His Master's Noise: A Thoroughly Modern Tube Phono Preamp

Thanks SY for a nicely written article, from 13 years ago. I'm hitting my junk boxes and ordering parts.

I have some questions.
  1. With all the loads being Constant Current, giving close to zero current fluctuations in the power supply, why bother with separate DC supplies? (cross-talk/channel isolation is a figure of merit, with becomes moot point with two speakers in the same room )
  2. With constant current sources, why not use cathode resistors for bias? There will be no negative feedback with no cathode voltage variation, needing a big capacitor, right?
  3. "blocking" = overload , clipping?
  4. "Build-out resistor", is this output series resistor? I guess in this case the R7 value is trimmed to compensate for the variation in source impedance from calculated value? Would a cathode follower buffer have eliminated this? Maybe even a 12AX7?
  5. Big old choke in the DC HT supply, why? The load is a constant current fed by DC voltage regulators that would absorb any noise, like any 120Hz rectifier noise getting past the 470uF cap. (Bad recovery characteristics in rectifiers can excite resonances in the transformer and give a nasty noise, using soft recovery diodes, or bypass these with caps and/or resistors would tame those diodes. LM317 is series with a HV MOSFET works too. )
  6. "carbon resistors for the gate and grid-stoppers", do you mean carbon film or carbon composition?
  7. When you say the cartridge has source resistance of 15 Ohm, being transformed x100, you mean impedance? (My HMC10 has a DC resistance of 6 Ohm and a recommended load impedance of greater than 20 Ohm. A 1:10 transformer would then require minimum resistance of 2k. As exact power transfer is not a central issue, but noise and voltage gain, we can forgo the RF regimen for impedance matching.)
 
Uptick

I read through your comments/questions and they all sound familiar. I built this and Sy’s more recent MM preamp. They are both challenging, matching parts and the ccs’s is tricky and is more than a novice or the faint at heart would expect for sure. Looking back at the builds for both I am reminded of how much I learned and the almost constant careful consideration they required. At the end of both I built pretty much exactly what was in the schematics. With HMN I run carts from 2-40 ohms and they all sound marvelous.

I think Sy was/is way ahead of the commercial world and the more common diy’er offerings. I don’t know how many have succeeded in crossing the finish line, But it is well worth the effort.

David
 
I have some some thought/questions about this design. I have made a board that I got ideas from this thread for.

DN2540 are in short supply, I have a bag of the SMT version on hand from years ago, but the current setting resistor value is likely to have to be tested out first, as the spread is quite wide, gate-to-source off voltage -1.5V to -3.5V for 10uA, and then the thermal factor has to be considered ~ +/- 10% around 25C, maybe not a showstopper.

The now expensive/rare input pentode, has a very fine grid structure with sever current limits, like 0.5uA in the Philips data sheet. They recommend 1K in series with the grid for fixed bias, with more than 10uF over a self bias resistor. The transformer here has a very low DC resistance ~100 Ohms, why not use a coupling cap and ground the grid thru a high value resistor maybe 500K?
The MC pickup load resistor can be on either primary or secondary of the transformer, as it may need some damping.

The buffer stage may not be needed with the output connected to the Q3's source. The input stage can also have the load connected to Q1's source. That dirty semiconductor business in still in pie anyway.

Active plate followers do not behave like resistors, so why the high plate supply voltage? If the output max is say 10V p-p and the input is 30 times less, the CC followers can swing close to the power supply value. Maybe 150V is OK?
 
The advantage of a Constant Current source is that the loadline is horizontal, which typically means the triode's most linear operating "region". I did the MOSFET+LM431+LD150+Zener HV constant current source over thirty years ago, before I saw it anywhere.
Attaching the load to the FET's source means you have a source follower and it can operate nicely with very crazy low loads.
A resistor loaded tube may very well be OK for a very low voltage swing, like that D3a. Not a challenging requirement for 0.5V output.

For a big voltage swing, like a 805, or 211 triode grids, more is needed and a resistor loaded tube cannot operate well with signal approaching the rail voltage. I designed a driver back then that could swing 600V p-p with 0.1% THD using sweep tube pentodes at output followers, loaded by chassis mounted resistors. Big, heavy and hot!
 
The manufacturers data sheets (Philips, Ericsson) for D3a/7721 all have a test condition that includes a self bias resistor (400R) and 100K in series to positive grid voltage (+10V).
I guessing this may be for an RF application? (SPICE sim really blows for phono with that setup, Eb 190 and 20mA thru tube)
Can't find any RF circuit for this tube.
The Hickok Cardmatic's uses 240R and +7.5V, Eb 160V, E210F similar high Gm pentode
D3a_LM.png
.
 
I made a couple of different PCB's for the D3a + 6DJ8, but substituted the DN2540's with my LM431 (-type) circuit to a get precision current setting, along with 50dB better power supply rejection.
Not sure if my noise performance is as good as the DN2540, a SPICE simulation does not really give a good answer for noise.
Can anyone here confirm the number for the output noise for the SY design?
"...Total noise was measured .... and found to be (unweighted) 0.16mV..."
Various AC voltmeters/DMMs have different bandwidths which will affect the results.
 
Last edited: