I would consider to drop that circuit tout-court. Maybe you will get a much better preamp making a cascode + DC coupled cathode follower where you take the bias for upper valve of the cascode from a partition of the cathode resistor of the cathode follower and that will work as feedback network as well. You need 3 ECC88's in total for both channels. Such circuit is described in the Morgan Jones book.
Hi,
that's not the point. I have a cat 1 Ebay set, I got it as a birthday present, now I would like to listen to the Clone sound magic... ;-)
So my technical question remains: Change 1K to 1M, and keep the tapping point? Or do I need to change the tapping point? I guess, the lower tapping point is due to DC conditions...
And yes: I think the is a superior topology: Maybe use the Impasse input stage without transformer and a SLCF as second stage...
But this is my idea, not CAT 1 topology. And I would like to understand the topology behind...
that's not the point. I have a cat 1 Ebay set, I got it as a birthday present, now I would like to listen to the Clone sound magic... ;-)
So my technical question remains: Change 1K to 1M, and keep the tapping point? Or do I need to change the tapping point? I guess, the lower tapping point is due to DC conditions...
And yes: I think the is a superior topology: Maybe use the Impasse input stage without transformer and a SLCF as second stage...
But this is my idea, not CAT 1 topology. And I would like to understand the topology behind...
My two cents:
I think things with this circuit are much more complex than they appear at first glance.
My simulations show that if you put 1m instead of 1k the distortion doubles. For example the second harmonic goes up from 0,440 mV to 0,750 mV in the same measuring conditions. In addition I can easily observe some nonlinearity in the lower register.
The most specific and important thing about the CAT pre is that it’s a current feedback circuit. I think that makes all the active components and stages interdependent. I think one cannot easily change only one resistor value. Changes should be applied everywhere to work properly.
I think things with this circuit are much more complex than they appear at first glance.
My simulations show that if you put 1m instead of 1k the distortion doubles. For example the second harmonic goes up from 0,440 mV to 0,750 mV in the same measuring conditions. In addition I can easily observe some nonlinearity in the lower register.
The most specific and important thing about the CAT pre is that it’s a current feedback circuit. I think that makes all the active components and stages interdependent. I think one cannot easily change only one resistor value. Changes should be applied everywhere to work properly.
No it isn't. The feedback samples the output voltage, and subtracts a fraction of it from the input voltage. Normal global negative feedback.Asen said:
The increased distortion you see with the 'correct' circuit could be due to the 12AX7 being badly biased. If it is on a curvy part of its characteristic then running a similar identical circuit as an active load, as this circuit does, will give some cancellation of even-order distortion. This is a useful trick for non-linear valves but not necessary for the 12AX7, which is very linear when used properly.
Given that nobody would design a circuit like that, I can only assume that the 'designer' fiddled with it until he got acceptable results, being unaware of how the circuit actually functions.
Well, I am far from saying I understand the circuit completely, but I was referring to R14. I thought it makes the input stage sensitive to current as far as all the idle current of the end-stage tube flows through it, and after that through R8 together with the current of the input stage. Maybe I’m not reading the topology correctly.
It is slightly unusual to put the output CF quiescent current through the input stage cathode resistor, but that does not make it current feedback.
Two snags with doing it this way:
- the input stage bias could be disturbed by the output stage
- the feedback loop will try to maintain frequency response down to DC but it can't do this because of the interstage coupling cap so any large subsonic input will create distortion in the first stage (there is no input coupling cap to limit this problem).
This is one of those circuits which shows more and more mistakes the longer you look at it.
Two snags with doing it this way:
- the input stage bias could be disturbed by the output stage
- the feedback loop will try to maintain frequency response down to DC but it can't do this because of the interstage coupling cap so any large subsonic input will create distortion in the first stage (there is no input coupling cap to limit this problem).
This is one of those circuits which shows more and more mistakes the longer you look at it.
Hi.
