One example to get you thinking: Take a single ended output tube, and put a perfect current source in its cathode. The current source will keep the cathode current constant, no matter what the grid voltage is. That would include the DC grid bias setting if using fixed bias, and also the grid signal from the driver tube. There will be no change in the cathode current, and so no change in in the plate current, and so no output from the output transformer when the grid is driven by signal (music). That would be a good reason to bypass the cathode current source.
For an input stage, if the plate current source is perfect, the tube plate still has to drive the coupling capacitor that drives the next stage's grid resistor to ground. Therefore, there will be a change of plate current. the changing plate current means there will also be a change in cathode current. A cathode resistor that is unbypassed will have a changing current in it, and so have a changing voltage across that cathode resistor.
With a perfect plate current source, the voltage across the unbypassed cathode resistor will be according to the ratio of the cathode resistor versus the next stages grid resistor.
For an input stage, if the plate current source is perfect, the tube plate still has to drive the coupling capacitor that drives the next stage's grid resistor to ground. Therefore, there will be a change of plate current. the changing plate current means there will also be a change in cathode current. A cathode resistor that is unbypassed will have a changing current in it, and so have a changing voltage across that cathode resistor.
With a perfect plate current source, the voltage across the unbypassed cathode resistor will be according to the ratio of the cathode resistor versus the next stages grid resistor.
For an input stage, the anode still has to drive the next stage grid resistor too.
For an output stage, CCS cathode bias is a good way of ensuring that the bias is mostly wrong. This is because it fixes the wrong current: average current instead of quiescent current. A resistor would be better. Fixed voltage bias better still, although it will not self-adjust to different valve samples or valve ageing.
For an output stage, CCS cathode bias is a good way of ensuring that the bias is mostly wrong. This is because it fixes the wrong current: average current instead of quiescent current. A resistor would be better. Fixed voltage bias better still, although it will not self-adjust to different valve samples or valve ageing.
Some tube types have more heater to cathode leakage than others. Some tubes heater to cathode leakage degrades over time. 1mV of injected hum from the heater to the cathode, will produce the same amount of hum as if you injected 1mV of hum into the grid. In either of these cases, you may experience hum because you did not bypass the cathode resistor.
As for calculating an appropriate bypass cap- I'm reading that the cathode resistance, just for the tube itself, is (Ra+ra)/(mu+1). That's in parallel with the cathode resistor for calculating the bypass capacitor. Since the CCS has a huge output resistance, am I correct that it could be ignored and the bypass capacitor calculated by (Ra+ra)/(mu+1)?
tapehead ted,
I had not even considered eliminating the bypass cap when I modified a circuit recently. It now has a current source in the plate circuit, and the plate drives the RC coupled grid of the next stage.
Just for fun, I will measure what the change is with, and then without, the cathode bypass cap. The grid resistor of the next stage is 270k Ohms, and that in parallel with an equal or even higher impedance of the current source, and when that is divided by u +1, is going to be many many times the resistance of the self bias resistor. I am predicting a change far less than 1dB.
I probably do not have enough spur free dynamic range in my FFT to see any difference in the harmonic levels between bypassed and unbypassed self bias resistor in the triode that has a current source in the plate circuit. I also do not expect the gain to change significantly. I shall see.
Where I would expect to see a difference between bypassed and un-bypassed is if I overdrive that stage and draw grid current. The self bias should shift when the bypass cap is present. And that bias shift, no matter the size will have recovery time due to the capacitance and the unbypassed impedance of that node (the total RC time constant).
If there is noise on the filament secondary, and if it has lots of high frequency energy, I do not expect it to introduce significant noise onto the cathode, because the self bias resistor is much lower than the Xc of the small pF filament to cathode and the large resistance of the filament to cathode leakage.
I had not even considered eliminating the bypass cap when I modified a circuit recently. It now has a current source in the plate circuit, and the plate drives the RC coupled grid of the next stage.
Just for fun, I will measure what the change is with, and then without, the cathode bypass cap. The grid resistor of the next stage is 270k Ohms, and that in parallel with an equal or even higher impedance of the current source, and when that is divided by u +1, is going to be many many times the resistance of the self bias resistor. I am predicting a change far less than 1dB.
I probably do not have enough spur free dynamic range in my FFT to see any difference in the harmonic levels between bypassed and unbypassed self bias resistor in the triode that has a current source in the plate circuit. I also do not expect the gain to change significantly. I shall see.
