Giaime said:
In this case, ditch the Ultralinear connection and add that 10db(v) of local feedback as an inner loop from the final plates back to the grids. Make sure that the screen voltage is well regulated (VR tube or bypassed Zener string at least, or an active regulator would be better). Then you can add the 6db(v) of gNFB. That should do the job.
Thank you very much Miles.
How to calculate the voltage divider from the 6550 plate to its grid?
I would assume that output resistance is 6550 rp, the feedback resistor has to be determined, and the resistor from "negative input" to ground is 6CG7 output impedance?
I am referring to this page:
http://www.aikenamps.com/GlobalNegativeFeedback.htm
So Ro is 6550 rp, Rf has to be determined for 10dB of feedback and Ri is 6CG7 stage output impedance, right?
Please correct me if I'm wrong
How to calculate the voltage divider from the 6550 plate to its grid?
I would assume that output resistance is 6550 rp, the feedback resistor has to be determined, and the resistor from "negative input" to ground is 6CG7 output impedance?
I am referring to this page:
http://www.aikenamps.com/GlobalNegativeFeedback.htm
An externally hosted image should be here but it was not working when we last tested it.
So Ro is 6550 rp, Rf has to be determined for 10dB of feedback and Ri is 6CG7 stage output impedance, right?
Please correct me if I'm wrong
Anyway, I did the math:
schematic attached. I neglected 6550 rp, and assumed 6CG7 anode node as 13.5k to ground (as Tubecad says).
Open loop from 6550 grid to anode there should be about 10.4x gain, or ~20dB. Now there is 3.28x, or ~10dB, so 10dB of feedback is there, with the 51k resistor. I'm not sure how to calculate C5 and C6...
Then global negative feedback: with partial feedback applied, gain from 6CG7 grids to output of the amp is about 2x, closed loop is 1x, so 6dB of feedback.
What do you think? In the schematic I haven't included G2 supply, I'm still thinking what to do.
- zener regulation for each output tube
- another 6550 as a pass regulator for all the 4 tubes.
schematic attached. I neglected 6550 rp, and assumed 6CG7 anode node as 13.5k to ground (as Tubecad says).
Open loop from 6550 grid to anode there should be about 10.4x gain, or ~20dB. Now there is 3.28x, or ~10dB, so 10dB of feedback is there, with the 51k resistor. I'm not sure how to calculate C5 and C6...
Then global negative feedback: with partial feedback applied, gain from 6CG7 grids to output of the amp is about 2x, closed loop is 1x, so 6dB of feedback.
What do you think? In the schematic I haven't included G2 supply, I'm still thinking what to do.
- zener regulation for each output tube
- another 6550 as a pass regulator for all the 4 tubes.
Attachments
I think the value of R5 is too small to give good linearity with an ECC83. 180k would be better but you also have to increase the value of R6, which is effectively in parallel with R5. I would increase R6 to 1Meg.
C5 and C8 are unnecessary if you take the local NFB loops back to the plates of the splitter. This will improve the bass performance, because the NFB loop results in a very low impedance, which is better before the coupling caps C2 and C3 than after.
I think your global NFB loop has insufficient gain within it to get enough NFB for pentode operation. For effective NFB, you should also include the first stage, giving you a true global loop. Take the NFB resistor R12 back to the cathode of U1. You would also have to reverse the OPT secondary connection, to keep the feedback negative.
For the sake of stability, you should also remove the coupling cap between first and second stages. Direct coupling would be easy to achieve - just put a 0.33uF (or bigger) between R11 and ground, removing R10 at the same time (you won't need it if you take the NFB back to U1 cathode). You also won't need a negative line for the CCS, unless you want to use a pentode CCS.
Frankly, I think you might be better served with a pentode as the first stage, instead of that ECC83, so there will be enough loop gain, but that's just my opinion. With beam tetrode mode operation of the OP stage, you're going to need a lot of NFB. It's all very well to aim for minimum NFB with triode or UL modes, but for pentode mode that's just unrealistic.
Finally, unless you really intend using the OP tubes strictly in class A, it would be better to use separate cathode resistors and bypass caps for the OP tubes. That will allow you (limited) operation in class AB1 and would also provide better balancing of the OP tubes. Better still would be fixed bias, giving you the flexibility to choose the operating point that sounds best, but you would then need a negative line of about 75 V or more.
C5 and C8 are unnecessary if you take the local NFB loops back to the plates of the splitter. This will improve the bass performance, because the NFB loop results in a very low impedance, which is better before the coupling caps C2 and C3 than after.
I think your global NFB loop has insufficient gain within it to get enough NFB for pentode operation. For effective NFB, you should also include the first stage, giving you a true global loop. Take the NFB resistor R12 back to the cathode of U1. You would also have to reverse the OPT secondary connection, to keep the feedback negative.
For the sake of stability, you should also remove the coupling cap between first and second stages. Direct coupling would be easy to achieve - just put a 0.33uF (or bigger) between R11 and ground, removing R10 at the same time (you won't need it if you take the NFB back to U1 cathode). You also won't need a negative line for the CCS, unless you want to use a pentode CCS.
Frankly, I think you might be better served with a pentode as the first stage, instead of that ECC83, so there will be enough loop gain, but that's just my opinion. With beam tetrode mode operation of the OP stage, you're going to need a lot of NFB. It's all very well to aim for minimum NFB with triode or UL modes, but for pentode mode that's just unrealistic.
