeakers here some information, page 92 in the book, showing impulse responses...

www.diyaudio.com

 

[MaxwellsEq]

Yes, 130dB is roughly today's limit of single-path DACs (all DACs today). We're experimenting with multi-path. The low-path has a 40nVrms quiescent BB/UW noise floor, while the high path has +29dBu headroom. That's a dynamic range and linearity of 175dB. It's somewhat analogous to HDR photography, perhaps we could call it "HDR-A". Multi-path architecture can be applied to every link in the audio signal path, from microphones to power amplifiers, improving today's best systemic dynamic range from 120dB to 160dB. We're less interested in commercial viability (applicability?) and more interested in proving to ourselves that it can be done. It's probably more practical is areas of seismology, medical imaging, and test equipment. You can read more about multi-path architecture in various patents,

https://patentimages.storage.googleapis.com/e6/ef/7b/aa66c2a13634c3/US9871530.pdf

https://patentimages.storage.googleapis.com/c9/97/6a/003abcd65746e2/US9590648.pdf

https://patentimages.storage.googleapis.com/3b/c2/17/a72b56f9159d6d/US10256782.pdf

Thanks for this - an interesting approach!

 

[antcollinet]

Let’s say you have a pianist and a singer . You record it under the best possible microphone (s) 2m away . Will the recoding be able to be indistinguishable from reality if reproduced by a system ? If yes, what will be the characteristics of this system ? If the bottleneck is the recording itself , what are the characteristics of the system that will reproduce this recording with 100% accuracy? I mean technical accuracy not what everyone hears which is subjective

No, as pointed out by @Purité Audio the recording cannot be indistinguishable from reality.

Regarding a “perfect” system, again there is no such thing - mainly limited by speakers and room.

However, most electronics (streamers, DACs) are capable of audibly perfect reproduction. Most amps are pretty damn close (to the point of being indistinguishable from each other if level matched and operated within power limits)

Speakers are still the bottleneck, together with the way they interact with the room. These are the areas that should be focussed on for improving sound quality.

 

[ads_cft222]

.

 

[Miguelón]

Aha, the eternal battle audiophile vs professional million dollar question!

I’m musician (pianist) not an audiophile nor a professional in electrical sound.

For me there is a big difference between DACs, not sure one better than others apart from distortion and linearity, but it exists.

I play the piano since 7 years old, and know its sound better than the voice of my mother. Steinway, Bösendorfer, Fazioli, Yamaha, Kawai… I played them and I listened for all them on different concert halls.

I’m not sure that measures tell all of the history of a circuit, but can tell that for example in image testings two PC monitors can measure exactly the same in all parameters but a person can find a difference between them side by side.

Probably this has more to do with filters in the audio world than electronics itself: I have a Focusrite Scarlett and like how it sounds with my digital Kawai piano, but can be very fatiguing to casual listening. I use an Ifi Zen Dac Signature V2 for relaxing, don’t know how they filter any of both but I’m absolutely sure that they sound very different (both with balanced XLR cables to Genelec 8030).

What I’m sure also is that paying 2.000€ or something like that for audiophile DAC is absurd because after all is a matter of taste if sharp or slow phase filters goes better or worse. On an ideal world, as DACs are little computers will be nice to freely apply whatever filter you want. Oh, this already exists, we call it DSP !

(I’m not sure but maybe the Scarlett has no filter at all, highs and treble are super sharp to a fatiguing limit)

 

[SIY]

For me there is a big difference between DACs, not sure one better than others apart from distortion and linearity, but it exists.

Yes, some are larger, some have different colors, some weigh more...

 

[xray]

Yes, some are larger, some have different colors, some weigh more...

And newer models tend to sound better because they are newer and shinier.

 

[Blumlein 88]

Aha, the eternal battle audiophile vs professional million dollar question!

I’m musician (pianist) not an audiophile nor a professional in electrical sound.

