Ive been modifying an amp that uses P-P EL84. Ive made a couple of bias-blocks (CCS sub-circuits) along the lines used by Gingertube in his Baby-Huey amp, one per output valve.
Due to the high +B in this amp, Ive reduced bias on the EL84 to 30mA which equals 10.5W per tube at its current +B.
During static testing, I noticed a loss of power over what I had before, so checked cathode-voltages. These (at point of clip) were 25V. The idle voltage was 12.5 to 13V depending on which tube checked.
Ive placed a 14V 5W Zener diode clamp across the bias-block from tube cathode to deck. This has restored the power at full output to near what it was....
Thoughts....?
Due to the high +B in this amp, Ive reduced bias on the EL84 to 30mA which equals 10.5W per tube at its current +B.
During static testing, I noticed a loss of power over what I had before, so checked cathode-voltages. These (at point of clip) were 25V. The idle voltage was 12.5 to 13V depending on which tube checked.
Ive placed a 14V 5W Zener diode clamp across the bias-block from tube cathode to deck. This has restored the power at full output to near what it was....
Thoughts....?
Last edited:
What is a "bias-block"? If allowing the cathode voltage to rise under signal (like a CCS?) is what it is supposed to do, then adding a zener in parallel will stop it from working and restore normal operation.
Ok, I have just found it - see Baby Huey PP EL84 amplifier - diyAudio. It seems a "bias-block" is just a bias sub-circuit, not something for blocking bias.
The two BJT bias circuit appears to be roughly equivalent to a CCS with a 5k6 resistor in parallel, bypassed by a big cap. No advantage for Class A; disastrous for Class B. The best thing to do is replace it with a bypassed cathode resistor.
The two BJT bias circuit appears to be roughly equivalent to a CCS with a 5k6 resistor in parallel, bypassed by a big cap. No advantage for Class A; disastrous for Class B. The best thing to do is replace it with a bypassed cathode resistor.
yes--Its a CCS set to bias at 30mA.--Ive added that to the first post--sorry for the confusion!
The advantage is maintaining a fixed current through all the output tubes--even if they are not matched, and TBH it does sound better Bass seems deeper and more effortless at low to medium volume.
--The CCA is fine for staying in class A, but as you say, no good for anything demanding higher output than is possible with class A.
My thoughts were to have the benefit of having a 30mA idle (no matter what tube used) and the zener set just above the normal bias voltage prevents cathode-cap charging and overbias the valves....
Suppose I'm looking for the benefits of the both fixed and autobias.
There's nothing wrong with Fixed Bias,--I actually would have preferred it-- just no easily available neg supply to run it from in this set!
Last edited:
Even with perfect const current cathode biasing, the cathode voltage should not double from idle to full "just clipping" output. The extra current the tubes try to draw because their 2nd order term in the grid characteristic is not that great.
I suggest there is something wrong, perhaps a fault in the bias block design, or, more likley perhaps, you have output transformers with the wrong primary impedance for the HT volatge you are using.
However I agree with DF96 - cathode resistor bias works best, and is simpler and more reliable.
Re Jon's question: With Class B you have no option to use fixed bias. With Class A you may have to use a combination of fixed bias and cathode bias if you want to run the tubes at rating limits.
However, for domestic audio, fixed grid bias is NOT recommended. Tubes drift in their characteristics over time, affecting power output and distortion. Cathode resistor bias usefully automatically compensates for tube drift, effectively lengthening the life of the tubes. Const current cathode bias does this too, but at the expense of more distortion and poor transient recovery as the bias circuit overbiases the tubes in trying to keep the current down under load.
I suggest there is something wrong, perhaps a fault in the bias block design, or, more likley perhaps, you have output transformers with the wrong primary impedance for the HT volatge you are using.
However I agree with DF96 - cathode resistor bias works best, and is simpler and more reliable.
Re Jon's question: With Class B you have no option to use fixed bias. With Class A you may have to use a combination of fixed bias and cathode bias if you want to run the tubes at rating limits.
