Hey indaco,
I tried to run your .asc for the SLCF but it wouldn't go in my system. The error log said there are too many instances of IRF510 and IRF610 (??).
I think there are some problems with the .asc.
When I have some time, I'll copy the circuit into an all-new .asc to try to get rid of those errors, and run it. Unfortunately, it's Monday and it's time to get real life started.
I tried to run your .asc for the SLCF but it wouldn't go in my system. The error log said there are too many instances of IRF510 and IRF610 (??).
I think there are some problems with the .asc.
When I have some time, I'll copy the circuit into an all-new .asc to try to get rid of those errors, and run it. Unfortunately, it's Monday and it's time to get real life started.
Sorry that it doesn't work on your program the file seems ok. In case try to change the mosfet (if you want and when you have a spare time).
I used DN2540 because it gives the best results but IRF830 / 820 I think are fine too. Probably given the 100V of B + also the lower series, with no problem...
PS: Yes, that's why me too I'd prefer to avoid silicon components (not just because a tube purist...) but also I wonder if I would still have the perception of the "sound" of the tube with those hung-up. Thanks meantime for the file, I'll take a look...
I used DN2540 because it gives the best results but IRF830 / 820 I think are fine too. Probably given the 100V of B + also the lower series, with no problem...
PS: Yes, that's why me too I'd prefer to avoid silicon components (not just because a tube purist...) but also I wonder if I would still have the perception of the "sound" of the tube with those hung-up. Thanks meantime for the file, I'll take a look...
Last edited:
You can make a SLCF with all triodes with just two dual triodes a side. This is with an input triode up front for gain.
I built mine using ECF80's which allows the CCS on the bottom of the stack to be configured as a pentode for better performance.
4 Triodes and not a single transistor is about standard for many preamps.
My point would be, if you are going to go to all the trouble of placing an active load on your Cathode Follower - why not add a single extra triode/pentode and get even better performance. When you get to a circuit at this level of performance you will hear no "valve sound" at all.
My current main preamp is a SLCF with all Jfets and Mosfets - better than my valve version.
Last edited:
I do exactly the same, always multiply the input rms level by 1.414 before filling in the value of the voltage source.
I agree, it is very easy to get confused but I guess they do it because the input waveform is not necessarily a sine wave.
Cheers
Ian
That makes perfect sense. If you have it measure a 1.414V peak sine wave, it's going to show the fundamental at 0dB on the FFT display. If you give it a 1V peak square wave, it will be different. (If I was a real engineer, I'd know what the FFT would show. But I admit it... I'm a hack. er... hobbyist.)
I'm not trying to start a brawl, or anything of the sort, but... If you design a circuit with valves to have no 'valve sound' at all, then what exactly is the point of using valves in the first place? I mean... I actually like the "valve sound"... sometimes.
I'm finding that I like my phono (RIAA) preamp to be made with tubes, simply because I like the way tubes sound amplifying and EQing vinyl records. Too clean of a preamp and records sound like... just... records. The same as the equivalent CD but with some vinyl noise, clicks and pops. What fun is that?
For my digital source (a Raspberry Pi with Allo BOSS DAC, or a laptop with a modest Topping D10 DAC), I find that adding a 'tubey sounding' tube line stage just muddies up the sound. I love that cheap DACs have finally started to sound really good. No more 'tubefier' needed. It's either that or I've gone deaf. That is entirely possible. I was a lot younger in 1990. But my old SACD and CD players all sound worse to me than 16bit/44.1kHz FLAC files playing through the Raspberry Pi. Maybe I'm just talking myself into thinking that...
That would be a SLSF, wouldn't it? Not that there's anything wrong with that. 🙂 And I'm not at all surprised that it would be a better performer (objectively) than the tube version. A transparent buffer with <0.001% THD at 1V out into a 10k ohm load with 1000pF of cable capacitance? What's not to love?
I ask the same question. What's the point? The only time I prefer not having tube sound is in buffer circuits, when I want it as neutral as possible and capable of driving difficult load.
Hey indaco,
I re-drew your schematic into my LTspice installation, using the models I have installed on my system.
After some problems with the way my standard.bjt file was formatted, I got the simulation to run.
First, can you tell me what the purpose is of the resistor network composed of R12, R9, R2, and R10?
It looks to me like R8 (47k) comes right off the signal source, feeding straight into the base of Q1.
Then R12 and R9 are two 33k resistors in parallel, making 16.5k ohms.
