I am building an EL84 pp am from a circuit in which the original used a FET for current regulation with a 270 ohm resistor in the signal path. I had thought of using an LM 317 instead but to achieve the +/-30 ma that I require results in 47 ohm series resistorinstead. I am no expert but that would mean that the voltage at the anode would probably exceed the 300V limit. As I am using parts that I have would an additional 220 on the valve side bring the circuit back into 'balance'?
It might help to post the circuit. That remark about " the +/- 30ma that I require " just does not make sense a CCS just suplies one value of current.
Thank you for your answers. The circuit can be found at http://www.geocities.ws/tjacodesign/valveamp/valveamp.html.
The LM 317 is rated at over 1 amp so a current of 30 ma should be OK?
The original circuit uses a CCS supplying 20 ma but a 47 ohm resistor with an LM 317 is the nearest value that I have.
Alll help appreciated
The LM 317 is rated at over 1 amp so a current of 30 ma should be OK?
The original circuit uses a CCS supplying 20 ma but a 47 ohm resistor with an LM 317 is the nearest value that I have.
Alll help appreciated
The LM317 will be well within ratings - it will pass a current of 25.5 mA and drop about 9 volts.
Using a const current source to bias output tubes has a minor advantage in that it improves amplifier performance in teh face of overdrive. In cathode resistor biased output stage, overdrive can charge up the cathode bypass capacitor, so that when the overdrive stops the tube is overbiased and distortion is increased. Cathode current tries to increase with drive anyway, so with reistor bias, the bais is signal dependednt and cannot remain optimum under signal drive.
However, the life of an integrated circuit regulator will be very dependent on the individual output tube, and circuit layout. It is quite normal for output tubes to internally arc over once in a while. This casues a cathode-anode or cathode-screen short circuit. It usually lasts less than a microsecond and causes only a small click in the speakers which may not be noticed. When music was sourced from vinyl, nobody cared, as you got clicks from dust in the groove anyway.
But those sub-microsecond tube shorts can destroy IC regulators. You can aleviate this problam by carefull layout and a ceramic capacitor across the regulator, in parallel with the bypass electrolytic. Don't rely on the main bypass cap alone.
If you look at output tubes made in the 1930's and 1940's, there was often a gray coating ("dag") on the inside of the glass. Later, a glass process was developed in the USA and elsewhere that renders the glass inside surface very very slightly conductive.
The dag and the slightly conductive glass cuts down the incidence of internal sub-microsecond shorts. Power tubes have high anode current, and there is always some electrons that go to hit the glass instead on internal metal. On standard glass, as it is a virtually perfect insulator, there slowly (over hours or days) builds up a negative charge. When the charge gets high enough an internal breakdown occurs. This is non-destructive because the energy in minute, but it causes a click in the speakers.
The dag or slightly conductive glass prevents any charge build-up.
Why am I telling you all this? Because while NOS tubes from reputable American, English, and European manaufacturers have the special slightly conductive glass (or the dag), it appears Chinese tube do not, and Russian consumer grade tubes may not.
So, use an IC regulator with Chinese tubes, and it may not last long.
I personally, in a push-pull amplifier, prefer to use normal cathode resistor bias, one resistor per tube. It's simpler, cheaper, and more reliable. The distortion issue I mentioned above can be largely avoided by having a bypass capacitor from cathode to cathode instead of two capacitors each grounded. That also improves push-pull symmetry, also reducing distortion. The cap has no DC bias and that mandates a non-electrolytic, but that's a good thing too. Or you can have two electros in series, negative ends together and grounded by a large value resistor.
Using a const current source to bias output tubes has a minor advantage in that it improves amplifier performance in teh face of overdrive. In cathode resistor biased output stage, overdrive can charge up the cathode bypass capacitor, so that when the overdrive stops the tube is overbiased and distortion is increased. Cathode current tries to increase with drive anyway, so with reistor bias, the bais is signal dependednt and cannot remain optimum under signal drive.
However, the life of an integrated circuit regulator will be very dependent on the individual output tube, and circuit layout. It is quite normal for output tubes to internally arc over once in a while. This casues a cathode-anode or cathode-screen short circuit. It usually lasts less than a microsecond and causes only a small click in the speakers which may not be noticed. When music was sourced from vinyl, nobody cared, as you got clicks from dust in the groove anyway.