Where I would expect to see a difference between bypassed and un-bypassed is if I overdrive that stage and draw grid current. The self bias should shift when the bypass cap is present. And that bias shift, no matter the size will have recovery time due to the capacitance and the unbypassed impedance of that node (the total RC time constant).
If there is noise on the filament secondary, and if it has lots of high frequency energy, I do not expect it to introduce significant noise onto the cathode, because the self bias resistor is much lower than the Xc of the small pF filament to cathode and the large resistance of the filament to cathode leakage.
Last edited:
Sorry, I was thinking constant current source in the cathode circuit here, not at all sure I made that clear.
Very interested to hear the experimental results.
So I'm designing an input stage that is direct coupled to the next stage. So there is no coupling capacitor and no grid leak resistor to drive, but there will be a grid stopper. So will a bypass capacitor be required for a constant current sink in the cathode circuit of the input stage?
Right now I'm thinking of just using an anode resistor so it's easier to set a fairly precise operating point for the sake of the DC coupling.
Also, I'm concerned about keeping noise and especially signal from ground from getting on the cathode of the input stage. My ground looks to be messier than my HT in this regard, so I'm wondering if the constant current sink in the cathode circuit will better reduce noise and unwanted signal from the ground. It looks to me like I could have a very large resistance from ground to cathode, and then a much smaller cathode resistance (Ra+ra)/mu+1) to form an effective potential divider.
It looks like a bypass capacitor could only interfere with this perceived advantage in reducing noise and unwanted signal from ground.
Thanks for any light anyone can shed here.
Right now I'm thinking of just using an anode resistor so it's easier to set a fairly precise operating point for the sake of the DC coupling.
Also, I'm concerned about keeping noise and especially signal from ground from getting on the cathode of the input stage. My ground looks to be messier than my HT in this regard, so I'm wondering if the constant current sink in the cathode circuit will better reduce noise and unwanted signal from the ground. It looks to me like I could have a very large resistance from ground to cathode, and then a much smaller cathode resistance (Ra+ra)/mu+1) to form an effective potential divider.
It looks like a bypass capacitor could only interfere with this perceived advantage in reducing noise and unwanted signal from ground.
Thanks for any light anyone can shed here.
Think about what the voltage gain might be for a common-cathode circuit with an unbypassed CCS in the cathode. (Hint: it will be small).
When thinking about "a fairly precise operating point" bear in mind that valve parameters typically vary by 20-30% from sample to sample and maybe by 50% over the useful life of a particular sample.
When thinking about "a fairly precise operating point" bear in mind that valve parameters typically vary by 20-30% from sample to sample and maybe by 50% over the useful life of a particular sample.
Amplifier ground should be something to rely upon, not shied away from. Try getting this right as a layout priority.
I've been looking at an image of what happens to a triode as it ages, and picturing my operating point drifting accordingly. Also thinking in terms of "regions" with an operating characteristic, rather than a specific point. Thanks for helping me focus on that approach.
To me it looks like I'd be putting an infinitely large resistor in the cathode circuit with an unbypassed CCS there. But it's always possible I'm totally misunderstanding something basic.
I'd like to thank everyone for making this forum the resource that it is- I sure appreciate this being here!
To me this situation looks like a potential divider from ground with a small series resistance and a large shunt impedance. The small resistor and the very small capacitance make me think high frequencies are going missing, but where would they be ending up? At the cathode?
I'll be thinking on this one too.
tapehead ted,
1. Bypass Capacitor results for an amplifier of a particular topology and parameters
One section of a 12AU7 was used as a driver for an SE Triode Wired KT66. The amp is intended for near-field playback using a CD player for the music source.
12AU7 circuit: RCA Input jack 50k Ohm potentiometer (volume) 1k Ohm grid stopper Self bias resistor: 470 Ohms Self bias 3.1V Self bias bypass capacitor . . . way over the top . . . 1000uF @ 10V Plate load IXYS 900V rated constant current source set to 6.6mA 12AU7 plate @ 145V Next stage 0.1uF @ 600V driving 270k Ohm 1W grid resistor, and 1k Ohm grid stopper
The results of removing the cathode bypass capacitor: Stage gain without bypass cap 16.57, versus 16.95 with the bypass cap 0.2 dB loss of gain without bypass capacitor
I did not see any change in the harmonic distortion, there was significant signal voltage across the self bias resistor versus the signal to the grid, so the negative feedback ratio (cathode degeneration) was negligible with the bypass cap removed.