However, for domestic audio, fixed grid bias is NOT recommended. Tubes drift in their characteristics over time, affecting power output and distortion. Cathode resistor bias usefully automatically compensates for tube drift, effectively lengthening the life of the tubes. Const current cathode bias does this too, but at the expense of more distortion and poor transient recovery as the bias circuit overbiases the tubes in trying to keep the current down under load.
Last edited:
Hmm--I did test the CCS by running 'em in series with a voltage source (30V) and a series resistor or three.
Halving the resistance of the resistor resulted in much the same current within a milliamp or so, so don't think the CCS are at fault.
However, the (Chinese) amp as standard does have too high a +B for EL84 being at 350V and Ive no idea if the UL Transformers are exact for the EL84 impedances, as Ive not measured/calculated them, Maybe I should.......
Not exactly sure what you mean by--
............'because their 2nd order term in the grid characteristic is not that great..........'
Halving the resistance of the resistor resulted in much the same current within a milliamp or so, so don't think the CCS are at fault.
However, the (Chinese) amp as standard does have too high a +B for EL84 being at 350V and Ive no idea if the UL Transformers are exact for the EL84 impedances, as Ive not measured/calculated them, Maybe I should.......
Not exactly sure what you mean by--
............'because their 2nd order term in the grid characteristic is not that great..........'
The transfer characteristic of a vacuum tube is inherently non-linear. It can be modelled at any particular anode voltage (triode) or screen voltage (tetrode or pentode) approximately (quite a good approximation if the tube is not operated too near grid cut-off, and is not a remote cut-off radio type) by the quadratic equation:-
Ia = k0 + k1 x Vg + k2 x Vg^2
where Ia is the instantaneous anode current, Vg is the instanteous grid voltage, k0, k1, and k2 are constants determined by the physical dimensions of the tube structure and anode/screen voltage, and ^2 means "the square of".
k1 corresponds to the gm figure given in data books. The k2 x Vg^2 term is known, for obvious reasons, as the second order term. It is normally very small compared to the first order term (k1 X Vg) and gives a slight curvature (non-linearity) to the transfer as plotted on graphs.
The second order term gives rise to a second harmonic in the output, and also increases the average current in the tube, because a sinewave is what you get with sin(2pi.f) and because of the trig identity:-
sin(A)sin(B) = cos(A-B)/2 - cos(A+B)/2, which you learned in high school, which simplifies to:-
sin(A)^2 = cos(0)/2 - cos(2A)/2
cos(0)/2 is the added DC current and cos(2A)/2 the second harmonic
Real tubes generate tiny amounts of higher harmonics, though generally the 3rd is very small compared to the second and 4th and anything higher are entirely negligible. You can account for the third harmonic by adding a k3 x Vg^3 term (ie a third order term), but most don't bother. More accurate formulae can be used in computer simulations that account for the near-cutoff condition, and varying anode voltage, as well.
If you know k1, you can calculate gain. If you know k1 and k2 you can calculate fairly accurately the distortion, by use of 1st year high school trig. It happens that all sharp cutoff tubes intended for audio have k1 and k2 terms in about the same proportions, so you never see a k2 term listed in databooks. You can determine it from grid curve graphs though.
Last edited:
We are not suggesting that the CCS are faulty. On the contrary, we are suggesting that using a CCS here is a mistake. The better the CCS, the worse the mistake!Alastair E said:
Note that an ideal Class A stage can provide peak AC output equal to half the quiescent dissipation. A real Class A stage is more likely to struggle to go much beyond 30-40% of dissipation, so 10W idle means 3-4W per valve. A CCS-biased output will be worse, as the second-order distortion (see Keit's description above) will cause a bias shift too. Maybe 2-3W per valve?
Best output bias is fixed grid voltage. Second best is cathode resistor. Third best is CCS - but this may be necessary if a toroidal OPT is used. If you do use a CCS then you must set the quiescent bias as 'hot' as possible, so it can cope with some bias shift to cooler operation when a signal is present.
Facing to the same situation, I've used the same solution and was happy 🙂
Note however than this phenomenon occurs while testing at continuous hi power, much less notiçable on music program.