But those sub-microsecond tube shorts can destroy IC regulators. You can aleviate this problam by carefull layout and a ceramic capacitor across the regulator, in parallel with the bypass electrolytic. Don't rely on the main bypass cap alone.
If you look at output tubes made in the 1930's and 1940's, there was often a gray coating ("dag") on the inside of the glass. Later, a glass process was developed in the USA and elsewhere that renders the glass inside surface very very slightly conductive.
The dag and the slightly conductive glass cuts down the incidence of internal sub-microsecond shorts. Power tubes have high anode current, and there is always some electrons that go to hit the glass instead on internal metal. On standard glass, as it is a virtually perfect insulator, there slowly (over hours or days) builds up a negative charge. When the charge gets high enough an internal breakdown occurs. This is non-destructive because the energy in minute, but it causes a click in the speakers.
The dag or slightly conductive glass prevents any charge build-up.
Why am I telling you all this? Because while NOS tubes from reputable American, English, and European manaufacturers have the special slightly conductive glass (or the dag), it appears Chinese tube do not, and Russian consumer grade tubes may not.
So, use an IC regulator with Chinese tubes, and it may not last long.
I personally, in a push-pull amplifier, prefer to use normal cathode resistor bias, one resistor per tube. It's simpler, cheaper, and more reliable. The distortion issue I mentioned above can be largely avoided by having a bypass capacitor from cathode to cathode instead of two capacitors each grounded. That also improves push-pull symmetry, also reducing distortion. The cap has no DC bias and that mandates a non-electrolytic, but that's a good thing too. Or you can have two electros in series, negative ends together and grounded by a large value resistor.
Last edited:
That 47 ohm resistor with a Lm317t should come out ~26ma instead of 20ma but that would probably be ok as most people run that tube with more than 20ma. And at 26ma per tube instead of 20ma your anode voltage will be a few volts less but I wouldn't worry about it.
Thank you for your comprehensive answers - Regret I am no expert but I was worried that the ht with a simple bridge rectifier and a 300 volt transformer would exceed the 300v limit on the el 84s. I have four telefunken tubes (new) - from the 50s? so I will follow the advice on coupling a ceramic cap in parallel with the 100mf cap.
Best regards to both....
Best regards to both....
What you say would be true for constant quiescent current bias, but not for the much more common constant average current bias. The latter is worse than a resistor, not better. Fortunately unless overdriven the effect is small so many people do not realise that they have spent money and time on a more complex but inferior system to a humble resistor.Keit said:
So do I - or mixed bias: fixed voltage plus cathode resistor.
Er, do you meant a transformer with a 300V spec'd secondary? Using that, a silicon bridge rectifier, and a capacitor input filter, the HT voltage will be approx 420 V.
1950's Telefunken's (new) are good provided not excessively gassy.
Please explain.
Yep. That's true for various things. True also that folk spend money on things that cannot make any audible difference.
Yeah - if you need to run tubes hard that may be more or less essential. For music reproduction, it's better to use bigger tubes or parallel tubes though.
Interesting reading. The transformer in the original circuit (see above) has a published secondary of 290 volts, which was why I questioned the 270 ohm resistor in the CCS - concerned about the resulting anode voltage.
If you run a valve with fixed voltage bias (via the grid or the cathode) then its quiescent current will be fixed, but even if in Class A its average current will increase when a signal is present due to second-order distortion. This is normal behaviour and no problem - in PP the second harmonic will cancel in the OPT but the DC effect is still present. 10% second (cancelling to maybe 1% in the output) has 10% current shift too.Keit said:
Run the same valve with a cathode resistor and the same effect occurs, but now the voltage rises to partially compensate. Average current rises with signal but quiescent current falls. If the resistor is bypassed then the reduction in quiescent current continues for a while after the signal has gone. This effect was implied in the classic Mullard book on audio amps, where for some circuits it suggested different bias points depending on whether you were just going to listen to music or do full-signal sine wave testing. If you want correct bias with full signal then you have run the valves hotter than 'correct' for no signal, so there is some wiggle room for the bias shift to cool them with signal without creating too much crossover distortion.
A naive CCS in the cathode enforces constant average current, so the quiescent current has to reduce under signal even more than for resistor bias. This is because the difference between average current and quiescent current is set by the valve and the signal (plus anode load etc.) so if you fix one the other has to move.
