add gain before or after phase splitter?
I've built a little stereo amp around a pair of gov't surplus 815 tubes I found at a junk shop. The tubes look cool and are dual beam power tubes designed for push-pull outputs.
My one-tube 12ax7 phase splitter works great, but with a little bit of feedback from the OPT output doesn't have enough gain for a normal 1 volt input to drive the 815 to full power, which i think should take something like 30vrms.
Can someone educate me on the advisability of a) adding a preamp before the phase splitter vs. b) adding drivers after the phase splitter to drive the output tube grids?
I have really no experience in audio design or really even anything analog. I'm just amusing myself with a new-to-me technology. It sounds a lot better than I'd expected, actually!
I've built a little stereo amp around a pair of gov't surplus 815 tubes I found at a junk shop. The tubes look cool and are dual beam power tubes designed for push-pull outputs.
My one-tube 12ax7 phase splitter works great, but with a little bit of feedback from the OPT output doesn't have enough gain for a normal 1 volt input to drive the 815 to full power, which i think should take something like 30vrms.
Can someone educate me on the advisability of a) adding a preamp before the phase splitter vs. b) adding drivers after the phase splitter to drive the output tube grids?
I have really no experience in audio design or really even anything analog. I'm just amusing myself with a new-to-me technology. It sounds a lot better than I'd expected, actually!
If you need quite a bit of gain I would say add the gain stage before the PI. Then apply NFB to the cathode of that stage instead. Any distotion taking place within that stage will then be identical at each power valve grid.
If you only need a little bit of extra gain (like a gain factor less than 10) I would say add gain stages between PI and power valves. You can apply local feedback on these stages to increase their input impedance and linearity so that the PI is always constantly loaded, and so that you obtain a very low output impedance to drive the power valves. Then global NFB can be brought back to the PI and you have 'nested' feedback and a nice little amp.
DC coupling would be a nice touch in both cases.
If you only need a little bit of extra gain (like a gain factor less than 10) I would say add gain stages between PI and power valves. You can apply local feedback on these stages to increase their input impedance and linearity so that the PI is always constantly loaded, and so that you obtain a very low output impedance to drive the power valves. Then global NFB can be brought back to the PI and you have 'nested' feedback and a nice little amp.
DC coupling would be a nice touch in both cases.
If I were you I'd do both. To get the best out of the 815, you can run it in Class AB2 and that needs a direct coupled cathode follower (or source follower) driver to supply the grid current.
This would provide also a high impedance load for the 12AX7, which is a very wimpy tube to be used as a driver for beam tetrodes, even in class AB1, because of the low resistance they need in the grid (usually 100k or less, especially with fixed bias). Stick a good linear voltage gain stage on the front to feed the 12AX7 splitter, direct coupled, and use global NFB to the cathode of the first stage.
This would provide also a high impedance load for the 12AX7, which is a very wimpy tube to be used as a driver for beam tetrodes, even in class AB1, because of the low resistance they need in the grid (usually 100k or less, especially with fixed bias). Stick a good linear voltage gain stage on the front to feed the 12AX7 splitter, direct coupled, and use global NFB to the cathode of the first stage.
I actually only need a small amount of gain, perhaps a factor of 6 or so. I experimented with a simple triode gain stage in front with lots of local feedback to reduce the gain, however that seems to result in a big decrease in high-frequency response.
Merlinb suggested DC coupling - to get the output and inputs at the same DC voltage level, I take it I add resistance below the cathode of the following stage to bring the whole thing up to the plate voltage of the driving stage?
Is the benefit of DC coupling that it eliminates frequency-dependent phase shift caused by the coupling capacitors? If so - what is the harmful result of that phase shift?
Merlinb suggested DC coupling - to get the output and inputs at the same DC voltage level, I take it I add resistance below the cathode of the following stage to bring the whole thing up to the plate voltage of the driving stage?
Is the benefit of DC coupling that it eliminates frequency-dependent phase shift caused by the coupling capacitors? If so - what is the harmful result of that phase shift?
I had not considered going in the class AB2 direction and letting the grids be driven positive. Doesn't that produce lots of distortion? Also, I'm not sure I need that much power - I mostly listen to classical music at moderate volume in my living room.
I *think* the phase splitter is able to produce enough undistorted output. Part of my problem is that I'm not really sure how to determine what resistance I should have on the grids of the output tubes. I'm using 470k to ground on them, now, and watching the grids of the output tubes with a scope, I don't see any visible ugliness until it starts trying to drive the grids positive, and summing the phases gives a nice flat line.
I have the tubes cathode biased to around -20v, since that puts them at about 70% of rated plate dissipation at idle, and I've been trying to stay in the class A neighborhood.
I hadn't anticipated this much interest or information! I'll certainly post a schematic when I get home. Please feel free to tell me to bug off and read a book if this is all too basic; I'm working from about one semester of electronics in college, here.
thanks very much.
eric
I *think* the phase splitter is able to produce enough undistorted output. Part of my problem is that I'm not really sure how to determine what resistance I should have on the grids of the output tubes. I'm using 470k to ground on them, now, and watching the grids of the output tubes with a scope, I don't see any visible ugliness until it starts trying to drive the grids positive, and summing the phases gives a nice flat line.
