I have recently bought some quality SE OPTs that are 600R primary impedance. The thinking was that I could use 4 x EL86 in parallel with them since they are rated for 350mA current. EL86 is a high current low voltage EL84 and usually has 2.5K as plate load, so 4 in parallel matches well with the OPTs for impedance and current.
There are some good examples of Hi Fi amplifiers with paralleled tubes. Here is one from Patrick Turner with 3 x EL34 at the output tubes: Patrick Turner 100W ULAB1.
From my limited experience with this hobby, it seems that tubes parallel very well, as long as they are reasonably well matched, but their only foible is the Miller capacitance, which is capacitance in parallel so is the sum of each tube.
From the data sheet:
Pentodes solve the capacitance problem with additional grids, so maybe it would not be an issue with the 4 tubes in parallel? There is a thread here ...
DiyAudio: Pentode Miller Capacitance, and which spec value to use?
... and from that it looks like it should not be a problem with Miller capacitance, even when it is compounded 4 times. However, if I want to be able to switch to Triode output then it is going to complicate the driver.
I know I have a long way to go before I realise this amplifier, but I would like to hear some thoughts on the sort of driver circuit I will need to satisfy the needs of 4 tubes in parallel, or if there are any special considerations to take on board. (Maybe it is a stupid idea!). It looks like oscillations can become more likely, for instance, and protection circuitry is more important. The output tubes will be self biased, so that is one simplification.
This thread from @Diabolical Artificer was an interesting read: DiyAudio: 120w PP UL monoblocks , with a quad of EL34s.
There are some good examples of Hi Fi amplifiers with paralleled tubes. Here is one from Patrick Turner with 3 x EL34 at the output tubes: Patrick Turner 100W ULAB1.
From my limited experience with this hobby, it seems that tubes parallel very well, as long as they are reasonably well matched, but their only foible is the Miller capacitance, which is capacitance in parallel so is the sum of each tube.
From the data sheet:
Pentodes solve the capacitance problem with additional grids, so maybe it would not be an issue with the 4 tubes in parallel? There is a thread here ...
DiyAudio: Pentode Miller Capacitance, and which spec value to use?
... and from that it looks like it should not be a problem with Miller capacitance, even when it is compounded 4 times. However, if I want to be able to switch to Triode output then it is going to complicate the driver.
I know I have a long way to go before I realise this amplifier, but I would like to hear some thoughts on the sort of driver circuit I will need to satisfy the needs of 4 tubes in parallel, or if there are any special considerations to take on board. (Maybe it is a stupid idea!). It looks like oscillations can become more likely, for instance, and protection circuitry is more important. The output tubes will be self biased, so that is one simplification.
This thread from @Diabolical Artificer was an interesting read: DiyAudio: 120w PP UL monoblocks , with a quad of EL34s.
The OPTs are designed for 6C33, but I have a shed load of PL84 tubes, so thought it could be an interesting experiment.
As mentioned, it could well be a stupid idea!
As mentioned, it could well be a stupid idea!
A good mosfet follower should be able to drive pretty much anything.
I've built 3 amps with paralleled OP valves, the first similar to Pat turners schematic using 6 EL34's the 2nd two 4 807's. I haven't had any issues with Miller C, the last one I built was flat to 50khz open loop. In all 3 I used a LTP using beefy triodes like the ECC99/12BH7/6BL7 with a CCS off an -80/100v negative rail as per PT's schematic's.
So, if you follow Pat's general theme in my experience you'll have a cracking amplifier. Remember to halve (if using 4 valves) the grid leak R's. Such amps will need properly testing for stability but Pat covers all that in his articles he wrote.
Happy building, Andy.
So, if you follow Pat's general theme in my experience you'll have a cracking amplifier. Remember to halve (if using 4 valves) the grid leak R's. Such amps will need properly testing for stability but Pat covers all that in his articles he wrote.
Happy building, Andy.
Using a mosfet follower direct coupled to the grid(s) is well documented under the name Powerdrive, by Tubelab. Tube purists and nostalgics can use a cathode follower instead, preferably one based on a high gm tube. Correctly done, a follower between the coupling cap and the grid takes care of the Miller C, prevents blocking distortion and sometimes even increases the output power by pushing the output stage into class A(B)2.
