Yes, I have measured the 12AU7 under similar conditions. In fact I measured quite a range of tubes including the 12AX7 and a few 6N1P. The tests were inspired by reading Morgan Jones' first book in the days when I was looking for a mic pre design without NFB. This was more then 10 years ago. The results are in a notebook somewhere. I will see if I can find it and copy them here but the 12AU7 was definitely measurably worse than almost any other tube I tested.
Cheers
Ian
It did make it into the 9 pin world as the ECC87. But except for the amplifier/mixer in the link, I don't know of any other equipment in which the ECC87 was used:
Klangfilm KL 5418 Amplifier/Mixer - Radiomuseum
The data of the E80CC comes close to that of the ECC40.
Last edited:
Well, it does look like ECC82 is not as linear as other types.
But what if you have a sleeve of them and you want to build something?
This little line stage circuit looks like it should be OK. 2.5X gain, 0.025% THD at 1V rms output into 50k ohm load (THD looks like it nulls into a 10k load, but I don't trust that in simulation).
Granted, this is a mere simulation, but the real life circuit would be quite close, and it looks like it would be a pretty good line stage preamp thing.
It could be made better using ECC81, ECC88, ECC40, ECC99, etc. But if all you have is a sleeve of ECC82...
C1 could be removed if you're sure there's no subsonic garbage coming in and the power supply has very low impedance.
D4 and D5 could be removed if there are no solid state devices downstream.
D1 and D2 could be red instead of green.
But what if you have a sleeve of them and you want to build something?
This little line stage circuit looks like it should be OK. 2.5X gain, 0.025% THD at 1V rms output into 50k ohm load (THD looks like it nulls into a 10k load, but I don't trust that in simulation).
Granted, this is a mere simulation, but the real life circuit would be quite close, and it looks like it would be a pretty good line stage preamp thing.
It could be made better using ECC81, ECC88, ECC40, ECC99, etc. But if all you have is a sleeve of ECC82...
C1 could be removed if you're sure there's no subsonic garbage coming in and the power supply has very low impedance.
D4 and D5 could be removed if there are no solid state devices downstream.
D1 and D2 could be red instead of green.
Well, reading the pros and cons (mostly the latter) of using the 12AU7 in practice leads to discourage a circuit like the one proposed at the beginning. But as I said I'm not looking for a nit and I like to experiment with slightly different sounds at the expense of a slightly increased distortion.
In this regard, as described on Tubelab, it occurred to me to try a ccda with this configuration, where the role of the small positive feedback to the CF is praised. In addition I would like to try at the same time to add a GNF (about 200k) to further stabilize everything. The simulation so far looks promising ...
In this regard, as described on Tubelab, it occurred to me to try a ccda with this configuration, where the role of the small positive feedback to the CF is praised. In addition I would like to try at the same time to add a GNF (about 200k) to further stabilize everything. The simulation so far looks promising ...
Attachments
The CCDA ceases to have equal-but-opposite current draw if the load changes much. In practice, that's probably not a problem driving the input of a power amp. But if the volume control comes after this line stage, then that would be a problem. Solution is to put the volume control before the line stage.
The positive feedback to the cathode of V1 basically does the same thing as the LED cathode load for V1 in the circuit I posted previously. The results are almost identical. The amount of NFB is about the same. The resulting gain is about the same. The THD turns out to be slightly improved.
It looks like a worthwhile improvement.
6N1p tubes i tried and compared to others sounded bad no matter what was done with bias when the circuit was designed well. Several people did measure them finding high distortion, some of it higher order than with comparable tubes. Their low cost is probably the only reason Chinese or some beginners like them so much.
So it "creates distortion" when kneecapped by feedback shunning religious fanatics? Seems to me that using feedback to correct distortion and not overloading the tube with input signals theres plenty of extra gain available to put feedback around in order to create a nearly distortion free amplifier. So its really quite possible to make a low distortion amplifier using 12AU7 and multiple people have done it.
If you decide not to use feedback then the measured performance of just about everything you build is not going to be optimal...
You should contextualize your categorization of the 12AU7 as a "distortion creator" so as not to confuse the newbies. 12AU7 are cheap, plentiful and easy to work with. First time DIYers stand a great chance of success using simple cheap 12AU7s. In fact that's what got me excited about building tube amps. I had the tubes, and was able to make a really good preamp, for not a lot of money.