I have the tubes cathode biased to around -20v, since that puts them at about 70% of rated plate dissipation at idle, and I've been trying to stay in the class A neighborhood.
I hadn't anticipated this much interest or information! I'll certainly post a schematic when I get home. Please feel free to tell me to bug off and read a book if this is all too basic; I'm working from about one semester of electronics in college, here.
thanks very much.
eric
There's nothing wrong with what you're doing. Your use of cathode bias allows you to use a higher grid resistor. Whether 470k is OK for your tubes I don't know but it seems too high for a beam tetrode. Certainly, the typical beam tetrode grid resistor would not be good for 12AX7 to drive.
Bear in mind that some kinds of music have really big transients. I'm thinking of music played by full symphony orchestras and choral music. You need a lot of headroom to ensure these transients don't drive the amp into clipping or, worse still, blocking distortion.
A direct-coupled cathode follower driver is an excellent way of preventing blocking distortion and it allows the tubes to take grid current, even if only momentarily.
Bear in mind that some kinds of music have really big transients. I'm thinking of music played by full symphony orchestras and choral music. You need a lot of headroom to ensure these transients don't drive the amp into clipping or, worse still, blocking distortion.
A direct-coupled cathode follower driver is an excellent way of preventing blocking distortion and it allows the tubes to take grid current, even if only momentarily.
Ok, while the site was down I decided to try using 6SN7 drivers between the 12ax7 phase splitter and the power tubes. I'm now using 100k instead of 470k grid resistors.
It can put out something like 90v p-p with no problem, so the 40v the power tubes need seems to be no problem. I don't seem to be lacking in gain anymore, even with a reasonable amount of global feedback. I can even bias it much lower now and let it run well into AB1 and it still sounds decent. So I'm pretty happy. I'll probably leave it biased higher though, around 70% max plate dissipation, just so it's class A most of the time until it gets loud. It sounds really nice.
I will, however, experiment with a follower and DC coupling to the power tube grids, just to see what I can learn.
It can put out something like 90v p-p with no problem, so the 40v the power tubes need seems to be no problem. I don't seem to be lacking in gain anymore, even with a reasonable amount of global feedback. I can even bias it much lower now and let it run well into AB1 and it still sounds decent. So I'm pretty happy. I'll probably leave it biased higher though, around 70% max plate dissipation, just so it's class A most of the time until it gets loud. It sounds really nice.
I will, however, experiment with a follower and DC coupling to the power tube grids, just to see what I can learn.
I'm glad it's working out well. Your decision about how much bias to use should be based on your own preference. You might find it's better to bias it a bit cooler in the summer, though!
P.S. Those tubes are fascinating. If I see any, I'm tempted to try them out. Please don't forget to post the schematic when you can.
P.S. Those tubes are fascinating. If I see any, I'm tempted to try them out. Please don't forget to post the schematic when you can.
I'll put up a schematic once I learn how to make it look decent in LTspice or whatever..
Anyway, i'm quite happy with it. I see that there's been a lot of discussion about screen supplies on beam tetrodes. The output tube I'm using (815) runs at a plate voltage of 400v but the screen can only go to 225v. One thing I noticed was that my screen supply (just a voltage divider off B+) was not stiff enough, causing lots of distortion as I turned the volume up.
Though I have no doubt that it's a far from ideal solution, by using smaller resistances (30k to B+, 50k and a cap to ground) it seems to be improved to the point that I'm running out of power from the power transformer at around the same point the screen voltage flopping around begins causing problems
I considered trying to properly regulate the screen supply, but then I suspect I'd have to quit using cathode bias and start regulating the bias supply too, and then it just all gets too complicated...
The 815 tubes are just too cool looking, as well.
Anyway, i'm quite happy with it. I see that there's been a lot of discussion about screen supplies on beam tetrodes. The output tube I'm using (815) runs at a plate voltage of 400v but the screen can only go to 225v. One thing I noticed was that my screen supply (just a voltage divider off B+) was not stiff enough, causing lots of distortion as I turned the volume up.
Though I have no doubt that it's a far from ideal solution, by using smaller resistances (30k to B+, 50k and a cap to ground) it seems to be improved to the point that I'm running out of power from the power transformer at around the same point the screen voltage flopping around begins causing problems
I considered trying to properly regulate the screen supply, but then I suspect I'd have to quit using cathode bias and start regulating the bias supply too, and then it just all gets too complicated...
The 815 tubes are just too cool looking, as well.
Schematic!!!
If you are using a split load phase inverter then read on - if not, ignore this post.
There are real benefits to having the gain stage after the split load phase inverter since its outputs only remain balanced while their individual loads remain balanced.
If your output stage is running Class AB then that is not the case. As soon as the output transitions from Class A to Class B (one tube cut off) then the loads on the anode and cathode of the PI become unbalanced.
