For your fun: an all-positive phase preamp. For the acronym hunters: APP preamp
But it has a lot of 3rd harminix, which is not nice - transssistor sound, you know . . .
But it does improve once a CCS of 2 mA is used instead of the 4K7: 2ndH -40 dB, 3rd -68 dB. So that is also a good contender this way for a simple 12AU7 preamp. I made this a few years ago.
But it has a lot of 3rd harminix, which is not nice - transssistor sound, you know . . .
But it does improve once a CCS of 2 mA is used instead of the 4K7: 2ndH -40 dB, 3rd -68 dB. So that is also a good contender this way for a simple 12AU7 preamp. I made this a few years ago.
- There are several CARY PCB's for sale, just change one part (R --> CCS) and leave another (C --> shunt wire) out.
Last edited:
Illuminating the second version, in which you combined with the first stage ... I had tried it too but in an approximate way (however, as here I have seen that with Vg-mos ≈ 9 / 10Vb + I had better results).
It would also be interesting to know the output impedance for each of these circuits. It should be below 100ohm (or a few tens). Maybe to think about for a headphone amplifier ...
Rongon, thanks for the files. Given my older version I take into account the different models for trials.
I wonder about this focus on linearity; Langford-Smith build a laboratory pre-amplifier with an amplification factor of 20 having a distortion of 0.002% using a 12AT7 and 6BQ5. According to many who "know" both of these tubes are not the most linear around and something better should be used. 0.002% sounds good enough to me.
AWV Radiotronics January 1958 VOl. 23 page 14
The late Patrick Turner liked below preamp. You have to go to the archive.org website to find his archived website and all the details.
AWV Radiotronics January 1958 VOl. 23 page 14
The late Patrick Turner liked below preamp. You have to go to the archive.org website to find his archived website and all the details.
Last edited:
I agree. MOSFET source followers are a very good thing for tubes.
A while ago, to teach myself how to use a pcb layout app, I made up a pcb for a 12AT7 DC-coupled to a source follower, with a shunt NFB loop.
I forgot to provide pads for split plate resistor and coupling cap for bootstrapping the 12AT7 plate load (oops). Maybe another time.
I have a lot of vintage 12AT7s, so I want to play with them.
Besides, I like tubes. I have poor taste, I know. I wear plaid pants too.
(The blue trace is the input waveform, at the grid of U1. The green trace is the output waveform, at OUT.)
0.007% THD at 1V rms output (1kHz) into 10k ohm load. Not bad for this simple of a circuit.
I need to build this, to goose the output from my input switchbox/AVC. Yeah, I have really poor taste. I know. I just gotta be me, ya know?
Edit to add: My reasoning for this circuit is to try to take advantage of the nice advantages an anode follower (common cathode with parallel NFB) provides, including adjustable gain (could switch in another 182k resistor in series with R5 to double the gain), decently high input impedance (~60k ohms), and low THD. However, the anode follower has drawbacks. The input impedance could be higher (I could raise it to 100k by changing R2 to 100k and R3 to 300k) and the tube has to drive both the negative feedback loop and the external load. That's where the source follower comes in. It unloads the tube from the external load and the negative feedback loop's current demands, so now it only has to drive the gate of the source follower, which will have both very high input impedance and very low input capacitance. The THD with the source follower added is far lower than a simple 12AT7 anode follower. This looks pretty good to me, and it sure is simple.
I whipped up a simulation of the above circuit but using a J309 into an IRF510, with 48V B+. The results are almost identical.
Last edited:
I ran valve based SLCF for over a decade. First one had a MOSFET on the top. I was very happy and they served me well. However when I built my fet based one I noticed no drop in performance but loved the energy saving and the absence of needing to source valves every three years or so.
Allen Wright added woo and complexity for his fans to a solid circuit
Last edited:
You are way too sensible for the Tubes/Valves forum. We're kind of wacko here.
Yes, that 'Super Linear Source Follower' circuit looks like a real winner. By far the best of the ones I listed. It could be scaled for use with lower voltage transistors and an nJFET in front with a parallel NFB loop ("drain follower"?). Just one B+ of only 48VDC, no heaters, no negative supply. Thank you for sharing that idea. That's (yet) another item on my list now.
