• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Tube output buffer for DAC....needs advice!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
HI!

I have built the "Elektor DAC 2000" some time ago. Efter modding it with AD797 instead of OPA627 the sound has improved a great deal.

I would like to have switchable tube output though. Like the Shanling CD-players have.

I need advice on:

1. What kind of tube design to add (SSRP, etc.)?
2. What kind of tubes to go fore? I have a few 12B4A, 6N1P, E80CC to try out.
3. Where and how to break in to mu exsisting circuit to add the new tube output?

Any ideas?

Best regards,

Peter Lund
 

Attachments

  • dac-output.jpg
    dac-output.jpg
    95.8 KB · Views: 663
this particular lower end dac isn't too bad, but I would waste the time to build one up from scratch.

the design considerations will be how good the PSRR of the tube stage you select.

Personally, a DAC with balance audio out would be better to build up a tube line stage with.

the I to V resistor network would be something to look at. OEMs have a tendency to go on the cheap here on those parts.

the clock is very very important. this is the other place they go cheap at. A PLL type clock is far superior to the basic crystal.
 
DavesNotHere said:
A PLL type clock is far superior to the basic crystal.
PLLs are always inferior to the crystal clock to which they are synchronised. PLLs are therefore used when some degree of adjustment is needed, as in an off-board DAC box. The PLL in the DAC needs to lock onto the crystal clock in the CD player (or other digital source), and is thus dependent on the low frequency jitter behaviour of that clock.
 
DavesNotHere said:
crystal clocks are not accurate. I guess with consumer products it doesn't matter that much.

but the the clocks used to record it is much more critical:
A good crystal clock is plenty good enough for professional and domestic audio. Low jitter is what matters: obtained by good design and high crystal quality. Long-term frequency accuracy is irrelevant as even a poor crystal clock is better than our ears or any musical instrument.

A PLL, by definition, has to lock onto something else (that is what the first L is: 'lock'). It will have poorer jitter than the thing it locks to.

mytek 192cx, apogee big ben, BLA Microclock --- all PLL.
Just out of interest I googled the first clock you mentioned (mytek 192cx). It is a high quality clock plus dividers and buffers. It can take an external clock input too. The blurb does not mention PLL. If in fact it does include a PLL (for frequency multiplying?) then this will degrade the quality of the external clock input, as I have said.
 
crystal clocks are not accurate. I guess with consumer products it doesn't matter that much.

but the the clocks used to record it is much more critical:

mytek 192cx, apogee big ben, BLA Microclock --- all PLL.

Crystal clocks can be as accurate as you want. I have one that has long term error of about one part in 10^13. That is totally "nuts" level of accuracy. Cost about $200 to get to that level.

Whenever you talk about more than consumer level, you have to have a "standard" to compare to. We can't talk about accuracy unless you can measure it. Lucky for us that there is now GPS.

As for a cycstal vs. a PPL. The PPL can never be as good as the source it is locked to. Even the best has to add some noise. We can argue if this matters and past a point clear it does not (after all your head moves because of the blood pumping through it) You a part in a billion.

Generally a high-end crystal oscillator will be "tuned" with a varactor diode and can have it's frequency pushed or pulled over some range. THe assembly is housed in a temperature controlled box and the output monitored and compared to a standard, GPS is the most common standard. A system like this is about as good as it gets. but not really needed for audio work
 
Thats just silly talk. I test many CDCE and CDCM jitter cleaner EVMs as they are made.
The PLL can lock onto a frequency many times the crystal. When divided back down to
a frequency similar to the original, the phase noise is in hundreds of femtoseconds...

Long term phase noise is still no better than the crystal, but the short term can be
many times better. I see it. I measure it. Not that type of EVM every day, but often
enough...

Temperature of the crystal would be the thing to tighten up on. I see cheap crystals
drift all over, compared to either of my rubidium refs...
 
Last edited:
A PLL is necessarily no better than the device it locks to for low frequency jitter. At mid frequencies jitter is set by the loop filter, so it could be better or worse. At high frequencies jitter is set by the PLL's own oscillator which in most cases will be worse than a crystal. Even if the PLL uses a crystal itself it needs to vary the frequency and this will worsen the jitter as it is an extra source of noise.

Long term drift is irrelevant for audio, so there is no need for rubidium references. Short-term jumps in frequency might be a problem, but decent crystal quality should avoid this.

Perhaps some of the confusion arises because to me 'PLL' is a circuit architecture, but to others 'PLL' might be a frequency source or type of oscillator.
 
What is the definition of long term/short term phase noise, and how do you measure it? I guess you need a super accurate and ultra low noise in the frequency and/or amplitude domain test rig.

This is a good question because there is an exact answer that does not depend on opinion.

Error in the phase of a sine wave can be expressed ad some added frequency component. Phase noise is defined as the power of those added components per Hz of signal.

It is probably better th=o think of "jitter" which is 100% that same thing as phase noise but expressed in the time domain. Jitter is simply the error in the period of the clock. For example a 1000Hz clock should have a period of 0.001 second. But might have a 1 sigma error of .00001 seconds.

We all know that periods and be converted to and from frequency. Phase noise is just looking a "jitter" from the point of view of frequency. If you one you know both.

So there you are clocking samples out of a DAC. The question is how bad can the error in the clock get before you can hear it. Tiny errors in the clock produce frequncy errors that no tweeter on Earth can reproduce, For example nanosecond level errors produce frequncy errors in the radio spectrum

On the other hand a millisecond level error is dead in the middle of the human voice range. Clearly we can hear the voice and clearly can can't hear microwaves. So the threshold of tolerable error is some place between.

No the thing is that even a cheap crystal has nanosecond level or better error.

Next topic LONG term error. This is easy to understand. The crystal is very tiny "jitter" but what if the freqnucy of "off". This causes a tiny shift on pitch. Can you hear it? that depends on how much "off" the crystal is. But again even a cheap crystal is at worst 100 parts per million. Far better then any's ability to hear.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.