I am currently studying Cathode Bias with a regulator as CCSink. (see picture).
Now this has been discussed before, but I find different recommendations for the cathode bypass cap in this setup.
e.g. the RH84 had just 47µF in that spot, a later revision 100µF.
(I know there is a lot of controversy about the RH but the low value of the bypass was never mentioned)
Mr. Brosky assumed 10.000µF in a similar setup, though PP.
In my case the current would be 50mA at 10V, which is equivalent to some 200ohms.
If it were a plain resistor this would call for something like 1000µF.
So, how does the RH get away with it's 100µF and what is the proper number,
100, 1000, 10000 ?
Now this has been discussed before, but I find different recommendations for the cathode bypass cap in this setup.
e.g. the RH84 had just 47µF in that spot, a later revision 100µF.
(I know there is a lot of controversy about the RH but the low value of the bypass was never mentioned)
Mr. Brosky assumed 10.000µF in a similar setup, though PP.
In my case the current would be 50mA at 10V, which is equivalent to some 200ohms.
If it were a plain resistor this would call for something like 1000µF.
So, how does the RH get away with it's 100µF and what is the proper number,
100, 1000, 10000 ?
Attachments
No, you may be confusing 'DC resistance' with slope resistance. There is no 200 ohms in the circuit.payloadde said:
The value of the cathode bypass cap is set by the impedance seen at the cathode and the desired LF rolloff for that stage. The impedance seen at the cathode is roughly 1/gm + RL/mu, where RL is the load seen by the anode. Mu can be regarded as infinite for a pentode. This assumes Class A.
For Class B you may need a bigger cap as it has to maintain a low impedance at music envelope frequencies. However, a single CCS is completely wrong for biasing Class B - it is not entirely right for Class A either.
I suggest you take a look at the CCS employed in the finals of the Baby Huey amp. Has been used by hundreds of people and has proven itself. IIRC bypass caps are a trade off between low frequency response where larger values offer lower distortion at low frequencies, and recovery time after overload conditions. Large caps get charged beyond the bias point, bossing the output stage more to class B with large signals. After overload large caps retain the colder bias point for a longer time.
Class A makes things a bit simpler I guess, as there is no situation in which the tubes go into cut off and thus there is no mechanism for the cathodes of the finals to creep up in voltage.
Class A makes things a bit simpler I guess, as there is no situation in which the tubes go into cut off and thus there is no mechanism for the cathodes of the finals to creep up in voltage.
I can only think of one reason for using CCS bias in an output stage: it is push-pull with a toroidal OPT so needs precise matching of quiescent current. Everywhere else a resistor would be better, as it gives less bias shift with signal. Valves do not need a precise current bias setting, and CCS delivers precise quiescent bias at the expense of more variable 'with signal' bias than a resistor so you are paying more for less. In those situations where a resistor is not good enough then fixed bias should be used.
Exactly.
You have to understand why you might want to use a CCS in an output stage and the case of the toroidal is about the only valid reason I can think of.
The justification given by Kitic is that it allows swapping of output valves (different valve types not just brands) without changing the cathode resistor value. Even attempting this however is going to be a compromise since changing valves type would undoubtedly require changing transformer in most cases.
So if you intend to change the valve type on a regular basis, then go ahead and use a CCS, otherwise a resistor is by far the better option.
Shoog
I switched to fixed bias in my higher powered single ended too. It is much more flexible if you want to try different power tubes, kt66 or kt88 or kt90 , etc. Downside is the extra parts for another power supply and the extra circuitry to make it adjustable.
The input impedance of the CCS is near infinite then, and we ignore the RL/mu part.
O.K., so let's assume, gm is 10mA/V, then 1/gm = 100ohms which is even lower than 200.
To get a LF rolloff < 1 Hz this would require something like 2000µF then.
and 100µF (15Hz) is most likely not enough ...