For me there is a big difference between DACs, not sure one better than others apart from distortion and linearity, but it exists.

I play the piano since 7 years old, and know its sound better than the voice of my mother. Steinway, Bösendorfer, Fazioli, Yamaha, Kawai… I played them and I listened for all them on different concert halls.

I’m not sure that measures tell all of the history of a circuit, but can tell that for example in image testings two PC monitors can measure exactly the same in all parameters but a person can find a difference between them side by side.

Probably this has more to do with filters in the audio world than electronics itself: I have a Focusrite Scarlett and like how it sounds with my digital Kawai piano, but can be very fatiguing to casual listening. I use an Ifi Zen Dac Signature V2 for relaxing, don’t know how they filter any of both but I’m absolutely sure that they sound very different (both with balanced XLR cables to Genelec 8030).

What I’m sure also is that paying 2.000€ or something like that for audiophile DAC is absurd because after all is a matter of taste if sharp or slow phase filters goes better or worse. On an ideal world, as DACs are little computers will be nice to freely apply whatever filter you want. Oh, this already exists, we call it DSP !

(I’m not sure but maybe the Scarlett has no filter at all, highs and treble are super sharp to a fatiguing limit)

Funny, Scarletts have a rep for being muffled and restrained in the highs. Frequency response doesn't support that or being sharp. I know for a fact they have filtering. I've measured it on both ADC and DAC side. I believe it is minimum phase. Recording interface ADCs often use this type filter for lower latency.

 

[Miguelón]

Funny, Scarletts have a rep for being muffled and restrained in the highs. Frequency response doesn't support that or being sharp. I know for a fact they have filtering. I've measured it on both ADC and DAC side. I believe it is minimum phase. Recording interface ADCs often use this type filter for lower latency.

Yes, you’re right: was the inverse, I find the Ifi Zen fatiguing and the Scarlett more neutral. I wrote it reversed

 

[Miguelón]

Funny, Scarletts have a rep for being muffled and restrained in the highs. Frequency response doesn't support that or being sharp. I know for a fact they have filtering. I've measured it on both ADC and DAC side. I believe it is minimum phase. Recording interface ADCs often use this type filter for lower latency.

I like so much the Focusrite’s headphone output with my Sennheiser HD 560, I’m thinking of upgrading to HD 600 but don’t know if the Scarlett will hand the 300 ohms headphones.

 

[antcollinet]

Aha, the eternal battle audiophile vs professional million dollar question!

I’m musician (pianist) not an audiophile nor a professional in electrical sound.

For me there is a big difference between DACs, not sure one better than others apart from distortion and linearity, but it exists.

I play the piano since 7 years old, and know its sound better than the voice of my mother. Steinway, Bösendorfer, Fazioli, Yamaha, Kawai… I played them and I listened for all them on different concert halls.

I’m not sure that measures tell all of the history of a circuit, but can tell that for example in image testings two PC monitors can measure exactly the same in all parameters but a person can find a difference between them side by side.

Probably this has more to do with filters in the audio world than electronics itself: I have a Focusrite Scarlett and like how it sounds with my digital Kawai piano, but can be very fatiguing to casual listening. I use an Ifi Zen Dac Signature V2 for relaxing, don’t know how they filter any of both but I’m absolutely sure that they sound very different (both with balanced XLR cables to Genelec 8030).

What I’m sure also is that paying 2.000€ or something like that for audiophile DAC is absurd because after all is a matter of taste if sharp or slow phase filters goes better or worse. On an ideal world, as DACs are little computers will be nice to freely apply whatever filter you want. Oh, this already exists, we call it DSP !

(I’m not sure but maybe the Scarlett has no filter at all, highs and treble are super sharp to a fatiguing limit)

Do proper controlled blind testing (of DACS), statistically relevant to confirm you can here a genuine difference in the sound-waves, rather then being fooled by your very human auditory system and its associated perception biases.

Then we might have something worthwhile to discuss.