Just be honest is all I'm asking.
Used within its design parameters a 12AU7 is quite good, and low distortion amplifiers are not only possible but have been selling like hot cakes for going on 70 years using 12AU7.
One example being a McIntosh MC30/40/240.
Distortion creator: Hogwash.
If you decide not to use feedback then the measured performance of just about everything you build is not going to be optimal...
You should contextualize your categorization of the 12AU7 as a "distortion creator" so as not to confuse the newbies. 12AU7 are cheap, plentiful and easy to work with. First time DIYers stand a great chance of success using simple cheap 12AU7s. In fact that's what got me excited about building tube amps. I had the tubes, and was able to make a really good preamp, for not a lot of money.
Just be honest is all I'm asking.
Used within its design parameters a 12AU7 is quite good, and low distortion amplifiers are not only possible but have been selling like hot cakes for going on 70 years using 12AU7.
One example being a McIntosh MC30/40/240.
Distortion creator: Hogwash.
I wrote 6N1P was tested and showed pretty bad distortion content, not ECC82.
My tube journey started with ECC82 preamp where one was half was adding gain, the other half served as a buffer. It had way too much gain we didn't know how to lower at the time, today i would use NFB if same schematic and tube was all i could work on.
Or rather cheap Edcor transformer at the output because it lowers gain, makes galvanic isolation, make things sound interesting, etc.
My tube journey started with ECC82 preamp where one was half was adding gain, the other half served as a buffer. It had way too much gain we didn't know how to lower at the time, today i would use NFB if same schematic and tube was all i could work on.
Or rather cheap Edcor transformer at the output because it lowers gain, makes galvanic isolation, make things sound interesting, etc.
In your post #11 you wrote:
"Pardon my ignorance but by what force of nature does a 12AU7 "create distortion" I've used them and they dont "create distortion" any more so than any other vacuum tube in my admittedly minimal experience.
So in order to put this silly rumor to rest why don't you link us all to done raw data that shows the 12AU7 to be a "distortion creator".
..."
So I posted some raw data for the ECC82/12AU7, just like you asked for. It shows that the ECC82/12AU7 is less linear than the ECC83 and the ECC40.
If your "feedback shunning religious fanatics" was (also) intended for me, than please explain how you came to that description. I didn't mention 'feedback' in my posts at all
Thanks for your contribution, on the first inverting stage wouldn't be more appropriate a higher Ia though (i.e. with Ra 30-40k and Rk ~300Ω)? A steeper slope is usually preferable in the graph as far as I know...
Looks like we're lapsing back into the 'NFB vs. zeroNFB' conflict.
I think it's been pretty well proven that NFB is a good thing if implemented correctly.
One of the definitions of 'implemented correctly' is as few reactive elements inside the loop as possible. Direct coupling helps. The simple CCDA preamp circuit has only one capacitor inside the feedback loop, all other components are either resistors or triodes. It should work well and 'sound good' with some NFB around it.
Another supposedly horrible distortion maker is the humble 12AT7/ECC81. It's not very linear on its own, but it has a lot of gain (high mu). Use it with some NFB and that gain becomes an advantage. For this use, 12AT7 comes out ahead of 12AX7 because of its lower rp and higher gm (can drive external loads better).
If you make a CCDA with 12AT7 and some NFB, it could be really clean. Simulation says 0.009% THD at 1Vrms out into 50k ohms (1kHz). Freq resp -0.03dB at 20Hz and 20kHz. That's not quite op-amp clean, but it's getting there...
If you use a 12AT7 with a MOSFET source follower and forget about the CCDA, you could get the ideal version of the above. The MOSFET drives the NFB better, and the 12AT7 supplies lots of voltage gain. The MOSFET would have to be one with low gate capacitance though.
A 12AX7 could be used instead of 12AT7 with a MOSFET source follower, but then the MOSFET would need really low gate capacitance. It can be done...
Edit to add:
For the first triode/inverting stage, a steeper slope (higher transconductance) is better. However, the cathode follower has very high input impedance, so it's easy to drive. I don't think gobs of gm are absolutely needed here. High gain does help, though. One could certainly experiment with a couple of D3a or 6J9P in triode and throw both high mu and a lot of gm at the problem. But I thought we were talking about 12AU7s and the like. That's why I brought up 12AT7. They're 'guitar amp' tubes. You can buy them at the nearest musical instrument store.
