C13 shouldn't impair the RIAA compliance.
D1, as Chris points out, should be an IR LED with Vf of 1.2V.
D1, as Chris points out, should be an IR LED with Vf of 1.2V.
Does anyone know what the supply voltage is? 300 V? For minimum RIAA- and A- or ITU-R 468-weighted noise, the input 12AX7 triode should run at roughly 0.75 mA to 0.9 mA of anode current (calculated based on Merlin Blencowe's noise measurements).
Regarding the 2 % wrong time constant of the first pole due to my approximation, I've checked what happens when you try to correct for it by aiming for a 2 % too small time constant. That trick works, the pole then ends up almost exactly where it should, and the only component value that changes by more than 0.02 % is the value of R26. Its calculated value goes from 139.715 kΩ to 135.443 kΩ, the nearest E96 value changes from 140 kΩ to 137 kΩ.
That is, if you are perfectionistic enough to worry about small deviations below 50 Hz, you can tweak R26 to correct for both the inaccuracies due to the valve's internal resistance not being precisely as estimated and the inaccuracies due to my approximation. However, unless the load is very high impedance, the output coupling capacitor C10 probably causes greater deviations.
Regarding the 2 % wrong time constant of the first pole due to my approximation, I've checked what happens when you try to correct for it by aiming for a 2 % too small time constant. That trick works, the pole then ends up almost exactly where it should, and the only component value that changes by more than 0.02 % is the value of R26. Its calculated value goes from 139.715 kΩ to 135.443 kΩ, the nearest E96 value changes from 140 kΩ to 137 kΩ.
That is, if you are perfectionistic enough to worry about small deviations below 50 Hz, you can tweak R26 to correct for both the inaccuracies due to the valve's internal resistance not being precisely as estimated and the inaccuracies due to my approximation. However, unless the load is very high impedance, the output coupling capacitor C10 probably causes greater deviations.
Last edited:
It would be interesting to see what improvements could be possible by making the second stage an anode follower with RIAA wrapped around it, ala MarcelvdG's published phono stage. Without an output (cathode) follower, the very low end EQ gets pretty dependent on second stage's open loop gain, so maybe leave its cathode degenerated with a 1K0 cathode resistor. External load would also have to be assumed to be pretty high, at least 100k, ideally maybe 250k. More is just never available in modern equipment.
For anyone interested, the series "build-out" resistor just after the first stage can be surprisingly large without serious noise penalty - I've used 750k without issue. This helps to keep RIAA reactive loading on the second stage manageable (very small caps). Then first stage anode load can be upped to 250k to 300k, battery bias or IR diode bias, no degeneration. Second stage anode load 100k? Tough call, as always.
All good fortune,
Chris
For anyone interested, the series "build-out" resistor just after the first stage can be surprisingly large without serious noise penalty - I've used 750k without issue. This helps to keep RIAA reactive loading on the second stage manageable (very small caps). Then first stage anode load can be upped to 250k to 300k, battery bias or IR diode bias, no degeneration. Second stage anode load 100k? Tough call, as always.
All good fortune,
Chris
Hi Jack,
Yes.. Chris is correct.. I’ve tried IR LED’s in a mod before.. I’m just thinking if it’s still too (Vf=1.2V) much of a high bias voltage Vs. the Vf=~0.7V I’m getting for the input stage right now..
These little tiny bias voltages are also a tough call. Approaching 0 VDC, even while still negative, grid current can become significant, and with higher source impedances can be a surprise source of distortion - grid current is very non-linear with voltage, and isn't especially predictable, varying with manufacture, age, phase of the moon.
All good fortune,
Chris
All good fortune,
Chris
Hi Marcel,
B+ is <250V, C13 is spec’d at 47uF/250V.. But I used 350V caps..
Thanks again for adding more insight and the more accurate R26 trimming.. Actually, I used 470nF for C10 as that’s what I have in stock.. I did not think a slightly higher value would hurt for the output cap..
Hi Chris,
Is an IR LED still worth a try? Also IR LED’s in series for the 2nd stage? The B+ is on the lowish side..