About the only reason I can think of where naive CCS bias would be appropriate for a PP output is when a toroidal OPT is used so DC currents have to balance quite closely. Even then a zero-crossing bias servo would be better, but few people use these - I guess because few realise why they are needed.
For most people cathode resistor bias is best. Nothing to adjust, and it automatically copes with valve ageing and sample variation. There is no reason to set an output current to some precise value, as valves are not that fussy. However, people like to feel they are doing something more complicated than naive boring engineers do, and surely something more complicated must sound better?
Many thanks for your comprehensive answer. I guess that I decided to use a CCS because the author used one. Do you have any comment on the mains transformer ht of 290v? This appears to be rather high!
The original circuit uses a toroidal OPT, so that might explain the CCS bias. 20mA seems rather low for Class A, and CCS bias fails completely for Class B so I am puzzled*.
290V RMS (leading to around 400V DC) seems rather high for an EL84.
Don't assume that a published circuit is well-designed. Magazine editors don't have the time to check everything, and some may even lack the knowledge to do so reliably.
* When I am puzzled by a circuit it either means the other designer is much smarter than me, or I am somewhat smarter than him.
290V RMS (leading to around 400V DC) seems rather high for an EL84.
Don't assume that a published circuit is well-designed. Magazine editors don't have the time to check everything, and some may even lack the knowledge to do so reliably.
* When I am puzzled by a circuit it either means the other designer is much smarter than me, or I am somewhat smarter than him.
Ahah! That is what you were on about. Your terminology was a bit obscure in your first post, and in your use of the word "quiescent". However I can't think of a way just now to put it any better.
More importantly, does this downward shift in "quiescent" current (equivalent to an increase in bias voltage) result in distortion? Well, as far as the tube is concerned, it's an increase in bias the same as the increase in current in a bias resistor, except somewhat more so.
So, while transient handling is improved with a CCS, steady state harmonic distortion can be expected to be worse than with cathode resistor bias.
You are a sharp thinker, DF96. I never thought of it that way until now.
I most certainly agree. But there are two ways to do cathode resistor bias in push-pull circuits:
a) A common resistor for both tubes;
b) Separate resistors, one for each tube.
(b) is best because it prevents a strong tube from overbiasing a weak tube. Many enginners however never realised that that didn't have to mean two cathode bypass capacitors. If you include a single cpacitor between the cathodes, dymanic balance is improved as well as better overdrive recovery.
QFT. Thanks for the great post!
I shall never again use an IC regulator for cathode CCS.
I have a couple of Cheap Chinese Tubes that first took out the 78xx regulator in the cathode CCS circuit, then the overcurrent melted them as well.
My circuit looked exactly the same, with two output pairs in parallel on a toroidal OPT. It lasted about six hours.
Dead right.
Even the respected English magazine Wireless World in its heyday when the editor and his staff were qualified engineers well able to assess things, they still published dud circuits.
I've had articles published myself in WW/EW and elsewhere. Naturally I reckon I wrote good stuff. But I do know from that editorial contact that magazine editors find it hard to get enough material to publish. They write begging letters to known good authors. And even a bad circuit may draw some interest and stimulate some minds.
Having said that, some editors insist on evidence (photos, test data) of a succesful prototype, and will arrange for another to be built from the draft article independently of the author. That's good, though their testing is usually just a functional test without instruments.
I assume you meant that the tubes took out the regulator chip, which shorted, and teh resulting zero bias led to the tubes melting. (ie the regulator(s) didn't melt).
Yes, the tubes did melt, not the regulators. There were small cracks in the glass at the bottom of the envelope. I assume that was the chain of events.
The 78xx regulators have a nasty habit of failing short-circuit, which I did not take into account when slapping the thing together, nor did I account for the consequences.
The 78xx regulators have a nasty habit of failing short-circuit, which I did not take into account when slapping the thing together, nor did I account for the consequences.
Gentlemen, a fascinating discussion! As an ex language teacher who relies on what I have read, building myself for fun with many mistakes along the way, I am reliant on such information! I will replace the CCSs with resistors, even if it means disconnecting the boards and removing them...
A thought (getting old and slow) - transformer manufacturers quote loaded voltages so I guess that the 290v would be the loaded value and therefore quite safe and appropriate for el84s?
A thought (getting old and slow) - transformer manufacturers quote loaded voltages so I guess that the 290v would be the loaded value and therefore quite safe and appropriate for el84s?
- Status
- Not open for further replies.