From there, R2 is a 1k resistor in series with R12||R9, and then R10 is in series with all of that, connecting to the grid of U1 (the 12AU7).
That basically makes an 18.5k resistors in series between the signal source and the grid of U1.
Why?
I simplified all that into a single 18k grid stopper for U1. The end result looks like this:
Does that look right to you?
I simulated that circuit, and I get a huge oscillation at well over 5MHz. (Notice that I ran the .ac AC analysis with a 10MHz bandwidth so I could see the oscillation.)
The FFT simulation is taking a really really long time to finish. Something's not right.
Last edited:
Hello Rongon,
in reality I copied the diagram from a forum ... there was a potentiometer (50k) at the input for which I inserted a resistive divider at 50% (even if I don't see its usefulness, being a buffer). Instead the other value I inserted it was by mistake.
My simulation hadn't problems of long-time run...have you set all parameters in proper way?
Anyway basically you obtained about my same results, with the pitch in the HF range (3.3MHz). I noticed this may be tamed through a small bypass capacitor (1-10nF) on R6 (above the mosfet on the bootstrapped triode). E.g. with 3.3nF, THD improved further a little bit.
Here's the plot after this modification.
PS: Kaypirinha, as agreed I sent the three boards this morning with priority mail
in reality I copied the diagram from a forum ... there was a potentiometer (50k) at the input for which I inserted a resistive divider at 50% (even if I don't see its usefulness, being a buffer). Instead the other value I inserted it was by mistake.
My simulation hadn't problems of long-time run...have you set all parameters in proper way?
Anyway basically you obtained about my same results, with the pitch in the HF range (3.3MHz). I noticed this may be tamed through a small bypass capacitor (1-10nF) on R6 (above the mosfet on the bootstrapped triode). E.g. with 3.3nF, THD improved further a little bit.
Here's the plot after this modification.
PS: Kaypirinha, as agreed I sent the three boards this morning with priority mail
There's M1 DN2540 between plate and +Vdc. What does this FET exactly do? Is it also a CCS? If so, I'll have to ask what the triode might do in between two CCS's?
Best regards!
Best regards!
When you tried to mimic the volume control, you didn't do it correctly. The volume control should be a voltage divider, with one series resistor (R1 in the circuit below) and one shunt resistor (R2).
The reason my simulation wouldn't render was because I was running the .trans directive to measure THD. LTspice dutifully rendered the oscillation on the waveform, which takes a lot of processing time.
I'll try the 6922 version in the image from your last post and see if that performs any differently. But now we've left the topic of this thread, which is 'how to make a good line preamp using ECC82', correct?
Now that we've gone off-topic, I modeled a plain old cathode follower using a 12GN7A wired triode, with a 250V B+ and a 10k cathode load resistor. I used a self-biased arrangement with a red LED for cathode bias, a 470k ohm grid leak resistor, and a 0.22uF input DC blocking capacitor. I used a 4.7uF output cap, with a 470k output load resistor. I used a 10k ohm external load and 500pF in parallel to simulate the input of a Class D amp with a couple meters of interconnect cable. I got THD of 0.005% and very wide bandwidth. No oscillation to tame, so no need for any extra shunt capacitors.
That's what happens if you use a tube that's a lot better than a 12AU7.
12GN7A-triode has low rp of less than 2000 ohms, high gm of greater than 15mA/V, higher mu of more than 30 (but high input capacitance).
The lower rp makes it very good at driving tougher loads.
The higher mu gives it more open loop gain to lower distortion closed loop.
The higher gm is how it gets that higher mu and lower rp.
12GN7A in triode makes for a pretty darn good cathode follower, even without active loads.
Add an active load (a MOSFET CCS in the cathode) and it gets really good. Like op-amp good. But it's a tube.
However... it's not an ECC82.
Last edited:
You're right, absorbed in the circuit I was forgetting the basics... the shunt R of course.
I do not know this type of tube you mentioned, I saw that it is a pentode (equiv. 12hg7). So I guess it is connected in a triode configuration in that circuit.
From the analysis of your simulation I trust that it is more suitable, but then it would actually fall outside the thread...
I do not know this type of tube you mentioned, I saw that it is a pentode (equiv. 12hg7). So I guess it is connected in a triode configuration in that circuit.
From the analysis of your simulation I trust that it is more suitable, but then it would actually fall outside the thread...
