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

[ELS122]

I came across a PrimaLuna DAC that uses a tube for the oscillator.
Why? is there any logical reason for this? Is jitter of any concern at all with modern oscillators? Is there any feedtrough from the oscillator to the audio?

[DimitriP]

Primaluna + DAC + Tube = a lot of dollars.
How would YOU do it?

Once upon a time in a place far away I was one told that signal from FM transmitters using tubes "sounds better" and "goes further".

Using a tube solves most problems, if not from an engineering perspective, it gives the sales/marketing people something to talk about.


Wouldn't  you feel "wierd" if you paid almost $4K for a DAC and all it used was solid state components like everyone else's for a lot less ?

I could have save me a lot of typing and said:
Primaluna; I rest my case.

But where would be the fun in that ?

[ataradov]

It is an audio-fool device, of course it will have tubes in random places. How else do you justify $4K price tag?

The construction looks like shit with random capacitors zip tied to the frame. But I guess this impresses certain audience.

Audio review sites that come up in a search are hilarious.  They all copy the same description text verbatim from the vendor site.

[coppice]

Don't you know audio equipment only sounds good if it produces a nice cheery glow. If the natural glow of a filament is insufficient you can always supplement it with a bunch of orange LEDs.

[TimFox]

Clock jitter in DAC systems for audio is more important than one might think, but low-jitter crystal oscillators can be made from either vacuum tubes or solid-state components.
https://www.analog.com/en/technical-articles/analyzing-audio-dac-jitter-sensitivity.html
The military ended up with a huge stock of "subminiature" tubes, as shown in your photo, which can be used in audio applications.
They can have advantages over larger tubes for microphonics, but transistors rarely exhibit microphonic problems.

[ataradov]

It is not clear what that tube even does here. It sits next to 10 cent crystal and some Nuvoton MCU.

There is a good chance that only the filament is connected to the power. It looks like there are traces from al the pins of the bulb, at least this is good.

Quote from https://www.hifinews.com/content/primaluna-evo-100-tube-dac-usb-tube-dac-tube-clock:

This master clock is still referenced to a standard 24.576MHz crystal – the small metal can just visible at the bottom of the inset picture – while a miniature 6S6B triode is used in the subsequent widebandwidth oscillator.

[coppice]

Clock jitter in DAC systems for audio is more important than one might think, but low-jitter crystal oscillators can be made from either vacuum tubes or solid-state components.
https://www.analog.com/en/technical-articles/analyzing-audio-dac-jitter-sensitivity.html
The military ended up with a huge stock of "subminiature" tubes, as shown in your photo, which can be used in audio applications.
They can have advantages over larger tubes for microphonics, but transistors rarely exhibit microphonic problems.

At 48ksps 1ps of jitter is of a similar order to dropping from 24 to 23 bits. So, low jitter is important, but 1ps is not hard to achieve these days. Just don't try using something the output of a PLL that's massively multiplying up from a low frequency clock source.

[coppice]

It is not clear what that tube even does here. It sits next to 10 cent crystal and some Nuvoton MCU.

There is a good chance that only the filament is connected to the power. It looks like there are traces from al the pins of the bulb, at least this is good.

Of course its clear. It glows. People like a cheery glow. That's why so much vacuum tube kit now has orange LEDs with their light projected onto the vacuum tubes to increase the effect.

[TimFox]

Clock jitter in DAC systems for audio is more important than one might think, but low-jitter crystal oscillators can be made from either vacuum tubes or solid-state components.
https://www.analog.com/en/technical-articles/analyzing-audio-dac-jitter-sensitivity.html
The military ended up with a huge stock of "subminiature" tubes, as shown in your photo, which can be used in audio applications.
They can have advantages over larger tubes for microphonics, but transistors rarely exhibit microphonic problems.

At 48ksps 1ps of jitter is of a similar order to dropping from 24 to 23 bits. So, low jitter is important, but 1ps is not hard to achieve these days. Just don't try using something the output of a PLL that's massively multiplying up from a low frequency clock source.

Naïvely, one would think that 1 ps of jitter with roughly 20 us/sample would be trivial, but the jitter from a bad oscillator can be measurable on an audio signal.
As discussed in the Analog Devices app note above, when you get jitter into the ns range, it is important.