Basically it seems that Kitic simply copied the typical values used in old SE designs. This was when 47uf was a big cap. For me it shows a fundamental lack of understanding of the basics. There is however a possible explanation for his approach, the RH designs are intended for salvaged console iron which tend to be small and will have a tendency to saturate if presented with a sufficiently low frequency signal. Choice of a 47-100uf cap would tend to keep this to a minimum.
Almost all historic SE designs, and many modern designs to boot, have chronically undersized cathode bypass caps. When a 50V 4700uf cap costs peanuts why would you choose anything less if thats what the calculation points to ?
Shoog
Almost all historic SE designs, and many modern designs to boot, have chronically undersized cathode bypass caps. When a 50V 4700uf cap costs peanuts why would you choose anything less if thats what the calculation points to ?
Shoog
Last edited:
I use CCS loads because I have toroidal power transformers as outputs and they require a balanced load. I use this like Broskie documented. It works well but I ditched the zeners. Apparently as shown they destruct when their voltage is exceeded. Probably something to do with 50V+ being shunted. As specified by John Broskie they short and fail shorted. Apparently 50V@10,000mf exceeds 5W... Who knew? 
FWIW I get great results with these toroid power transformers as output. You just need to know the math to pick the proper coil... And thanks to those of you on the forum who've given me guidance etc. As a reasult I have a sub $1000CAD ~30W design in class a that is cheap to implement. If you can build this yourself you end up with a 30W class amp that costs under 1000$ and sounds as good if not better than a mac...
FWIW I get great results with these toroid power transformers as output. You just need to know the math to pick the proper coil... And thanks to those of you on the forum who've given me guidance etc. As a reasult I have a sub $1000CAD ~30W design in class a that is cheap to implement. If you can build this yourself you end up with a 30W class amp that costs under 1000$ and sounds as good if not better than a mac...
Last edited:
So with a CCS attached to a class A, as the tube ages the bias voltage would be automatically adjusted to keep the standing current the same as before. Is this the right approach though? Is the ideal standing current for best linear operation always the same as the cathode emission declines with age?
A single shared and unbypassed CCS can be used in the cathodes of a push pull output pair so long as amp is strictly Class A ONLY. With signal applied, as one side drops current the other side increases current but total current remains constant.
Allen Wright did this with his DP300B push pull 300B Amp (Class A Triode PP)
For Class AB push pull you need a CCS for each tube (to set idle current) AND a bypass cap to handle the AC signal. That is what I did in the Baby Huey. With EL84 or 6V6 then 470uF bypass caps were good but 4700uF were better.
For the best Baby Huey version I went to fixed bias - why? because the CCS biased output stage has the worst overload recovery of any bias scheme I've ever tried. If you tend to push your Class AB amp hard then stay away from CCS bias. (It is no good for a Git Amp).
For Single Ended (like the RH designs) you need a bypass cap to pass the AC signal.
Cheers,
Ian
Allen Wright did this with his DP300B push pull 300B Amp (Class A Triode PP)
For Class AB push pull you need a CCS for each tube (to set idle current) AND a bypass cap to handle the AC signal. That is what I did in the Baby Huey. With EL84 or 6V6 then 470uF bypass caps were good but 4700uF were better.
For the best Baby Huey version I went to fixed bias - why? because the CCS biased output stage has the worst overload recovery of any bias scheme I've ever tried. If you tend to push your Class AB amp hard then stay away from CCS bias. (It is no good for a Git Amp).
For Single Ended (like the RH designs) you need a bypass cap to pass the AC signal.
Cheers,
Ian
all true, GT, thanks for summarizing it concisely. I wanna add these not that obvious points of my experience:
- for a common CCS/PP case the tubes MUST be matched, otherwise one will run out quick;
- for same common CCS/PP it is strictly NO GO to remove (accidentally or not) one of output stage tubes, bc another one will take all double current and might pass away in a matter of few minutes;
- for the separate CCS/PP we basically achieve 3 things: a) the non-matched tubes can be used, also applies to the aging process, b) the OPT is always well-balanced, c) we get the tubes running a bit cooler, I'd found the optimal is -25% of the class A quiescent current;
- for the separate CCS/PP + Zener altogether with above 3 we can run the output tubes even more cooler (up to -(35-40)% of class A idle), to almost like AB2 mode without actually grid-driving them, proven. Literally, I breadboarded two R+C bypassing Zeners on a switch and when connected the swing on OPT's primary increases up to shaggedelic 30%, not bad !