I think should be asked to a expert guy for such a configuration, in the case is Shoog (who first revisited the circuit above).
The FET is there to ensure that the voltage across the triode is constant. Since the triode has a constant voltage across it and a constant current through it - it operates in its most linear way. The SLCF was developed by one of the big tech companies to present a stable load to the front end of an oscilloscope and a very linear drive to the screen tube.
Indeed, if that is a sink circuit it's as you say...I realized that I have incorrectly copied the schematic regardless (then the simulation fooled me).
I think there must be more wrong than just that. I ran an AC analysis and it still has a +13dB spike at ca. 6.5MHz.
I found that if I raise the value of the gate stopper on the DN2540 to 22k ohms or more, that kills the oscillation.
Perhaps the DN2540 is not the best part to use as the plate constant voltage source. Maybe DN2540 has too high gate capacitance or Crss?
Maybe. I substituted an STN3N45 MOSFET in the plate of the 12AU7 and the oscillation went away.
I think the problem with the DN2540 was that the B+ of 100V was too low for it. It was only dropping 10V drain to source. I find that device wants 50Vds, at least.
I raised the B+ to 150V and now the DN2540 works.
I ran a transient response on this and it shows THD with 1V rms out at 0.0155%. It's all H2. That's half the THD of a plain old 12AU7 cathode follower with a resistor cathode load. Still, the 12AU7 Bootstrapped Pair with shunt NFB shows lower THD and gives you adjustable gain, and uses no transistors.
I think all this is dependent on implementation. If you want to make a really good cathode follower, use a better triode than 12AU7 — unless you're going for 'tube color'. A 12GN7A-Triode (or a similar RF pentode like EF184 in triode) will make a really good cathode follower with just a cathode resistor for its load. But if you want a cathode follower with op-amp-like low THD, try a 12GN7A-triode in a SLCF. Or a 6922 or 6DJ8. Something with higher mu and higher gm than 12AU7.
Last edited:
I think I found another mistake in the schematic for the SLCF...
MPSA92 is a PNP transistor, but is shown as an NPN.
I put an MPSA92 in as the top transistor in the 12AU7 cathode (on top of the BC550) and the circuit's not working.
I think MPSA92 is the wrong part there. That has to be a mistake.
A 2N3019 works as the top transistor.
Now that I have this SLCF simulation working, I tried it with a 6DJ8. It works far better with that tube (higher gm, higher mu than 12AU7).
With a 6DJ8 the THD is below 0.005%.
With a 12AU7 THD is 3x higher (0.015%).
I would expect THD to be somewhat higher in real life, in both cases.
MPSA92 is a PNP transistor, but is shown as an NPN.
I put an MPSA92 in as the top transistor in the 12AU7 cathode (on top of the BC550) and the circuit's not working.
I think MPSA92 is the wrong part there. That has to be a mistake.
A 2N3019 works as the top transistor.
Now that I have this SLCF simulation working, I tried it with a 6DJ8. It works far better with that tube (higher gm, higher mu than 12AU7).
With a 6DJ8 the THD is below 0.005%.
With a 12AU7 THD is 3x higher (0.015%).
I would expect THD to be somewhat higher in real life, in both cases.
Last edited:
Yes, I then noticed that the diagram taken contained some mistake. Reading in the forum it was said that instead of the MPSA092 it was the MPSA042 in reality.
I inserted the DN2540 into the simulation thinking of an excellent component in terms of characteristics, not taking into account the low B + of the circuit, moreover even the current does not enter the optimal operating range.
Have you also tried changing the C for bootstrapping (e.g. lower value) to keep V as constant as possible at the triode? As far as I know then usually R gate stopper between 100R-1K should be fine.
Just for curiosity, if you want to take a look here, Allen Wright designated phono stage, among other things makes use of a "common" IRF830...
https://keith-snook.info/riaa-stuff/Secrets of the phono stage - Allen Wright.pdf
I inserted the DN2540 into the simulation thinking of an excellent component in terms of characteristics, not taking into account the low B + of the circuit, moreover even the current does not enter the optimal operating range.
Have you also tried changing the C for bootstrapping (e.g. lower value) to keep V as constant as possible at the triode? As far as I know then usually R gate stopper between 100R-1K should be fine.
Just for curiosity, if you want to take a look here, Allen Wright designated phono stage, among other things makes use of a "common" IRF830...
https://keith-snook.info/riaa-stuff/Secrets of the phono stage - Allen Wright.pdf