[coppice]

Clock jitter in DAC systems for audio is more important than one might think, but low-jitter crystal oscillators can be made from either vacuum tubes or solid-state components.
https://www.analog.com/en/technical-articles/analyzing-audio-dac-jitter-sensitivity.html
The military ended up with a huge stock of "subminiature" tubes, as shown in your photo, which can be used in audio applications.
They can have advantages over larger tubes for microphonics, but transistors rarely exhibit microphonic problems.

At 48ksps 1ps of jitter is of a similar order to dropping from 24 to 23 bits. So, low jitter is important, but 1ps is not hard to achieve these days. Just don't try using something the output of a PLL that's massively multiplying up from a low frequency clock source.

Naïvely, one would think that 1 ps of jitter with roughly 20 us/sample would be trivial, but the jitter from a bad oscillator can be measurable on an audio signal.

Naive means being unable to roughly divide 20us by 2^24 in your head and get a number around 1ps. In 2023 maths standards suck very badly.

[TimFox]

The opposite of naïve is not having been born yesterday.
In my day at high school, we used to compete with each other on doing binary powers quickly in our heads.
However, applying random variations to signal degradation is a bit more complex than that.
I believe it was in the 1990s that serious attention started being paid to clock jitter in digital audio systems.
An article from 1993:  https://www.stereophile.com/reference/1093jitter/index.html

[ELS122]

Could noise impact jitter? some of those soviet pentodes have been known to have less than 300nV EIN.

Primaluna has the waveforms of the oscillator posted on their website: https://www.primaluna-usa.com/primaluna-evo-100-tube-dac
and the waveform looks.... bad. around 80V/us slope. which evem some audio op amps can manage
So the only logical reason for using a tube (other than marketting of course), that I can think of is noise.

[ataradov]

The tube is 6S6B. 6J52P is huge and has rigid pins.

It is there for marketing. All of this is achievable using solid state devices. But you have to use up old Soviet junk somehow, I guess.

[coppice]

The opposite of naïve is not having been born yesterday.
In my day at high school, we used to compete with each other on doing binary powers quickly in our heads.
However, applying random variations to signal degradation is a bit more complex than that.
I believe it was in the 1990s that serious attention started being paid to clock jitter in digital audio systems.
An article from 1993:  https://www.stereophile.com/reference/1093jitter/index.html

We were seriously applying ADC and DAC jitter analysis to radar signals in the 70s. Its not rocket science. Its the basic tradeoff of how far the signal can have changed within the jitter width. Perhaps you are thinking of the effects of dithering. That is more complex. and some aspects of it are far from obvious and intuitive.

[T3sl4co1l]

Noise is strictly worse, comparing devices of comparable ratings and bandwidth, more or less simply because the cathode is several times hotter than ambient (absolute temp).  Which incidentally also determines the exponential cutoff slope (tubes, like FETs, have an exponential subthreshold cutoff region; above threshold for tubes goes as Iout ~ Vin^(3/2); FETs, ~Vin^2).

A given device might have lower noise, or in a given frequency range (RF tubes can be quite good; TWTs are still a good choice for certain microwave applications, mostly aerospace; those might finally be getting displaced by GaN these days, I don't know), but pound for pound so to speak, transistors win; and cryogenically chilled transistors, even moreso.

For anything audio, rest assured the exclusive and sufficient answer is: marketing.  A case could kinda be made for competitive purposes in things like guitar amps, where That Tube Sound(TM) is easier to produce, in terms of design effort, than designing a DSP filter of equivalent performance.  But the DSP itself, ADC, DAC, and the chip amp following it, are far cheaper (and more efficient, and more mechanically robust, and more reliable..) in any kind of quantity (100s? 1000s?) where that design effort is amortized away.

Tim

[TimFox]

The opposite of naïve is not having been born yesterday.
In my day at high school, we used to compete with each other on doing binary powers quickly in our heads.
However, applying random variations to signal degradation is a bit more complex than that.
I believe it was in the 1990s that serious attention started being paid to clock jitter in digital audio systems.
An article from 1993:  https://www.stereophile.com/reference/1093jitter/index.html

We were seriously applying ADC and DAC jitter analysis to radar signals in the 70s. Its not rocket science. Its the basic tradeoff of how far the signal can have changed within the jitter width. Perhaps you are thinking of the effects of dithering. That is more complex. and some aspects of it are far from obvious and intuitive.

No, I am just stating that historically, during the transition from analog to digital audio systems, the effect of clock jitter became apparent to the audio fraternity (not only the demonized audiophiles).