There was less than 100uV hum and noise from the amplifier output, so there was no significant hum and noise introduced by the unbypassed 12AU7 cathode. The 470 Ohm self bias resistor was low enough by itself to take care of any filament cathode leakage and capacitance.
As discussed earlier in the thread, this stage works good without the bypass capacitor, because the plate load is the current source in parallel with 270k Ohms (a high impedance load).
2. CCS in the cathode circuit
For a common cathode circuit, the CCS would have to have a bypass capacitor or there would not be any gain. Also, you can not use a CCS in the cathode circuit, and also a CCS in the plate circuit (at the same time).
Perhaps it makes more sense to put the CCS in the plate circuit. The stage will have more gain, and lower distortion than with a plate load resistor. And for that circuit, using a cathode self bias resistor makes sense.
A grounded grid circuit can use a CCS in the cathode, and it should not be bypassed by a cap. A cathode follower circuit can use a CCS in the cathode, and it should not be bypassed by a cap.
1. Bypass Capacitor results for an amplifier of a particular topology and parameters
One section of a 12AU7 was used as a driver for an SE Triode Wired KT66. The amp is intended for near-field playback using a CD player for the music source.
12AU7 circuit: RCA Input jack 50k Ohm potentiometer (volume) 1k Ohm grid stopper Self bias resistor: 470 Ohms Self bias 3.1V Self bias bypass capacitor . . . way over the top . . . 1000uF @ 10V Plate load IXYS 900V rated constant current source set to 6.6mA 12AU7 plate @ 145V Next stage 0.1uF @ 600V driving 270k Ohm 1W grid resistor, and 1k Ohm grid stopper
The results of removing the cathode bypass capacitor: Stage gain without bypass cap 16.57, versus 16.95 with the bypass cap 0.2 dB loss of gain without bypass capacitor
I did not see any change in the harmonic distortion, there was significant signal voltage across the self bias resistor versus the signal to the grid, so the negative feedback ratio (cathode degeneration) was negligible with the bypass cap removed.
There was less than 100uV hum and noise from the amplifier output, so there was no significant hum and noise introduced by the unbypassed 12AU7 cathode. The 470 Ohm self bias resistor was low enough by itself to take care of any filament cathode leakage and capacitance.
As discussed earlier in the thread, this stage works good without the bypass capacitor, because the plate load is the current source in parallel with 270k Ohms (a high impedance load).
2. CCS in the cathode circuit
For a common cathode circuit, the CCS would have to have a bypass capacitor or there would not be any gain. Also, you can not use a CCS in the cathode circuit, and also a CCS in the plate circuit (at the same time).
Perhaps it makes more sense to put the CCS in the plate circuit. The stage will have more gain, and lower distortion than with a plate load resistor. And for that circuit, using a cathode self bias resistor makes sense.
A grounded grid circuit can use a CCS in the cathode, and it should not be bypassed by a cap. A cathode follower circuit can use a CCS in the cathode, and it should not be bypassed by a cap.
There's an example in Valve Amplifiers where there is a two-transistor cascade in the cathode circuit, many million of ohms, and a single transistor CCS in the plate circuit, less than 200K. According to the book, the superior cathode CCS dominates and sets the operating conditions. FWIW!
One of the things I wondered about your experimental circuit is whether or not the increased output impedance without the bypass cap would affect the next stage, either in frequency response or gain. 470 ohms wouldn't make too much of a difference, I expect, and the mu of the 12au7 is small as well, so (ra+(mu)rk) is not so different than just ra in your experiment. Kind of a perfect situation for minimum impact in that regard.
On the other hand 270k is kind of small grid leak, but it seems to be more than plenty by your results. I'm always looking to direct couple one of these CCS plate stages or at least follow it with a cathode follower- I'm reluctant to surrender any of my gazillion ohms! Probably totally unnecessary, but that's the trend of my thought.
Thanks!
1. Current sources are are generally not exactly matched.
To use a current source at the top and the bottom of a triode tube, you would have to have them matched, both for current, temperature effects, dissipation effects, and voltage effects. When B+ changes, or when the tube ages or changes, or when one current source is dissipating more than the other current source, you have a problem.