One more addition:
12AU7 is not as linear as 12AX7 or ECC40.
However, 12AU7 is not much (if any) worse than some other popular triodes like 6DJ8/ECC88,
I enjoy the sound of zero-NFB circuits. They sound more 'alive' to me than amplifiers with lots of NFB. I understand I'm enjoying a distortion. I can still like it. I'm allowed.
I think it's been pretty well proven that NFB is a good thing if implemented correctly.
One of the definitions of 'implemented correctly' is as few reactive elements inside the loop as possible. Direct coupling helps. The simple CCDA preamp circuit has only one capacitor inside the feedback loop, all other components are either resistors or triodes. It should work well and 'sound good' with some NFB around it.
Another supposedly horrible distortion maker is the humble 12AT7/ECC81. It's not very linear on its own, but it has a lot of gain (high mu). Use it with some NFB and that gain becomes an advantage. For this use, 12AT7 comes out ahead of 12AX7 because of its lower rp and higher gm (can drive external loads better).
If you make a CCDA with 12AT7 and some NFB, it could be really clean. Simulation says 0.009% THD at 1Vrms out into 50k ohms (1kHz). Freq resp -0.03dB at 20Hz and 20kHz. That's not quite op-amp clean, but it's getting there...
If you use a 12AT7 with a MOSFET source follower and forget about the CCDA, you could get the ideal version of the above. The MOSFET drives the NFB better, and the 12AT7 supplies lots of voltage gain. The MOSFET would have to be one with low gate capacitance though.
A 12AX7 could be used instead of 12AT7 with a MOSFET source follower, but then the MOSFET would need really low gate capacitance. It can be done...
Edit to add:
For the first triode/inverting stage, a steeper slope (higher transconductance) is better. However, the cathode follower has very high input impedance, so it's easy to drive. I don't think gobs of gm are absolutely needed here. High gain does help, though. One could certainly experiment with a couple of D3a or 6J9P in triode and throw both high mu and a lot of gm at the problem. But I thought we were talking about 12AU7s and the like. That's why I brought up 12AT7. They're 'guitar amp' tubes. You can buy them at the nearest musical instrument store.
One more addition:
12AU7 is not as linear as 12AX7 or ECC40.
However, 12AU7 is not much (if any) worse than some other popular triodes like 6DJ8/ECC88,
I enjoy the sound of zero-NFB circuits. They sound more 'alive' to me than amplifiers with lots of NFB. I understand I'm enjoying a distortion. I can still like it. I'm allowed.
Last edited:
Oh, sorry... I didn't fully understand your comments.
Yes, you could increase the Ip of the first 12AU7 (U1) and make the circuit a CCDA. The thing is, I don't know if that improves anything. You want as much gain as possible from that first stage, to provide negative feedback. I was going for as much voltage gain as possible from U1 since the U2 cathode follower presents such a light load to U1. You could rearrange things to make Ip of U1 = Ip of U2.
Also, 12AU7 is weird in that it's more linear at low Ip (like 2mA) than it is with higher Ip (like 5mA or more).
The idea of a CCDA is that the balanced current draw from U1 and U2 make the circuit's current draw appear more constant to the power supply. A higher impedance from the power supply becomes less of a drawback. If you have a good regulated psu with low output impedance, then this may become a moot point. Or maybe the CCDA 'sounds better' because it makes the regulator work less hard? I really don't know.
ECC82 is absolutly fine in any circuit , when you need low gain , don't turn it in to a distortion histeria 
Be assured you won't hear or see on the scope nothing unusual using it .
In all designs the key is to optimizes the circuit so it could generate much higher voltage than you need , let's say 55Vrms , even when you only use 20V . Then the distortions are proportionally lower .
Be assured you won't hear or see on the scope nothing unusual using it .
In all designs the key is to optimizes the circuit so it could generate much higher voltage than you need , let's say 55Vrms , even when you only use 20V . Then the distortions are proportionally lower .
In fact, I have to change my mind for a moment, I relied mainly on simulated assumptions rather than theory.
In reality, the ccda predicts the Va to be about half (or slightly less) of Vb +.