Sure, the effect of "overbias" is less pronounced when using a resistor rather than a CCS.
It exists even with fixed bias if the bias source (including the grid leak resistor) has a too high impedance, in this case it is the link capacitor who strore the rectified voltage produced by grid current at high level.
At least, a CCS in each cathode eliminate the need of manually matching plate current of the 8 output tubes (4 x EL84 per channel).
(And matched tubes do not stay matched for ever 🙁)
Nothing is perfect 😛
Yves.
Thanks Yves, Glad someone else has done similar to good effect.
I'm quite happy with the way the amp runs at the moment. Sounds much better than it did in its original configuration, and the CCS bias-block definitely does improve the sound.
I need to do some tidying up of the internals, some of my modifications are point-to-point on the PCB etc, looks rather nasty. I'll possibly do a post/thread on whats been done to it....
--Not much is left of the original design TBH!
It was pretty obvious it had been made down to a price, the omission of any grid-stoppers on all the valves was one thing that gave me some fun chasing an oscillation on both channels after I had re-designed the gain-stage and replaced the Transistor phase-inverter with a 6H16B valve below decks as I didn't want to make holes in the thing.....
Such is the inscrutable delights of El-cheapo Chinese Yaqin EL84 P-P amps! At least it Looks nice straight out of the box, after much messing about, now this one sounds nice too!
I'm quite happy with the way the amp runs at the moment. Sounds much better than it did in its original configuration, and the CCS bias-block definitely does improve the sound.
I need to do some tidying up of the internals, some of my modifications are point-to-point on the PCB etc, looks rather nasty. I'll possibly do a post/thread on whats been done to it....
--Not much is left of the original design TBH!
It was pretty obvious it had been made down to a price, the omission of any grid-stoppers on all the valves was one thing that gave me some fun chasing an oscillation on both channels after I had re-designed the gain-stage and replaced the Transistor phase-inverter with a 6H16B valve below decks as I didn't want to make holes in the thing.....
Such is the inscrutable delights of El-cheapo Chinese Yaqin EL84 P-P amps! At least it Looks nice straight out of the box, after much messing about, now this one sounds nice too!
No, I did suggest the CCS bias block might be faulty. The OP reported that the cathode voltage doubled under load, just at the point of clipping. Even with the stupid unnecessary CCS, it shouldn't vary that much! However the OP sensibly tested his bias blocks so he probably has an incorrect output transformer or some other problem.
I know you and I agree. I'd just put it this way:-
The lowest distortion comes with fixed grid bias. Cathode resistor bias is not quite as good, but with sensible design its not much worse. CCS cathode biasing is worst of all.
If you use a toriodal output transformer, solve the balance problem with a bias balance pot. And/or select a transformer with a generous core size. Don't use a common cathode resistor, as this promotes differential tube ageing. Each tube should have its own bias resistor. If the output tubes are the same brand and date code, and are purchased new or trustworthy NOS, they are then unlikley to significantly drift apart in characteristics, so if the bias balance pot is adjusted when the tubes are installed, the post is unlikely to need attention until the tubes are worn out and are replaced.
TBH--I would have Preferred to use a fixed-bias system. Unfortunately, there was no straightforward way to implement it in this amp without adding a transformer for the bias rail, and the associated rec, smoothing and even maybe regulation--which would have been difficult--even for a small transformer as the chassis is only about 1.5" deep--and already pretty full! To add it above decks would have made it look horrible.
This Yaqin EL84 amp is a pretty small footprint--About the size of the Philips CD104 its currently sitting on top of....
As to which is to be preferred in the bias regime, Personal taste and requirements/constraints of the design have more to do with it than out and out performance on paper sometimes--Life is a series of compromises.......
This Yaqin EL84 amp is a pretty small footprint--About the size of the Philips CD104 its currently sitting on top of....
As to which is to be preferred in the bias regime, Personal taste and requirements/constraints of the design have more to do with it than out and out performance on paper sometimes--Life is a series of compromises.......
