Is there any logical reason to use a tube as the oscillator in a DAC?

[SiliconWizard]

Yes $4000 is a ridiculous amount of money for even a very good and well-designed DAC, but I can certainly see much better use of money for the design nonetheless.
The final price has to be put into perspective in terms of margin - for high-end/luxury items, a 80%-90% margin is not unusual. We're very far from commodity gear with maybe a 20% margin if planets are aligned.

As I said, I'm pretty convinced this design does actually significantly worse than a simple CMOS TCXO - not even speaking of an OCXO which would still make it relatively cheap.

[vk4ffab]

Its irrelevant when the jitter is less than 500ns because most normal people cannot hear it.

I think you mean 500ps. Even the tone deaf would hear 500ns.

No i meant ns. That is about the limit where most average people can no longer hear jitter on digital audio and when you get down to 250ns even trained sound engineers can no longer hear jitter. There is plenty of research that has been done in this field. This is just one often cited paper, but there are others and none of them claim the human ear can hear jitter on digital audio in the ps range. So, as long as the dac/adc is using a clock with jitter in the low ps range, there will be no detectable jitter in the recovered audio even with a 24 bit dac sub optimally clocked.

Detection threshold for distortions due to jitter on digital audio ASHIHARA et al 2005



There are also consumer grade dacs on the market for a lot less than the one in this thread that have perfect jitter responses, MOJO comes to mind here, and no one is ever complaining about jitter with these sorts of things, they complain about a lot of other more important reasons for audio degradation. Certainly not their clock needs a tube to improve jitter Claiming their low jitter tube clock somehow has improved audio from a dac is nothing more that pseudoscience audio foolery and not based in reality because their 5ps clock will sound identical to a 100ps clock as both will have jitter distortion we just cannot hear.

[ataradov]

Capacitors are dancing all over the place. Must improve the sound too.

Also, their claim that it uses "Military grade" components is wrong. That OTK stamp indicates it is a regular commercial part.

[TimFox]

From a quick look at the PCB near the crystal, it looks like a normal Colpitts crystal oscillator circuit.
The Meacham Bridge circuit would be better, if it mattered.

[bdunham7]

That is about the limit where most average people can no longer hear jitter on digital audio and when you get down to 250ns even trained sound engineers can no longer hear jitter. There is plenty of research that has been done in this field. This is just one often cited paper, but there are others and none of them claim the human ear can hear jitter on digital audio in the ps range.

It depends a lot whether it is random or correlated jitter, as well as the program material and other circumstances.  Random jitter can be equated to noise floor.  The level below which audibility is possible is probably debatable, but the lowest number I recall seeing from a reputable source is 5ns, and I think that was a theoretical argument vs a demonstration.   500ns is a bit high, I think 500ns of jitter correlated to 1kHz might be readily audible with the right program material. 

As has been pointed out, it isn't unreasonable to hold recording systems to a bit better standard, but single-digit or fractional picoseconds is probably a bit much.  You can do the math for a 24-bit or 32-bit system, but no audio system has anywhere near the actual 144dB (or 196db!) dynamic range that these imply.

[TimFox]

144 dB of dynamic range would be excessive, since the threshold of pain is around 130 dB(SPL), the threshold of hearing is 0 dB(SPL) in an anechoic chamber, but you are very lucky if the background ambient sound level in your house is < 20 dB(SPL).

[boB]

Dumb idea although the audio philes might like it

[vk4ffab]

That is about the limit where most average people can no longer hear jitter on digital audio and when you get down to 250ns even trained sound engineers can no longer hear jitter. There is plenty of research that has been done in this field. This is just one often cited paper, but there are others and none of them claim the human ear can hear jitter on digital audio in the ps range.

I think 500ns of jitter correlated to 1kHz might be readily audible with the right program material.