An old rule of thumb was 1 Volt for medium sized valves, but that was also regularly broken in lots of famous historical preamps by famous historical manufacturers. Grid leak bias was also commonly used, and although grid voltage couldn't be measured directly, one could work back from measured anode voltage and published curves to get numbers in the half a Volt range. A separate issue is that LEDs at under a milliAmp are way down in their knee, so very non-linear. Both of these effects are mitigated by the small signal levels. A small enough section of a curve is a tangent.
The classic 12AX7 phono preamp is endlessly challenging. It kinda falls into the same category as a dog walking on two legs - usually a bit awkward, but then there's a Golden dancing the merengue.
Battery bias is well suited for phono stages, too. Or just well filtered DC from supply.
All good fortune,
Chris
The classic 12AX7 phono preamp is endlessly challenging. It kinda falls into the same category as a dog walking on two legs - usually a bit awkward, but then there's a Golden dancing the merengue.
Battery bias is well suited for phono stages, too. Or just well filtered DC from supply.
All good fortune,
Chris
Why is the bias on the second stage (U2) so cold? Two red LEDs in series would drop about 3.2V or a bit more. Two IR LEDs in series would drop about 2.4V, which is still pretty cold for a 12AX7 with B+ = 300V.
Also, R31 at only 68k ohms is a low value for a 12AX7. I'd expect it to have low gain and higher distortion.
Also, what does this circuit drive? What is the input impedance and input capacitance of the following stage? One can't look at the stage in isolation, ignoring whatever might be interacting with it downstream (or upstream, for that matter). Is this part of an integrated amp with a switching network and volume control after this phono stage? What value of volume control pot?
Also, R31 at only 68k ohms is a low value for a 12AX7. I'd expect it to have low gain and higher distortion.
Also, what does this circuit drive? What is the input impedance and input capacitance of the following stage? One can't look at the stage in isolation, ignoring whatever might be interacting with it downstream (or upstream, for that matter). Is this part of an integrated amp with a switching network and volume control after this phono stage? What value of volume control pot?
I used to work on these a lot. I don’t have the schematic but here’s what I remember.
The whole phono stage is over on that sideways board using the two 6SL7’s. And that’s what the schematic that’s shown is drawn for. Not 12AX7’s. It was a very simple phono stage using just the two 6SL7 tubes.
Everything on the PC board is all line stage. The two 12AX7’s are the gain stage. And then there’s two I believe, 6SN7’s as the final buffer output stage. If I remember right, at the front of the PC board, up near the volume control, the lines of resistors, diode, small cap are all for powering the switching and LEDs. It’s been a while so it’s hard to remember exactly. I’d have to see the whole thing again to be positive of what everything was on the lie stage board.
The phono stage over on the sideways metal piece was always a pain to work on. Easy to get a little shock.
The whole phono stage is over on that sideways board using the two 6SL7’s. And that’s what the schematic that’s shown is drawn for. Not 12AX7’s. It was a very simple phono stage using just the two 6SL7 tubes.
Everything on the PC board is all line stage. The two 12AX7’s are the gain stage. And then there’s two I believe, 6SN7’s as the final buffer output stage. If I remember right, at the front of the PC board, up near the volume control, the lines of resistors, diode, small cap are all for powering the switching and LEDs. It’s been a while so it’s hard to remember exactly. I’d have to see the whole thing again to be positive of what everything was on the lie stage board.
The phono stage over on the sideways metal piece was always a pain to work on. Easy to get a little shock.
B+ is less than 250 V and the diodes are ordinary diodes, presumably silicon ones, see posts 1, 4 and 26.
If folk want to use LEDs, or any forward biased diode, they can always bias them up onto the flat part of their voltage curve and get very reasonable linearity. Just a low voltage, quiet supply and a resistor - electrolytic capacitors as bypasses have their own issues at very low bias voltages. IME, early ESR growth in only a couple decades.
Still campaigning for battery bias! If not, a quiet, small DCV supply is easy enough, for the benefits it gives. (And adjustable!) Just cause Hank didn't play it that way is no reason we can't. I'll bet Hank would have insisted.