One current source will be dissipating more power, for example a CCS in the cathode circuit has only a little voltage across it, while the CCS in the plate typically has much more voltage across it.
Of course it is possible to have the two CCS with unequal currents, but then the current has to go somewhere, like the grid; Or one current source will be collapsed with not enough voltage across it.
An ideal concertina might be an exception to the different dissipation of one CCS versus the other CCS.
There is a way around this, it is to parallel a resistor across the current source that always has less current than the other current source. But now we would be changing that associated current source (and resistor) to have a lower resistance, which defeats the reason to use a current source.
2. I used 270k Ohms in the KT66 circuit because it is a conservative value, and the tube will not go into a current run-away condition (unless it becomes very gassy). The amp I have been talking about has 7 mV to 10 mV on the grid. That is the sum of the coupling cap leakage, and the grid current. 7 mV = 26nano Amps, 10mV = 37nano Amps.
If you do tube rolling, a different tube may not be as likely to run-away.
If you ever have had a tube go into run-away current condition (I have), you may find yourself being more conservative in your selection of grid resistors. I have seen Red Plates on tubes that started to run-away.
A 270k Ohm load on a 12AU7 plate is 35 times the 7.7k Ohm plate resistance. You will get low distortion, and the gain will be very very close to u (about 97% of u).
To use a current source at the top and the bottom of a triode tube, you would have to have them matched, both for current, temperature effects, dissipation effects, and voltage effects. When B+ changes, or when the tube ages or changes, or when one current source is dissipating more than the other current source, you have a problem.
One current source will be dissipating more power, for example a CCS in the cathode circuit has only a little voltage across it, while the CCS in the plate typically has much more voltage across it.
Of course it is possible to have the two CCS with unequal currents, but then the current has to go somewhere, like the grid; Or one current source will be collapsed with not enough voltage across it.
An ideal concertina might be an exception to the different dissipation of one CCS versus the other CCS.
There is a way around this, it is to parallel a resistor across the current source that always has less current than the other current source. But now we would be changing that associated current source (and resistor) to have a lower resistance, which defeats the reason to use a current source.
2. I used 270k Ohms in the KT66 circuit because it is a conservative value, and the tube will not go into a current run-away condition (unless it becomes very gassy). The amp I have been talking about has 7 mV to 10 mV on the grid. That is the sum of the coupling cap leakage, and the grid current. 7 mV = 26nano Amps, 10mV = 37nano Amps.
If you do tube rolling, a different tube may not be as likely to run-away.
If you ever have had a tube go into run-away current condition (I have), you may find yourself being more conservative in your selection of grid resistors. I have seen Red Plates on tubes that started to run-away.
A 270k Ohm load on a 12AU7 plate is 35 times the 7.7k Ohm plate resistance. You will get low distortion, and the gain will be very very close to u (about 97% of u).
Last edited:
i can confirm that, gyrator/ccs gives (almost) max gain. so adding capacitor is not helping much.
but with capacitor somehow sound gets worse...
zlab,
You said: "Do you mean CCS on plate & cathode at the same time, are you serious".
No, I am not serious. Read my whole post # 16 and analyze it.
Then refer back to post #15. See what is proposed. I did not even have to research the literature to know that "as stated", it is not going to work. It will have to have a resistor in parallel with one of the CCS's, and that CCS will have to be set to a lower current than the other CCS. That means the CCS with the parallel resistor is no longer a current source. The conclusion is that we now effectively have only 1 CCS. Kind of different than what was stated in post # 15 to be 2 CCS's.
I am sorry, I should have been more clear, so that nobody had to analyze the circuit, and conclude the actual truth. Did I miss something in the literature?
You said: "Do you mean CCS on plate & cathode at the same time, are you serious".
No, I am not serious. Read my whole post # 16 and analyze it.
Then refer back to post #15. See what is proposed. I did not even have to research the literature to know that "as stated", it is not going to work. It will have to have a resistor in parallel with one of the CCS's, and that CCS will have to be set to a lower current than the other CCS. That means the CCS with the parallel resistor is no longer a current source. The conclusion is that we now effectively have only 1 CCS. Kind of different than what was stated in post # 15 to be 2 CCS's.
I am sorry, I should have been more clear, so that nobody had to analyze the circuit, and conclude the actual truth. Did I miss something in the literature?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.