Yes, that 'Super Linear Source Follower' circuit looks like a real winner. By far the best of the ones I listed. It could be scaled for use with lower voltage transistors and an nJFET in front with a parallel NFB loop ("drain follower"?). Just one B+ of only 48VDC, no heaters, no negative supply. Thank you for sharing that idea. That's (yet) another item on my list now.
I need to leave for work, but I whipped up a sim of J310 into a SLSF of 2x IRF610 with DN2535 source CCS, with 100V B+.
I get similar performance to the 12AT7-MOSFET circuit posted in #205. Not actually any better, except that it doesn't need a heater supply and regular parts replacement won't be necessary. I think the higher distortion of the individual devices makes it a wash between FET with more NFB vs. triode with less NFB, at least in regard to THD at 1V out.
I get similar performance to the 12AT7-MOSFET circuit posted in #205. Not actually any better, except that it doesn't need a heater supply and regular parts replacement won't be necessary. I think the higher distortion of the individual devices makes it a wash between FET with more NFB vs. triode with less NFB, at least in regard to THD at 1V out.
Of course interesting these different variations on the theme with the flavour of hybrid with mosfets or fets (in particular the "new" SLSF),
Anyway I was positively surprised by that 1958 circuit. I tried to simulate it and it actually has a very low THD, up to a relatively high Vp-p (out).
The cathode-cathode feedback being between "odd" tubes is negative, right? The resistive network on the V3 grid is also ingenious to optimize the Z (out) as they claimed.
That circuit from 1958 is a Ring of Three, which we discussed earlier. The nice feature here is that the output buffer (the third stage), uses a big honkin' 12W power pentode (EL84) wired triode, and there is that odd way of DC-coupling the cathode of the EL84 to the cathode of the first 12AT7. But it's still a Ring of Three.
Rk of the EL84 is 5.6k ohms, which is a huge value of Rk for an EL84. My guess is that the large Rk will cause degeneration (negative feedback) in that stage. If that's all the NFB there is, doesn't this circuit have a lot of gain?
The feedback from the cathode of of the cathode follower is non-inverting compared to the input.
The cathode of the first 12AT7 is also non-inverting.
That does look like positive feedback, doesn't it...
A nice feature is the 'noiseless bias' on the grid of the cathode follower.
You'll notice that the NFB loop contains two poles, one of .025uF and 1M ohms, the other of 0.05uF and about 1M ohms.
Rk of the EL84 is 5.6k ohms, which is a huge value of Rk for an EL84. My guess is that the large Rk will cause degeneration (negative feedback) in that stage. If that's all the NFB there is, doesn't this circuit have a lot of gain?
The feedback from the cathode of of the cathode follower is non-inverting compared to the input.
The cathode of the first 12AT7 is also non-inverting.
That does look like positive feedback, doesn't it...
A nice feature is the 'noiseless bias' on the grid of the cathode follower.
You'll notice that the NFB loop contains two poles, one of .025uF and 1M ohms, the other of 0.05uF and about 1M ohms.
Last edited:
Ah ok, now I recall "the ring of three" you mentioned before...
Yes the Rk for V3 is 5.6kohms. In the simulated circuit I noticed how the feedback influences greatly (as an improved fashion): without that trick waveforms become not good and THD increase a lot.
Yes the Rk for V3 is 5.6kohms. In the simulated circuit I noticed how the feedback influences greatly (as an improved fashion): without that trick waveforms become not good and THD increase a lot.
Oh my goodness, that was one colossal brain fart I had! Well duh.
Now that that has finally sunk into my thick skull, I realize that the feedback loop is essentially the same as in a Ring of Two, with the feedback signal coming off the plate of the second triode and is applied to the cathode of the first triode.
Since the cathode follower is non-inverting, the signal at the output is basically the signal from the plate of the second triode, but buffered by the cathode follower.
Therefore, just as in the Ring of Two topology, the feedback in this circuit is negative feedback.
It is not a problem if there is only one value of gain required. You can always make he circuit stable. That is one reason why early audio mixers used fixed gain stages with attenuators in between.
Cheers
Ian
I just noticed that the Langford-Smith Lab Amp has gain of 20X (26dB) closed loop, which is far too much for a typical line stage these days.
I wonder what would happen if you tried to reduce the gain to 3X by increasing the NFB by another 16dB.
My guess is that you'd have to start adding compensation caps to suppress very low and very high frequency resonances/oscillations.