See: http://www.pmillett.com/A2_buffer.html
This shows the idea of a mosfet follower with a CCS (10M45s or 10M90S) driving a power tube. Depending on the circuit you may or may not need a capacitor between the follower and the power tube’s grid. Of course this works for “pure” Class A operation as well.
Regards, Gerrit
This shows the idea of a mosfet follower with a CCS (10M45s or 10M90S) driving a power tube. Depending on the circuit you may or may not need a capacitor between the follower and the power tube’s grid. Of course this works for “pure” Class A operation as well.
Regards, Gerrit
The most important problem of parallel tubes is current hogging. The circuit may work properly from the start, but it is a matter of time when one tube experiences thermal runaway. Individual cathode bias for each tube is the only solution, but it is not an optimal design since fixed bias gives most power with least distortion.
Make sure you put individual grid stoppers on each of the output tubes, to prevent oscillation (risk of motorboating). Otherwise driving them with a CC sink loaded follower (tube or MOSFET) will be fine. Naturally, you don't need a coupling capacitor for each output tube.
That was am interesting comment. All the circuits have looked at do have a separate coupling capacitor for each output tube. Are there pros and cons of a single coupling capacitor vs separate ones, apart from the component count/cost maybe?
With the circuit above, there are separate bias pots for each tube, so separate coupling caps are needed. With cathode bias, they should balance well enough with a single cap and grid return. Of course separate resistor and bypass cap for each. I don't see any advantage for fixed bias in SE.
You could also add a simple Mosfet follower between the driver tube and the coupling cap: Drain to B+, Gate to the anode (with gate stopper resistor and a zener diode for protection) and Source to ground through a sufficient Source resistor. A tube cathode follower would also work, perhaps you could use something like an ECL84 with a high mu triode (gain stage) and a high gm pentode (CF) in the same bottle?
Increase the coupling cap by 4x and decrease the grid leak for the output tubes to 1/4 of what the datasheet states for one tube. Separate cathode resistors and -caps is the only reasonable bias solution for four small(ish) tubes in parallel, and of course separate grid stoppers on each tube.
Increase the coupling cap by 4x and decrease the grid leak for the output tubes to 1/4 of what the datasheet states for one tube. Separate cathode resistors and -caps is the only reasonable bias solution for four small(ish) tubes in parallel, and of course separate grid stoppers on each tube.
I have ~25 NOS Zaerix if you need more...
Cost. I should add though that if you are doing fixed bias of each individual output tube then you certainly will need individual coupling capacitors.
You forget the cost (and effort) of each individual -ve bias circuit you will need for each tube in addition... Then contrast that with the quality (cost) of coupling capacitor you might like to use. Cathode power resistors and electrolytic by-pass caps don't need to be that expensive.
To be fair, I certainly do prefer fixed bias for Push-Pull, even though it is a little more work to implement. If it was only 4 Tubes per channel (2+2) then I would design for fixed bias.
Anything beyond 4 tubes per channel, I would probably go for cathode bias with a single coupling cap (with individual grid stoppers) and periodically test the tubes to see they are still reasonably well matched. With all due respect to Patrick Turner.
Last edited:
Then again... I have my own digitally calibrated tube tester so testing 12 tubes in a row (then re-sorting them to get equal sides) is less tedious for me than checking and adjusting bias of 12 tubes (6 per channel) on a running amp.
Patrick's design is good in that it uses easy to obtain tubes as well. They wouldn't be my personal choice, but I understand where he is coming from.
Patrick's design is good in that it uses easy to obtain tubes as well. They wouldn't be my personal choice, but I understand where he is coming from.
Thanks for the offer, but i am forever buying stuff for future projects, and have accumulated about a similar amount. I bought a tube tester a few years ago, and the guy said 'do you want the tubes and the transformers?' when I coillected it, so I have plenty of things to keep me going up to and including the eventual tomb and afterlife 🙂
ECL84 would be interesting. I checked and PCL84 is the same (they don't always correlate). I have a few of them.
Now I need to clear the decks a bit and do some planning. The transformers are auction finds, so did not break the bank, but are huge. The PSU isn't too demanding with TV tubes. I have some 70V 100VA toroids, so a doubler for the output tubes and an extra section to make a trippler for some head room for the driver and gain stage.
I get stuck with the finishing details in my previous projects, plus like to make the builds compact which requires a lot of planning. This wiill be a breadboard so I can experiment, but looking at those transformers maybe it needs to be a butchers block :-/