Considering the circuit without positive feedback and without C (byp) from my calculations:
Iq = 300 / (47000 + 8000 + [21] 470) = 4.6mA
But the Va is about 2/5 of Vb +, so to have the value recurring to 150V:
Rk = (47000 - 8000) / 21 ≈ 1.8K, corresponding to 3.2mA of Ia, a value that would be in agreement with the "goodness" of the results between 2-3mA (therefore not such a crazy thing).
I hope this interpretation is right in the context, then obviously we have to consider the real circuit and the presence of the feedback that leads to lowering both Rk and Zout ...
To make the 12AU7 CCDA with the positive feedback to V1 cathode and NFB from output to V1 grid, you have to arrange for V1 and V2 to draw the same plate current (Ip).
1. Choose the plate resistor value for V1.
2. The cathode resistor for V2 will be the same value.
3. You will need the Vp (plate voltage) of V1 to be 1/2 of B+ voltage. If B+ = 300V then Vp for V1 will be 150V.
4. Now you can figure out the Ip of V1 and the value of Rp for V1.
Let's say you decide on 30k ohms for Rp. That 30k ohms will need to drop 150V.
Ip = Vp/Rp
Ip = 150V/30,000R = 0.005 (5mA)
Now you need to look at the 12AU7 plate curves to see what the grid bias will be for Vp = 150V and Ip = 5mA. That looks like about -4.6V to me.
4.6V/0.005A = 920R
I think 910R is the closest standard value.
Let's not worry about the positive feedback yet. Let's put together our prospective circuit in its most simple and basic form. Plugging the circuit values into LTspice, I came out with this (using Adrian Immler's model of a PhilipsECG 12AU7):
Yes, the method works. Voltages and currents came out very close to predictions.
Now we add the positive feedback. The bottom of R6 gets disconnected from ground and then connected to the cathode of V1.
The combined plate currents of both V1 and V2 (about 10mA) will now be flowing through R2. The value of R2 will need to be reduced enough for V1 to see 4.7V at its cathode. Since V1 and V2 will be drawing 10mA, the math is easy.
4.7V/0,01A = 470R
R2 needs to be changed to 470 ohms. In simulation it came out close, but actually 510R got the two triodes to equalize at 5mA Ip each. It's hard to get the two triodes to draw exactly the same current, so I think it's not useful to split hairs. Either 470R or 510R will work.
With positive feedback in place, the gain goes up, but so does THD. So we want to add negative feedback to reduce gain and lower THD. For that, we can use an 'anode follower' circuit, but with the cathode follower acting as a buffer (which is a good thing). That's plate to grid 'shunt' or 'parallel' NFB.
Let's say our target is a gain of about 2.5X. We'll make the values of the two feedback resistors define a 3:1 ratio. That would define 3X gain if the open loop gain was infinite, but it's not even close. The losses should put us close to 2.5X gain closed loop. I'll pick a voltage divider made of a 300k feedback resistor and 100k series resistor. Like this:
The feedback is now defined by a virtual ground at the input grid of V1, made from R3 in series and R1 in parallel. R1 is the feedback resistor ('shunts' the amplified and inverted signal from the output back to the input), and R3 is the series resistor forming the voltage divider that defines how much output signal is fed back to the input.
The DC conditions don't change at all. Gain is reduced from about 15X down to 2.256X. That's about 16dB of NFB.
Now that we know the gain is 2.256X, we can figure out what input signal voltage will be required to get 1V rms (1.414V peak) at the output of our little amplifier.
1.414V/2.256 = 0.627V (627mV peak)
I put a 627mV 1kHz sine wave signal to the input of the simulated circuit, and it says the output will be 1.4135V peak. Close enough.
It also says the THD at 1kHz will be 0.0264%, which is quite decent. It's all 2nd harmonic, of course.
The frequency response looks like it will be flat down to DC and -3dB down at 365kHz. That's in simulation, of course. It's likely to be not that good in real life. But it should be close and more than good enough.
It's a feedback amplifier, so now you have to be careful of stability. You don't want it to ring or oscillate. However, there are no reactive elements inside the negative feedback loop except for the output coupling cap C1. LTspice predicts no nasty ringing or oscillations, because the open loop gain isn't that high (a 12AU7 only has mu of 15 to 20).