Perhaps this would be a worthwhile circuit to try using a 12AU7 for the first and second stages, so that less NFB is needed to get down to 3X gain.
I wonder what would happen if you tried to reduce the gain to 3X by increasing the NFB by another 16dB.
My guess is that you'd have to start adding compensation caps to suppress very low and very high frequency resonances/oscillations.
Perhaps this would be a worthwhile circuit to try using a 12AU7 for the first and second stages, so that less NFB is needed to get down to 3X gain.
I found the article with the Langford-Smith Laboratory Amplifier. I think there has to be a mistake in the schematic.
As drawn, the feedback loop comes straight off the cathode of the EL84, lifted up at least 100V above ground.
The feedback loop then goes through a 22k resistor to the cathode resistor of the first stage 12AT7.
That forms a DC voltage divider as well as an AC one. The voltage at the cathode of the first 12AT7 will likely be +6VDC.
This will bias the first 12AT7 extremely cold, perhaps in cutoff. In simulation the Ip looks like about 140uA (micro-amperes).
Why would you do that? To reduce the gain?
Don't you think the feedback should be taken off the output after the output DC blocking cap (32uF or 1uF)?
Or did he do that because of the two different paths to output blocking caps?
I'm mystified by that.
As drawn, the feedback loop comes straight off the cathode of the EL84, lifted up at least 100V above ground.
The feedback loop then goes through a 22k resistor to the cathode resistor of the first stage 12AT7.
That forms a DC voltage divider as well as an AC one. The voltage at the cathode of the first 12AT7 will likely be +6VDC.
This will bias the first 12AT7 extremely cold, perhaps in cutoff. In simulation the Ip looks like about 140uA (micro-amperes).
Why would you do that? To reduce the gain?
Don't you think the feedback should be taken off the output after the output DC blocking cap (32uF or 1uF)?
Or did he do that because of the two different paths to output blocking caps?
I'm mystified by that.
Last edited:
Playing some more with the LabAmp in LTspice, I can see that it's a bear to make it stable if you apply more NFB to reduce the gain. But the circuit as published does make about 17X gain with 0.005% THD at 2V rms out into a 47k ohm load.
I swapped in a 12AU7 for the first two stages and that reduced the gain a little, about -1.5dB. THD doesn't change much.
If you move the feedback to come from after the DC blocking output cap, the gain from the first stage goes way up, and so does the open loop gain from the whole amp. But that causes sharp resonances both at the ultrasonic and infrasonic ends of the spectrum with NFB applied.
The ultrasonic response would now require compensation by bypassing the NFB resistor with a small cap, adjusting the value while observing the output of the amp on as scope with square wave input.
The infrasonic response now requires a huge value for the DC blocking output cap from the cathode follower. A big peak at under 1Hz only starts to go away with C = 100uF.
This is an amazing circuit in that it plays with triodes in such a crazy way. The gain changes with the value of the parallel NFB resistor because a smaller value relative to Rk of the first stage 12AT7 changes the bias on that triode, and the gain with it, along with the NFB.
I think the circuit might be too dependent on individual tube characteristics, so making a stereo pair matched in gain might be more trouble than it's worth.
Still, quite an interesting circuit, and the lowest THD around (like 0.0006% at 1V out into 50k ohms).
I swapped in a 12AU7 for the first two stages and that reduced the gain a little, about -1.5dB. THD doesn't change much.
If you move the feedback to come from after the DC blocking output cap, the gain from the first stage goes way up, and so does the open loop gain from the whole amp. But that causes sharp resonances both at the ultrasonic and infrasonic ends of the spectrum with NFB applied.
The ultrasonic response would now require compensation by bypassing the NFB resistor with a small cap, adjusting the value while observing the output of the amp on as scope with square wave input.
The infrasonic response now requires a huge value for the DC blocking output cap from the cathode follower. A big peak at under 1Hz only starts to go away with C = 100uF.
This is an amazing circuit in that it plays with triodes in such a crazy way. The gain changes with the value of the parallel NFB resistor because a smaller value relative to Rk of the first stage 12AT7 changes the bias on that triode, and the gain with it, along with the NFB.
I think the circuit might be too dependent on individual tube characteristics, so making a stereo pair matched in gain might be more trouble than it's worth.
Still, quite an interesting circuit, and the lowest THD around (like 0.0006% at 1V out into 50k ohms).
Last edited: