And that explains why Morgan Jones did what he did with the Crystal Palace (a light bulb just went on!). He took the input LTP and AC-coupled that to the driver stage diff amp. Then the driver stage was DC-coupled to cathode followers and the output stage. So the AC coupling was between the two diff pairs, where the DC balance is not important.
But why is it not a "fail" that the driver stage diff and the push-pull output stage are DC-coupled? What do the cathode followers do, or does the fixed bias do something that allows good balance from the output stage? Or is it that the output stage is transformer loaded?
Thanks for the info so far.
--
And what happens if the LTP is DC coupled to a PP driver stage with a center tapped choke as plate load?
But why is it not a "fail" that the driver stage diff and the push-pull output stage are DC-coupled? What do the cathode followers do, or does the fixed bias do something that allows good balance from the output stage? Or is it that the output stage is transformer loaded?
Thanks for the info so far.
--
And what happens if the LTP is DC coupled to a PP driver stage with a center tapped choke as plate load?
Last edited:
Morgan incorporated a DC balance pot in the driver stage and used a lot of matching of single triodes. I'm rather down on the idea of running 10 mA through a pot wiper.
Look at the ETF lecture from Morgen Jones about transformers:
http://www.triodefestival.net/uploads/images_uploads/Measuring_transformer_leakage_inductance.pdf
Strange that so less transformer manufactures do give the information people like to have about there transformers....
Just Tango, Monolith Magnetics, Menno van der Veen give detailed information about there transformers.
Thanks for all this info, SY.
It's a bit daunting to find out how many compromises have to be made when designing something as simple as a push-pull triode power amp.
So from what I think you've been saying, you think a single LTP with a common-cathode stage driving it is a less error-prone design than an LTP DC-coupled to a PP diff driver stage?
Therefore, the worst of all worlds would be a 3-stage PP amp, with an input LTP DC-coupled to a PP diff driver stage, DC-coupled to the PP output stage?
--
It's a bit daunting to find out how many compromises have to be made when designing something as simple as a push-pull triode power amp.
So from what I think you've been saying, you think a single LTP with a common-cathode stage driving it is a less error-prone design than an LTP DC-coupled to a PP diff driver stage?
Therefore, the worst of all worlds would be a 3-stage PP amp, with an input LTP DC-coupled to a PP diff driver stage, DC-coupled to the PP output stage?
--
Therefore, the worst of all worlds would be a 3-stage PP amp, with an input LTP DC-coupled to a PP diff driver stage, DC-coupled to the PP output stage?
That can be made to work, but it will be much more complicated and twitchy. Look at everything MJ had to do for things like regulation and thermal compensation, then double it!
But why is it not a "fail" that the driver stage diff and the push-pull output stage are DC-coupled?
Putting it crudely: It is not a fail because the cathode followers are independent of one another. Small differences in their grid voltages will not amount to much difference in bias between them, as the large cathode resistors make them largely self compensating.
But with and LTP the triodes share a cathode resistor, so they cannot both come to a happy optimum bias at the same time.
Tubelab helped me develop an amp that has a very similar front end to the amp in discussion here. A dual diff amp DC coupled driving MOSFET source followers to PP 6L6. As Sy mentioned, DC balance was an issue identified by Tubelab in the design, and he recommended a small compensation circuit to adjust for it. The whole discussion is in this thread http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html The circuit in discussion is here http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp-45.html#post2179774
Here is the of the schematic of the complete amp. You can see the circuit described.
Here is the of the schematic of the complete amp. You can see the circuit described.
Attachments
Putting it crudely: It is not a fail because the cathode followers are independent of one another. Small differences in their grid voltages will not amount to much difference in bias between them, as the large cathode resistors make them largely self compensating.
Remember that the diff amp is direct coupled to the cathode followers- that's the primary source of DC imbalance unless there's a balance pot in the diff amp cathode circuit.
Tubelab helped me develop an amp that has a very similar front end to the amp in discussion here. A dual diff amp DC coupled driving MOSFET source followers to PP 6L6. As Sy mentioned, DC balance was an issue identified by Tubelab in the design, and he recommended a small compensation circuit to adjust for it. The whole discussion is in this thread http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html The circuit in discussion is here http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp-45.html#post2179774
Here is the of the schematic of the complete amp. You can see the circuit described.
I remember that thread, and read it with much interest. Is there no gNFB loop around that amp? If so, I'm surprised that PP UL 6L6GC's would sound acceptable run that way.
My last stab at a PP amp had 12AT7 LTP DC-coupled > 12BH7A diff AC-coupled > 6L6GC-triodes. It needed some gNFB to get rid of 'nasties.' Might have been the Dyna ST70 OPT's. They're designed to have NFB around them.
--
Yes, admittedly any DC imbalance will be passed on to the power stage, but the cathode followers themselves will be happy.Remember that the diff amp is direct coupled to the cathode followers- that's the primary source of DC imbalance unless there's a balance pot in the diff amp cathode circuit.
Last edited:
noisy circuit
The problem with the original Mullard circuit using EF86 and 12AX7 is for sensitive speakers it has far too much gain and is noisy. To reduce the gain one way is to change the EF86 to a triode, but maybe best to change the 12Ax7 for the very linear 6CG7.
Phil
The problem with the original Mullard circuit using EF86 and 12AX7 is for sensitive speakers it has far too much gain and is noisy. To reduce the gain one way is to change the EF86 to a triode, but maybe best to change the 12Ax7 for the very linear 6CG7.
Phil
Just a few aside comments (since SY has scooped all the essential details! )
Folks seem to get over-absorbed with changes to the front end EF86 in comparison with rather greater gains (as in advantages, not µ!) regarding a phase splitter change.
Yes, the gain is high, and pentodes make more noise than triodes. But that is relative. One really has to stick one's ears into the loudspeaker to hear the difference in noise, other things being in order. To me the advantage lies in the almost 4x lower distortion for pentodes than triodes at low signal. As said low, but one must also consider the effect of mixing NFB signal on a cathode with signal on G1. Any non-linearity there is not cancelled by global NFB (but that is another subject).
A greater advantage of pentodes for me is in the lower influence of any Miller capacitance because of vartying input impedance of the feeding source. With triode input of the 12AX7 kind this can influence GNFB h.f. stability. Further, the lower anode operating voltage in the tube's linear region possible with pentodes than with triodes (and to boot, the gain step and phase functions possible with g2 bypass constants, should l.f. stability enter the fray.)
Going on to the phase splitter, Miller effect from high µ triodes on the first stage again comes into play. For these reasons I also prefer a lower µ lower Rp tube for the phase splitter (in fact, if folks promise not to pick up stones, I do not like 12AX7s anywhere in a power amplifier ). My choice has changed from 12AT7s to ECC88s, which I found somewhat more linear, but that will depend on other circuit values.
Thus I myself found the most satisfactory application, bandwidth and distortion-wise (open loop naturally), with an EF86 and ECC88. I myself have not indulged in using CCs', finding the old cathode resistor satisfactory, also over time (perhaps I was just fortunate in my double triodes aging in equal fashion ...) But the possibilities are multiple, even allowing for some taste.
Folks seem to get over-absorbed with changes to the front end EF86 in comparison with rather greater gains (as in advantages, not µ!) regarding a phase splitter change.
Yes, the gain is high, and pentodes make more noise than triodes. But that is relative. One really has to stick one's ears into the loudspeaker to hear the difference in noise, other things being in order. To me the advantage lies in the almost 4x lower distortion for pentodes than triodes at low signal. As said low, but one must also consider the effect of mixing NFB signal on a cathode with signal on G1. Any non-linearity there is not cancelled by global NFB (but that is another subject).
A greater advantage of pentodes for me is in the lower influence of any Miller capacitance because of vartying input impedance of the feeding source. With triode input of the 12AX7 kind this can influence GNFB h.f. stability. Further, the lower anode operating voltage in the tube's linear region possible with pentodes than with triodes (and to boot, the gain step and phase functions possible with g2 bypass constants, should l.f. stability enter the fray.)
Going on to the phase splitter, Miller effect from high µ triodes on the first stage again comes into play. For these reasons I also prefer a lower µ lower Rp tube for the phase splitter (in fact, if folks promise not to pick up stones, I do not like 12AX7s anywhere in a power amplifier ). My choice has changed from 12AT7s to ECC88s, which I found somewhat more linear, but that will depend on other circuit values.
Thus I myself found the most satisfactory application, bandwidth and distortion-wise (open loop naturally), with an EF86 and ECC88. I myself have not indulged in using CCs', finding the old cathode resistor satisfactory, also over time (perhaps I was just fortunate in my double triodes aging in equal fashion ...) But the possibilities are multiple, even allowing for some taste.
An externally hosted image should be here but it was not working when we last tested it.
Morgan incorporated a DC balance pot in the driver stage.
http://img153.imageshack.us/img153/3595/screenshot20100508at348.png
That's the way to go, I think. A pair of LEDs with a pot across them will give enough bias to balance the DC.
I could add a 100ohm resistor in series with C1 and ground and feed the GNFB there.
Last edited:
High plate load (a CCS works well for this), as little cathode degeneration as possible (an LED works well for this). Keep the load on the output as high impedance as possible. The linearity is startlingly good.
http://www.diyaudio.com/forums/tubes-valves/164920-guess-tube.html
http://www.diyaudio.com/forums/tubes-valves/164920-guess-tube.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
- Home
- Amplifiers
- Tubes / Valves
- MULLARD 5-20 build with mods