You could make C1 larger in value, perhaps 4.7uF, and that should increase LF stability a little. Or, you could install an input blocking cap of a large enough value to allow low audio frequencies through but filter out subsonic noise. Maybe 1uF would do well (it forms a high pass filter with R3). That would make LF stability very solid.
There's always more to a circuit than meets the eye. Feedback complicates things too. But in a simple circuit like this, you can see how everything goes together fairly easily.
--
1. Choose the plate resistor value for V1.
2. The cathode resistor for V2 will be the same value.
3. You will need the Vp (plate voltage) of V1 to be 1/2 of B+ voltage. If B+ = 300V then Vp for V1 will be 150V.
4. Now you can figure out the Ip of V1 and the value of Rp for V1.
Let's say you decide on 30k ohms for Rp. That 30k ohms will need to drop 150V.
Ip = Vp/Rp
Ip = 150V/30,000R = 0.005 (5mA)
Now you need to look at the 12AU7 plate curves to see what the grid bias will be for Vp = 150V and Ip = 5mA. That looks like about -4.6V to me.
4.6V/0.005A = 920R
I think 910R is the closest standard value.
Let's not worry about the positive feedback yet. Let's put together our prospective circuit in its most simple and basic form. Plugging the circuit values into LTspice, I came out with this (using Adrian Immler's model of a PhilipsECG 12AU7):
Yes, the method works. Voltages and currents came out very close to predictions.
Now we add the positive feedback. The bottom of R6 gets disconnected from ground and then connected to the cathode of V1.
The combined plate currents of both V1 and V2 (about 10mA) will now be flowing through R2. The value of R2 will need to be reduced enough for V1 to see 4.7V at its cathode. Since V1 and V2 will be drawing 10mA, the math is easy.
4.7V/0,01A = 470R
R2 needs to be changed to 470 ohms. In simulation it came out close, but actually 510R got the two triodes to equalize at 5mA Ip each. It's hard to get the two triodes to draw exactly the same current, so I think it's not useful to split hairs. Either 470R or 510R will work.
With positive feedback in place, the gain goes up, but so does THD. So we want to add negative feedback to reduce gain and lower THD. For that, we can use an 'anode follower' circuit, but with the cathode follower acting as a buffer (which is a good thing). That's plate to grid 'shunt' or 'parallel' NFB.
Let's say our target is a gain of about 2.5X. We'll make the values of the two feedback resistors define a 3:1 ratio. That would define 3X gain if the open loop gain was infinite, but it's not even close. The losses should put us close to 2.5X gain closed loop. I'll pick a voltage divider made of a 300k feedback resistor and 100k series resistor. Like this:
The feedback is now defined by a virtual ground at the input grid of V1, made from R3 in series and R1 in parallel. R1 is the feedback resistor ('shunts' the amplified and inverted signal from the output back to the input), and R3 is the series resistor forming the voltage divider that defines how much output signal is fed back to the input.
The DC conditions don't change at all. Gain is reduced from about 15X down to 2.256X. That's about 16dB of NFB.
Now that we know the gain is 2.256X, we can figure out what input signal voltage will be required to get 1V rms (1.414V peak) at the output of our little amplifier.
1.414V/2.256 = 0.627V (627mV peak)
I put a 627mV 1kHz sine wave signal to the input of the simulated circuit, and it says the output will be 1.4135V peak. Close enough.
It also says the THD at 1kHz will be 0.0264%, which is quite decent. It's all 2nd harmonic, of course.
The frequency response looks like it will be flat down to DC and -3dB down at 365kHz. That's in simulation, of course. It's likely to be not that good in real life. But it should be close and more than good enough.
It's a feedback amplifier, so now you have to be careful of stability. You don't want it to ring or oscillate. However, there are no reactive elements inside the negative feedback loop except for the output coupling cap C1. LTspice predicts no nasty ringing or oscillations, because the open loop gain isn't that high (a 12AU7 only has mu of 15 to 20).
You could make C1 larger in value, perhaps 4.7uF, and that should increase LF stability a little. Or, you could install an input blocking cap of a large enough value to allow low audio frequencies through but filter out subsonic noise. Maybe 1uF would do well (it forms a high pass filter with R3). That would make LF stability very solid.
There's always more to a circuit than meets the eye. Feedback complicates things too. But in a simple circuit like this, you can see how everything goes together fairly easily.
--