BUT, but, but. .. sonically an old-n-good RCZ just beats the CCS+C+Z. .. though that's up to you;
- a-and for a CCS/SE the CCS+C+Z also allows to run the output tube a bit cooler, but that has to be figured experimentally in each particular build, bc here everything adds into an equation with multiple X-es, so just trying is the easiest way. I would estimate it -10% of the class A (only applicable here) quiescent current, or in the best case scenario up to -15%.
- for a common CCS/PP case the tubes MUST be matched, otherwise one will run out quick;
- for same common CCS/PP it is strictly NO GO to remove (accidentally or not) one of output stage tubes, bc another one will take all double current and might pass away in a matter of few minutes;
- for the separate CCS/PP we basically achieve 3 things: a) the non-matched tubes can be used, also applies to the aging process, b) the OPT is always well-balanced, c) we get the tubes running a bit cooler, I'd found the optimal is -25% of the class A quiescent current;
- for the separate CCS/PP + Zener altogether with above 3 we can run the output tubes even more cooler (up to -(35-40)% of class A idle), to almost like AB2 mode without actually grid-driving them, proven. Literally, I breadboarded two R+C bypassing Zeners on a switch and when connected the swing on OPT's primary increases up to shaggedelic 30%, not bad !
BUT, but, but. .. sonically an old-n-good RCZ just beats the CCS+C+Z. .. though that's up to you;
- a-and for a CCS/SE the CCS+C+Z also allows to run the output tube a bit cooler, but that has to be figured experimentally in each particular build, bc here everything adds into an equation with multiple X-es, so just trying is the easiest way. I would estimate it -10% of the class A (only applicable here) quiescent current, or in the best case scenario up to -15%.
+1
This may be is the reason because my SE amplier works better with a old resistor and not a LM317 CCS.
Thank´s
To some extent it depends on whether the output stage is within a feedback loop. If it is then sometimes you need a higher rolloff in the output to maintain LF loop stability. However, in general a cathode CCS will need a bigger bypass cap than a cathode resistor because the resistor will maintain some LF gain down to DC while the CCS will not. Bypassed resistor gives you an LF shelf, but bypassed CCS gives a high pass.payloadde said:
A good question. Does anyone know the answer?Circlotron said:
Only true if there is no second-order distortion. A typical PP output could easily produce 10% or more 2nd on each side on signal peaks. This cancels in the OPT so you don't hear it. However, you still get a 10% shift in DC too which causes the CCS to reduce quiescent current. The result can be introduction of crossover distortion, but only on sustained loud passages. To avoid this the valves would need to be set at a 'hotter' bias than would be needed for resistor bias. A fat bypass capacitor just delays the problem.gingertube said:
CCS bias is inappropriate for Class AB. You can add a big cap as a sticking plaster, but basically the circuit does not work.
See? I said it doesn't work. The basic problem is that a CCS fixes the wrong current. It fixes average current, when what you actually want is fixed quiescent current. These two currents are only equal for a Class A output with no second-order distortion, or no signal. So, it only works under conditions where you don't need it.
So, to sum up, don't use a CCS for Class A unless it is PP with a toroidal OPT. Everywhere else CCS is either unnecessary or harmful. It is particularly harmful for Class AB. Adding a big bypass cap is merely a sticking plaster for a design based on a misunderstanding of how the output stage works.
I'm trying to understand how does this confront with the general idea that a tail CCS enhances AC balance in a LTP. Quiescent DC is AC isn't it? Maybe it is only true for input stages and phase inverters where 2nd order distortion can be low?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.