A simple diff amp (ECC99, 6N1P, 6SN7, 12AU7 in order of my decreasing preference) running off the split load phase inverter will fix that since it always stays in Class A and will give you that extra gain of 5 to 10 times. Tie the diffamp cathodes together and run a current source inplace of the tail resistor to force perfect balance.
Cheers,
Ian
If you are using a split load phase inverter then read on - if not, ignore this post.
There are real benefits to having the gain stage after the split load phase inverter since its outputs only remain balanced while their individual loads remain balanced.
If your output stage is running Class AB then that is not the case. As soon as the output transitions from Class A to Class B (one tube cut off) then the loads on the anode and cathode of the PI become unbalanced.
A simple diff amp (ECC99, 6N1P, 6SN7, 12AU7 in order of my decreasing preference) running off the split load phase inverter will fix that since it always stays in Class A and will give you that extra gain of 5 to 10 times. Tie the diffamp cathodes together and run a current source inplace of the tail resistor to force perfect balance.
Cheers,
Ian
yes, it's a split load phase splitter, it's basically this one:
http://www.angelfire.com/electronic/funwithtubes/images/Amp-Spl-2.gif
except with a 6SN7 as the second stage rather than a 12AU7.
The output balance looks very close to perfect. I can't see or measure any difference in the phases with my scope, and they sum to a flat line.
The output looks more or less like this:
http://mexico.limpoc.com/~eric/output.pdf
Sorry about the crappy looking schematic. Trying to do things the open source way may not be the best approach.
http://www.angelfire.com/electronic/funwithtubes/images/Amp-Spl-2.gif
except with a 6SN7 as the second stage rather than a 12AU7.
The output balance looks very close to perfect. I can't see or measure any difference in the phases with my scope, and they sum to a flat line.
The output looks more or less like this:
http://mexico.limpoc.com/~eric/output.pdf
Sorry about the crappy looking schematic. Trying to do things the open source way may not be the best approach.
Now we have the schematic and see that there is alread a gain stage/driver after the split load phase inverter.
There is a simple way to get a bit more gain (ONLY about 1.5 times but that may be enough)
The combined load for the split load inverter stage is 200k ohms.
The input stage is running a 100K load. This stage (1/2 of the 12AX7) could be run at lower current with a higher load resistor without much change and you get the benefit of a smidge more gain plus the currents in the input stage and the split load inverter being near equal but opposite - so less power supply interaction. Try 220k for the anode load on the first stage and then adjust the 1k8 in the cathode such that the DC voltage on the anode is between 1/2 and 2/3 of the rail voltage on the power supply side of anode load resistor. You'll probably end up with this resistor at about 2K7.
Something to think about anyway.
Cheers,
Ian
Cheers,
Ian
There is a simple way to get a bit more gain (ONLY about 1.5 times but that may be enough)
The combined load for the split load inverter stage is 200k ohms.
The input stage is running a 100K load. This stage (1/2 of the 12AX7) could be run at lower current with a higher load resistor without much change and you get the benefit of a smidge more gain plus the currents in the input stage and the split load inverter being near equal but opposite - so less power supply interaction. Try 220k for the anode load on the first stage and then adjust the 1k8 in the cathode such that the DC voltage on the anode is between 1/2 and 2/3 of the rail voltage on the power supply side of anode load resistor. You'll probably end up with this resistor at about 2K7.
Something to think about anyway.
Cheers,
Ian
Cheers,
Ian
Ah, just to be clear, at the time I was complaining about the lack of gain, it was just the 12AX7.. adding the drivers after it took care of the gain problem and it has plenty now.
I do wonder about power supply interaction with all the different dropping resistors... starting with 400v, down to 210 for the screen, 300 for the 6sn7, somewhat less for the 12ax7, etc... None of it is exactly well regulated
Here is a picture of it hacked together on a piece of an old radio chassis for experimentation.
http://mexico.limpoc.com/~eric/test.jpg
eric
I do wonder about power supply interaction with all the different dropping resistors... starting with 400v, down to 210 for the screen, 300 for the 6sn7, somewhat less for the 12ax7, etc... None of it is exactly well regulated
Here is a picture of it hacked together on a piece of an old radio chassis for experimentation.
http://mexico.limpoc.com/~eric/test.jpg
eric
qq-
Just came across this thread. I too, have been thinking of an 815 based amplifier. They are just crying out to be used!
I was thinking of going PPP 815's. Then I can play with ultra linear and triode connetion. But yes, the screen voltage would greatly limit these designs.
Ever thought of using a regulator tube for the screen? Sure would look purrdy...
I have a sofis tube curve tracer and was going to curve an 815 with both sections conneted.
Small world! I live in Berkeley.
Best- Kent
Just came across this thread. I too, have been thinking of an 815 based amplifier. They are just crying out to be used!
I was thinking of going PPP 815's. Then I can play with ultra linear and triode connetion. But yes, the screen voltage would greatly limit these designs.
Ever thought of using a regulator tube for the screen? Sure would look purrdy...
I have a sofis tube curve tracer and was going to curve an 815 with both sections conneted.
Small world! I live in Berkeley.
Best- Kent
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.