[TimFox]

I did a quick simplistic calculation to see how 1 ps of jitter compares with noise.
Assume 74HCT logic, with these performance parameters:
V_IH - V_IL = 1.6 V guaranteed logic levels
V_OH - V_OL = 2.4 V, measuring the endpoints for the transition time specification.
t_t < 15 ns, for the rise/fall time between those endpoints.
For noise purposes, we estimate BW = 0.35/t_t = 23 MHz;
For thermal noise, assume 50 ohms supplying noise (conservative) of 0.9 nV/rt Hz, or 4.3 uV rms in 23 MHz BW.
The slew rate of the signal between binary levels is 2.4 V/15 ns = 160 V/us.
Therefore, 4.3 uV rms of noise voltage would effect a time difference of 27 fs rms (using that slew rate).

[jonpaul]

Tubes/Valves and jitter/clocks are two DIFFERENT topics.

Mixed up here


http://www.audiophilleo.com/zh_hk/docs/Dunn-AP-tn23.pdf

http://www.nanophon.com/audio/

https://benchmarkmedia.com/blogs/application_notes/12142221-jitter-and-its-effects

j

[TimFox]

Yes, but the original question asked if the use of a vacuum tube in the crystal oscillator would improve the clock jitter in a digital audio system.
Most of us believe that it would not improve the jitter, and that there are ways to generate a clock signal that have very bad jitter using either tubes or solid-state devices in a bad design.
Over my lifetime, it took a while before practical solid-state devices had better noise than the best tubes, but for normal purposes there is now a better solid-state device (JFET or BJT depending on the application) than the best small-signal tubes, especially since no one is now improving the performance of tubes for this purpose.

[vk4ffab]

An audio fool and his money are quickly separated. I wonder if they used Zero Oxygen copper on the board rather than that dirty filthy 0.0005% oxygen copper, because we all know that 8.30269534 × 10-30 moles can make your jitter do the chicken dance and the only solution is to listen to audio in the vacuum that surrounds the flat earth where the warm glow from the nearest star will induce a magnetic field in the flux capacitor and improve the carrion effect as ravens and others come to scavenge free electrons from the oxygen cloud surrounding the tube plazma 

As a former sound engineer, cant say I have ever heard jitter even on very early implementations of DAT recorders from Sony.

[TimFox]

This kind of snarky comment about demonized audio enthusiasts without technical education is typical of these sites.
Feel free to discuss absurd ideas about cables sold to the unwary by charlatans.
However, the design of low-jitter clock circuits is a quantitative design problem.

Yes, OFE copper is specified to be 99.99% pure copper with < 0.0005% oxygen.
You may find it in Cat6 ethernet cables due to its mechanical properties, mainly resisting metal fatigue.
When I used it in good vacuum systems, the informal test of an alloy was to take a flat piece and clamp it in a vise.
Regular copper would fail after only a few flexes back and forth with a ViseGrips pliers, but OFHC could handle more than my boredom threshold.

[PCB.Wiz]

NJR (now Nisshinbo) make oscillator ICs, and they recently released a NJU6385 in DFN6 that is spec'd as a low jitter, low noise oscillator for Audio use.

It's unclear if this is a new design die, or a marketing re-package of existing parts ?
Their Icc/Vcc plots do not suggest any fancy regulation on-chip.

https://www.nisshinbo-microdevices.co.jp/en/products/quartz-crystal-oscillator-ic/

 NJU6385
Phase noise   Phase noise (-103dBc/Hz(Typ.) @10Hz Offset  -163dBc/Hz(Typ.) @1kHz Offset
RMS Jitter   0.05psec(Typ.) @12kHz to 20MHz
Operating Current   9.0mA(Typ.) @CL=15pF.

[TimFox]

0.05 ps = 50 fs is comparable to the rough quantitative value of 28 ps (using probably a low estimate of the voltage noise) that I posted above.

[SiliconWizard]

Of course this is just marketing. Bonus point if someone did a proper teardown and found out that the tube was actually not connected to anything.

Not even talking about the microphonic behavior of tubes and how even small vibrations could throw it all off. Maybe you need to mount this equipment on 10 tons of concrete or something, like the very high-end turntables.

If anyone here is able to design a tube-based oscillator that exhibits better frequency accuracy and lower jitter than a pure solid-state approach with modern ICs, I'll be happy to have a look at it.

[ataradov]

There are a high-res pictures of it. It is connected. It does buffer the output of that  24.576MHz crystal oscillator. And the traces go somewhere else. No idea where it is used though.