Exactly! And why would it move into Class B on peaks? Because the output transformer primary impedance is too low for the HT voltage in use!
Hmm--Time to test the Output Transformer's Impedance then, because gawd-knows what it actually is seeing there's been some pretty basic errors in the thing's design....!😀
Any other tests I can do to the transformers while I'm at it....?
--I may get around to this later--Let you know what it is.....
Last edited:
That (as described by the OP) was exactly my experience with the Baby Huey. When testing with continous sine wave at high power the cathode voltages went to about double the "normal" approx +12V.
As Yves says this is not a problem with normal music source but the CCS's do however compromise overload recovery.
This was why I did the fixed bias variant of the Baby Huey. Going fixed bias had other implications, the Rg1 values on the output tubes had to be reduced which then required buffers. I used MOSFET source followers with CCS loads direct coupled to the output tube grids (just a grid stop in line) between the phase splitter and the output tubes and applied the bias to the source follower gates.
Having said that I dug out the prototype Baby Huey with the CCS bias blocks just last week to lend to a fellow EE in the day job. His report was that it was "stunning" and had trouble believing me when I told him the fixed biased version was even better. The caveat on that is that this was his first experience with a tube HiFi amp although he has built and modified a lot of tube Git. Amps.
The Zener clamp is a good solution to assist overload recovery.
The output trannies should be anywhere between 6K and 10K Raa, I've used both 8K and 10K on the many BH's I've built.
Other comments on "stuff" above".
The bypass caps on the CCS's allow Class AB operation. Without the bypass caps you are indeed limited to Class A operation.
EL84 are quite happy biased at 12 watts idle.
Cheers,
Ian
As Yves says this is not a problem with normal music source but the CCS's do however compromise overload recovery.
This was why I did the fixed bias variant of the Baby Huey. Going fixed bias had other implications, the Rg1 values on the output tubes had to be reduced which then required buffers. I used MOSFET source followers with CCS loads direct coupled to the output tube grids (just a grid stop in line) between the phase splitter and the output tubes and applied the bias to the source follower gates.
Having said that I dug out the prototype Baby Huey with the CCS bias blocks just last week to lend to a fellow EE in the day job. His report was that it was "stunning" and had trouble believing me when I told him the fixed biased version was even better. The caveat on that is that this was his first experience with a tube HiFi amp although he has built and modified a lot of tube Git. Amps.
The Zener clamp is a good solution to assist overload recovery.
The output trannies should be anywhere between 6K and 10K Raa, I've used both 8K and 10K on the many BH's I've built.
Other comments on "stuff" above".
The bypass caps on the CCS's allow Class AB operation. Without the bypass caps you are indeed limited to Class A operation.
EL84 are quite happy biased at 12 watts idle.
Cheers,
Ian
It is a problem because teh CCS raises the cathode voltage with signal, so the tubes become overbiased and will distort more than they would on fixed or cathode resistor bias.
As DF96 said, if you are going to put in a zener, then it is pointless having a cathode CCS.
The Baby Huey design uses EL84/6BQ5 with a HT between 300 and 330V. If you do have 330V HT, the output trans primary impedance, which should be approx proportional to teh square of the voltage, should be 10 kohm anode to anode. If you use a 6 kohm transformer, no wonder the amp goes into Class B, pulls up its cathodes, and you'll get the distortion and very poor transient recovery that that entails.
Even a 6K tranie on 300 V will cause Class B operation on strong signals.
If you use one common CCS for both cathodes, the amp will work without a cathode bypass, though not especially well.
If you use separate CCS' for each cathode, there MUST be bypass capacitors, or there will be little or no output - the CCS's will prevent any signal current through the tubes.
It has nothing to do with whether you want Class A or Class AB operation.
In a triode output stage, you can be fairly sloppy about output transformer impedances. In a pentode amplifier, you need to be more carefull, but you can still get away with some variation in impedance.
That is, unless you do stupid things like sink cathode currents into a CCS.
That is, unless you do stupid things like sink cathode currents into a CCS.