All good fortune,
Chris
Still campaigning for battery bias! If not, a quiet, small DCV supply is easy enough, for the benefits it gives. (And adjustable!) Just cause Hank didn't play it that way is no reason we can't. I'll bet Hank would have insisted.
All good fortune,
Chris
Hello.. The part with a metal plate on its side is the 6SN7 CF output stage mounted sideways; it was added in a later revision.. The Phono has 12ax7’s on the main board, the linestage has 6SL7’s on the main board.. The 6SL7 linestage R values has been revised too (my unit at least) where the plate load R’s are 47k, now actually are more suited to 6SN7’s.. It sounds better with 6SN7’s there, and the gain is more manageable..
Hi Chris,
Yeah.. There are folks who also use SiC diodes, but those also only work well with higher plate currents..
Yeah.. There are folks who also use SiC diodes, but those also only work well with higher plate currents..
The classic objection to forward biased diodes in a signal path is that they have curves where only girls should have curves. Well above the knee they start to look reasonably resistive, but they're certainly not the magic bullet that modern fashion implies. Through the knee all forward biased semi-con diodes are some version of exponential (voltage vs. current) so are little distortion generators gathering in our midst. Without proper biasing (well above the knee), maybe adequate, maybe worrying, depending on one's religion.
Battery, or some equivalent substitute, biasing in the grid path is a century old, so seems impossibly strange today, but removes these issues. In fact, for a stereo two 12AX7 phono preamp, a single grocery store battery (1.56 VDC) can bias all four triodes fine. Bolt those cathodes to chassis and live free, brothers!
"Play 'La Marseillaise'. I said play it."
-Rick Blaine, Casablanca 1942
Battery, or some equivalent substitute, biasing in the grid path is a century old, so seems impossibly strange today, but removes these issues. In fact, for a stereo two 12AX7 phono preamp, a single grocery store battery (1.56 VDC) can bias all four triodes fine. Bolt those cathodes to chassis and live free, brothers!
"Play 'La Marseillaise'. I said play it."
-Rick Blaine, Casablanca 1942
Does this look like the circuit under discussion?
or with 6SL7:
It seems kind of mean to ask a 12AX7 or 6SL7 to drive a 50k ohm load from its plate.
or with 6SL7:
It seems kind of mean to ask a 12AX7 or 6SL7 to drive a 50k ohm load from its plate.
C1 and C6 should be 47 uF and 100 nF, respectively. Whether V1 is 250 V or less is unclear, as is the value of Rload. Cout has been increased to 0.47 uF, and for the RIAA correction, two possibilities are under discussion.
The other is:
No 820 kΩ, 1 MΩ grid leak from U2's grid to ground.
R26 = 137 kΩ
R28 = 14.3 kΩ
R29 = 69.8 kΩ
C6 remains 100 nF
C7 remains 22 nF
C8 remains 1 nF
The other is:
No 820 kΩ, 1 MΩ grid leak from U2's grid to ground.
R26 = 137 kΩ
R28 = 14.3 kΩ
R29 = 69.8 kΩ
C6 remains 100 nF
C7 remains 22 nF
C8 remains 1 nF
Oh, sorry -- forgot to mention that I had added C1 (just a decoupling cap) and reduced the value of C6 to flatten the simulated bass response. You can remove C1 and increase C6 to 0.1uF and not much would change.,
Increasing Cout to 0.47uF shouldn't change much either.
If B+ is down around 230V, then that will reduce performance a bit more.
Does hat Rload of 50k represent a volume control pot?
The EQ looks fine, of course. I'm sure there are better ways to go about it, but this is a pre-existing circuit, so I understand. Tweaking.
Increasing Cout to 0.47uF shouldn't change much either.
If B+ is down around 230V, then that will reduce performance a bit more.
Does hat Rload of 50k represent a volume control pot?
The EQ looks fine, of course. I'm sure there are better ways to go about it, but this is a pre-existing circuit, so I understand. Tweaking.