If i remember rightly, Dolby did a bunch of tests back in the day and found jitter became audible in the low ns range on a single tone sine wave. Now other than CW operators doing morse code, I do not know anyone who listens to a single tone sine wave, so in a practical sense, a single tone is irrelevant because its not how anyone listens to or appreciates audio.
 But even in that worst case scenario, a clock in ps range is not going to add much, if anything to audible distortion because things like quantized noise, non linear distortions and even the noise of individual components have a greater impact on audible distortion on a typical audio stream. So I do not have a problem with Ashihara and others findings that trained sound engineers using their own studios could not hear jitter much below 500ns.

So taking this back to the device this thread is about, raving that 5ps is great and makes the audio better because of tubes is nothing more than marketing wank, because its effect is so small that the jitter could be orders of magnitude higher and still we wont be able to hear it.

[radiolistener]

No i meant ns. That is about the limit where most average people can no longer hear jitter on digital audio and when you get down to 250ns even trained sound engineers can no longer hear jitter.

it depends on sound pattern which was used for testing and conditions (volume level, etc). Maybe for usual music you will not be able to recognize difference with clock jittter less than 250 ns, but for special test cases it still can be detected. You can combine some signals in the source in that way so distortions caused due to clock jitter can be easily recognized.

Human ear has pretty large dynamic range. Yes, it is not linear, but for perfect reproduce for all frequencies, your DAC needs for about 140 dB dynamic range. There is no DAC with so high dynamic range that can cover all cases for human ear, so... it's better to use 0.1 ps jitter if you want to build good sound DAC, in order to not loss DAC dynamic range.

If i remember rightly, Dolby did a bunch of tests back in the day and found jitter became audible in the low ns range on a single tone sine wave. Now other than CW operators doing morse code, I do not know anyone who listens to a single tone sine wave, so in a practical sense, a single tone is irrelevant because its not how anyone listens to or appreciates audio.

Good quality sound DAC has many use cases, not only listening sound. For example it can be used to produce test signals for measurements which needs to be clean enough.

And maybe someone like to listening clean sines...  who knows

[radiolistener]

144 dB of dynamic range would be excessive, since the threshold of pain is around 130 dB(SPL), the threshold of hearing is 0 dB(SPL) in an anechoic chamber, but you are very lucky if the background ambient sound level in your house is < 20 dB(SPL).

For example rifle gun shot has about 160 dB SPL, helicopter has noise about 120-140 dB SPL.
Some orchestral musical trumpets instruments produce sound level about 150-160 dB SPL.
Such sound level is not comfortable for human ear, but don't cause pain. But can be unsafe for long exposure.

By the way, with usual atmospheric pressure max possible sound level is about 192 dB SPL, so if you want to cover painful sound, there is no DAC that can cover 192 dB dynamic range 

[Gyro]

I suspect that, even if the DAC puts out 140dB+ dynamic range, the components down stream are going to have severe problems in maintaining it!

[Kleinstein]

The effect of jitter in the DAC timing is only relevant when there is a signal, especially a signal with high amplitude and high frequency. The ear may be quite good in the dynamic range with 1 tone at a time, but not when is comes to detecting a little noise in the presents of a very loud one. The dynamic range is pretty limited and it is easy to mask noise noise.  So the test showing some 0.5-1 µs as the level where jitter get hearable sounds totally plausible.

[radiolistener]

I suspect that, even if the DAC puts out 140dB+ dynamic range, the components down stream are going to have severe problems in maintaining it!

yes, there is no amplifier that can handle so wide dynamic range.

[radiolistener]

The ear may be quite good in the dynamic range with 1 tone at a time

This is not true, human ear is more complicated and can hear many tones simultaneously and even more it can distinguish even very weak signals on background of very loud one. But it has non flat frequency response and non flat sensitivity level, it depends on many factors. Include individual properties of ear, age factor and ability to hear different aspects of sound which needs to be trained.

So the test showing some 0.5-1 µs as the level where jitter get hearable sounds totally plausible.

with 0.5-1 uS clock jitter you will get bullshit sound with high distortions. Something like cheap Chinese kid toy.

I can say it’s very easy to get used to high-quality sound, but then it’s hard to listen to cheap sound, because you start to hear all the defects that you didn’t notice until you heard high-quality sound...