I believe that the way to design such a circuit is to first design it on paper, using published tube characteristics and a ruler. I know that 12AX7 has the reputation of being very linear, but that is on paper and not in an improperly biased circuit. I would then breadboard the circuit and verify how well it works. I would start without negative feedback, which is a little hard in a circuit with DC coupled feedback. When I have verified that the circuit operates properly and "clips" symmetrically for a slightly overloading input, I connect the feedback. The reason I do it like that, is that negative feedback hides so many design errors that are difficult to find with the negative feedback active. My design methodology does require the use of an oscilloscope and a sine-wave generator, but that should be minimum requirements for a design-effort.
Circuit simulation is in my opinion not replacements for reality. If simulation says that there is distortion in a circuit, that may mean nothing more than the simulator has distortion, and that is it. The models used for characterizing a device is nothing more than a model.
Hans J Weedon.
I believe that the way to design such a circuit is to first design it on paper, using published tube characteristics and a ruler. I know that 12AX7 has the reputation of being very linear, but that is on paper and not in an improperly biased circuit. I would then breadboard the circuit and verify how well it works. I would start without negative feedback, which is a little hard in a circuit with DC coupled feedback. When I have verified that the circuit operates properly and "clips" symmetrically for a slightly overloading input, I connect the feedback. The reason I do it like that, is that negative feedback hides so many design errors that are difficult to find with the negative feedback active. My design methodology does require the use of an oscilloscope and a sine-wave generator, but that should be minimum requirements for a design-effort.
Circuit simulation is in my opinion not replacements for reality. If simulation says that there is distortion in a circuit, that may mean nothing more than the simulator has distortion, and that is it. The models used for characterizing a device is nothing more than a model.
Hans J Weedon.
Ok , it's not high- end , but why it is so badly designed ? , it's good that the feedback is taken without any capacitor and directly to to the first stage , it will be better to place the feedback onto lower value cathode resistor ( of course then we have to make some other changes in the input circuit ) or via an resistor devider , and the input coupling cap it will not solve anything , plus that it will add some phase shift in the circuit .
I think, very low frequency disturbance will not occur. So only bad biasing may be the error. I suggest to reduce the cathode resistors of the second stage. What is an optimum current? May someone propose optimum biasing for this schematic and tubes, would be a very good tutorial... May someone propose the best current, I try to calculate the correct resistor value and explain my ideas...
Hi there.
Digging back in my memory-banks, (75 years of them) I remember that good rule of "thumb" is to bias a triode amplifier at about 2/3 of the supply-rail. in other words at 1/3 the voltage across the plate resistor and 2/3 the voltage across the tube.
That is reasonable for a resistive plate-load, but with a second tube as a load, my guess, without having used that configuration, would be mid-way between the supply rail and ground. My guess would be a reduced voltage swing equal to 1/2 the supply voltage, versus 2/3 the supply rail for a resistive load. The advantage of a properly biased tube as a load resistor would be significantly reduced even harmonic distortion.
Unfortunately human hearing is not very bothered by even harmonic distortion, so reducing that type of distortion does not improve the acoustical perception much.
I believe that even harmonic distortion falls into the non-descriptive term "warmth"
Hans J Weedon.
Digging back in my memory-banks, (75 years of them) I remember that good rule of "thumb" is to bias a triode amplifier at about 2/3 of the supply-rail. in other words at 1/3 the voltage across the plate resistor and 2/3 the voltage across the tube.
That is reasonable for a resistive plate-load, but with a second tube as a load, my guess, without having used that configuration, would be mid-way between the supply rail and ground. My guess would be a reduced voltage swing equal to 1/2 the supply voltage, versus 2/3 the supply rail for a resistive load. The advantage of a properly biased tube as a load resistor would be significantly reduced even harmonic distortion.
Unfortunately human hearing is not very bothered by even harmonic distortion, so reducing that type of distortion does not improve the acoustical perception much.
I believe that even harmonic distortion falls into the non-descriptive term "warmth"
Hans J Weedon.
Hi Hans,
but your rule just talks about voltage distribution, so it totally depends on choosen anode resistor values.