Just remember how we listened very distorted and noisy sound from tape in the past and it was "pretty good", but now, when everyone got used to better quality sound, the tape sounds like bullshit... 

[coromonadalix]

yeah  the higher resolution / precision etc ...  you will hear them more, im lucky to have very good hearing at my age

same for tv's  or else  loll    you see more defects ... impurities ...  artefacts ... i remember the complaints When Sony Trinitron came into market  loll  peoples saw too much loll


you get high end dacs,  but now the amp doesn't follow and the speakers too ...

i played with what they called high end stuff at the time,  like DAT  or high end cd players with external DACS,  or the DAC chip ... ones you could buy and change them on the main pcb's  etc ...

You could hear the difference in the cds recording mode : digital digital etc ... digital analog ....    on 20,000 $/ea  speakers   loll

Played with Carver,  high end amplifiers,  wow,   i had formation to repair them loooooong time ago loll

You have Mosfets based amplifiers with a bandwith and frequency response so incredibly straight from 10hz to 50 Khz ....
 
you alway end up missing something, unless you buy an huge kit,  once and for all   loll


For the tubes,   i always hear  a tube is "warmer" in sound vs electronic ones (transistors / mosfets  etc ... a class a-b class)    even heard those who are called G - H class, or the D i think,  it like an smps amplifier running at very high frequencies 1Mhz or more ...

Tubes amp could take more "beating" than others  ...  some of my old friends whant tubes amps, tube guitar amps  loll   end of an era  loll   

oh and yes  like the previous thread,  going down into normal "cheap stuff" is like .......... crying and banging you head on a wall 

[vk4ffab]

No i meant ns. That is about the limit where most average people can no longer hear jitter on digital audio and when you get down to 250ns even trained sound engineers can no longer hear jitter.

it depends on sound pattern which was used for testing and conditions (volume level, etc). Maybe for usual music you will not be able to recognize difference with clock jittter less than 250 ns, but for special test cases it still can be detected. You can combine some signals in the source in that way so distortions caused due to clock jitter can be easily recognized.

Human ear has pretty large dynamic range. Yes, it is not linear, but for perfect reproduce for all frequencies, your DAC needs for about 140 dB dynamic range. There is no DAC with so high dynamic range that can cover all cases for human ear, so... it's better to use 0.1 ps jitter if you want to build good sound DAC, in order to not loss DAC dynamic range.

If i remember rightly, Dolby did a bunch of tests back in the day and found jitter became audible in the low ns range on a single tone sine wave. Now other than CW operators doing morse code, I do not know anyone who listens to a single tone sine wave, so in a practical sense, a single tone is irrelevant because its not how anyone listens to or appreciates audio.

Good quality sound DAC has many use cases, not only listening sound. For example it can be used to produce test signals for measurements which needs to be clean enough.

And maybe someone like to listening clean sines...  who knows

The thing to remember is that the device we are talking about in this thread is a high fidelity audio device, its not a special test case, its not theoretical, its entirely practical and it revolves around the subjective reality of listening to audio, most commonly music. So in that context and that context only, 5ps clock jitter does not improve audio because humans cannot hear random jitter much below 500ns. There are certainly cases where that much jitter will be detrimental but those cases are not audio.

Its also not that hard to design a clock with 0.2ps jitter, which would be required for accurate reproduction of audio with a 24bit dac. Diodes Inc make them as a part off the shelf you can buy from mouser. Using that they could claim look at us, no jitter on our audio, but rather, they use a tube, have detectable jitter and claim their tube thing makes the audio better. Because its a sales gimmick for tube people, nothing more. The tube makes the jitter we cant hear sound warmer after all

To quote Jello Biafra, those dumb punk kids will buy anything.

[jonpaul]

rebonjour:

Misc notes:

Jitter audibility: See already referenced AES and AP papers by Steve HARRIS and Julian DUNN.

Reason to use a tube:

Nostalgia, appearance,

Jon