So my question is: What could be the "perfect" current flowing. Reducing the cathode resistors ramps up the current. How do I determine the optimized current???
In the ImPass preamp, an optimun current was choosen for the 6SN7...
What is desired for the 12AX7?
After I got a good value there, I recalculate the sensing resistor for getting the same drive for the upper tube...
but your rule just talks about voltage distribution, so it totally depends on choosen anode resistor values.
So my question is: What could be the "perfect" current flowing. Reducing the cathode resistors ramps up the current. How do I determine the optimized current???
In the ImPass preamp, an optimun current was choosen for the 6SN7...
What is desired for the 12AX7?
After I got a good value there, I recalculate the sensing resistor for getting the same drive for the upper tube...
OK. If a design looks like one thing but isn't (in this case looks like mu-follower but isn't) it is often a sign that the designer didn't know what he was doing.Dimitris AR said:
Read the data sheet or Vade-Mecum or RCA handbook. These give recommended working points for ECC83/12AX7 as a normal voltage amplifier. 260V is a bit low for a supply rail, but we have an active load so we can aim for about 1mA anode current. That means Rk should be about 1-1.5k. 3k is too small. The 47k can be smaller so it doesn't drop too much voltage. 10k should be fine. Then the 1k to the upper grid should be 1M or thereabouts. Finally, if you want it to be an actual mu-follower take the signal from the upper cathode (although this won't make much difference as it is driving a CF).HVfanatic said:
Why did the designer pick Rk=3k? I suspect he looked at the datasheet, where 200V and 250V supply options are given. Both use a large anode resistor (to get linearity) so both have to use low anode current and hence a largish cathode resistor. As we have an active load we can use a modified version of the datasheet options for higher supply rail - they run at higher current and get more gain and less distortion. Maybe the designer did not realise this, or just had a bad day. Anyway, having picked a bad value for Rk he then perhaps played around and found that he could reduce distortion by artificially forcing more balance between upper and lower parts of the 'mu-follower' - which by now was no longer a mu-follower but a bad active load. 'Bad' because the 47k drops DC voltage (which we don't want) but does not add much AC anode load (which we do want).
There are two ways to use an SRPP/mu-follower. A genuine SRPP is balanced and designed to match its load. Under these conditions the even-order distortion in the upper and lower valves are cancelled, so it really does act a bit like push-pull. This is the best way to use a non-linear valve such as ECC82/12AU7. For a linear valve such as ECC83/12AX7 or ECC88/6DJ8 it is better to use a mu-follower to provide a high anode load. No need for balance, as the valve is linear enough on its own.
Hi,
I got the point with the lower cathode resistor. Could you link the datasheet you used?
What I don't understand properly: I thougt, the drive resistor is determined by the wanted drive to the upper valve. Im my opinion, the 10K brings only 25% of drive, so modulation of the upper valve is poor. Or does the higher current improve the amplification factor of the valve, so that only 10 is needed??? Or does the voltage drop becomes unacceptable now with 47K?
I always thought, the SRPP hast to low drive to the upper valve, because the cathode resistor is only determined by bias requirements, and upper valve modulation is a happy sideffect...
I always thought, u-follower is more perfect push-pull then SRPP... And betafollower would be constant current anode load alike.... MHMMMM...
I got the point with the lower cathode resistor. Could you link the datasheet you used?
What I don't understand properly: I thougt, the drive resistor is determined by the wanted drive to the upper valve. Im my opinion, the 10K brings only 25% of drive, so modulation of the upper valve is poor. Or does the higher current improve the amplification factor of the valve, so that only 10 is needed??? Or does the voltage drop becomes unacceptable now with 47K?
I always thought, the SRPP hast to low drive to the upper valve, because the cathode resistor is only determined by bias requirements, and upper valve modulation is a happy sideffect...
I always thought, u-follower is more perfect push-pull then SRPP... And betafollower would be constant current anode load alike.... MHMMMM...