Finally, unless you really intend using the OP tubes strictly in class A, it would be better to use separate cathode resistors and bypass caps for the OP tubes. That will allow you (limited) operation in class AB1 and would also provide better balancing of the OP tubes. Better still would be fixed bias, giving you the flexibility to choose the operating point that sounds best, but you would then need a negative line of about 75 V or more.
Thank you very much!!!
Altought I have to disagree in some points. First, the first tube is an ECC88 and not ECC83. I would like to keep the design as a simple single ended gain stage without feedback, coupled to a unity gain power buffer (because if you do the math, that is).
This is to retain the preminently second armonic distortion of the first stage, so it can "cover" the higher order distortion of the power stage. That's a trick that I learned recently tweaking an amp I made: it measures much worse, and probably if it was an amp for me, I wouldn't do it, but it is for a friend that "likes" that single ended sound.
Anyway, an useful trick would be to connect the local feedback resistors from the anodes of the 6550 to the anode of the 6CG7, thus avoiding a capacitor and gaining NFB even at lower frequencies.
I don't know if the NFB, now, is enought. You made a good point, probably I'll need extreme gain and big nfb ratios to make pentodes sound good, but hey, maybe I can get good results this way.
By the way, fixed bias isn't an option, a precise desing objective is to keep cathode bias, because my friend doesn't want to regulate the bias after each set of tubes. Don't ask me, he said so!
However, I'll add separate cathode resistors.
I will think about it... for now, an updated schematic, I made some math errors before.
Altought I have to disagree in some points. First, the first tube is an ECC88 and not ECC83. I would like to keep the design as a simple single ended gain stage without feedback, coupled to a unity gain power buffer (because if you do the math, that is).
This is to retain the preminently second armonic distortion of the first stage, so it can "cover" the higher order distortion of the power stage. That's a trick that I learned recently tweaking an amp I made: it measures much worse, and probably if it was an amp for me, I wouldn't do it, but it is for a friend that "likes" that single ended sound.
Anyway, an useful trick would be to connect the local feedback resistors from the anodes of the 6550 to the anode of the 6CG7, thus avoiding a capacitor and gaining NFB even at lower frequencies.
I don't know if the NFB, now, is enought. You made a good point, probably I'll need extreme gain and big nfb ratios to make pentodes sound good, but hey, maybe I can get good results this way.
By the way, fixed bias isn't an option, a precise desing objective is to keep cathode bias, because my friend doesn't want to regulate the bias after each set of tubes. Don't ask me, he said so!
However, I'll add separate cathode resistors.
I will think about it... for now, an updated schematic, I made some math errors before.
Attachments
Here's the basic design problem as I see it: either feedback around the output stage is minimal, in which case it's not terribly useful, or it's significant (a CF output stage is the reductio ad absurdem) in which case the burden shifts to the driver stage. And you can count on the driver stage giving 6dB more distortion for every 6dB of feedback around the output stage.
Ray's point about the wisdom of including earlier stages in the loop is absolutely correct.
Ray's point about the wisdom of including earlier stages in the loop is absolutely correct.
Mmmm...
thanks SY, so, you suggest to convert to a more "basic" design, removing local fb and including the first stage in the loop?
My primary objective, is to keep Zout down. Maybe I have to investigate in cathode feedback, a bit like McIntosh 275. Since I will have the transformers custom wound, I can do that, but I precisely need the ratio.
Another time, thank you all very much for the help. I have to think.
thanks SY, so, you suggest to convert to a more "basic" design, removing local fb and including the first stage in the loop?
My primary objective, is to keep Zout down. Maybe I have to investigate in cathode feedback, a bit like McIntosh 275. Since I will have the transformers custom wound, I can do that, but I precisely need the ratio.
Another time, thank you all very much for the help. I have to think.
Giaime said:
It's not very obvious to me that looking at shunt feedback in terms of gain stage and voltage is the easiest approach... The voltages needed at the grid of the 6550s are similar with or without shunt feedback, but when shunt feedback is applied the driver stage need to deliver current into the output stage. In fact, the output stage moves into a transresistance stage (current to voltage converter).
Here is a text about shunt feedback that I found quite useful:
http://users.ece.gatech.edu/~pallen/Academic/ECE_6412/Spring_2003/L260-ShuntShuntFb-2UP.pdf
I made a shunt feedback calculator (Excel) based on that text:
http://www.veiset.net/miniPP/Shuntfeedback-Excel97.xls
Jan E Veiset
NO-6600
Hello!
First I have to thank you all for all the suggestion you gave me. Now, after finishing a fine sounding guitar amp, I attempt the redesign of the amp I need.
Switched to 6L6GC (or KT66, the circuit could be easily adapted), cathode feedback AB2 output stage. x = 0.08 for the cathode winding: it means that there is about the 8% of the total plate signal swing at the cathode. Now, this is just a SPICE simulation.
Check it out, and let me see what do you think. Excuse me putting the image on tinypic, but it's too big for the forum software.
http://i8.tinypic.com/255i93q.gif
Power output appears to be on the high side, about 45W at 0.7% distortion. But since I don't believe that 6L6GC grids can be drive up to +30V without strong distortion, let's say this is a 30W amp
First I have to thank you all for all the suggestion you gave me. Now, after finishing a fine sounding guitar amp, I attempt the redesign of the amp I need.
Switched to 6L6GC (or KT66, the circuit could be easily adapted), cathode feedback AB2 output stage. x = 0.08 for the cathode winding: it means that there is about the 8% of the total plate signal swing at the cathode. Now, this is just a SPICE simulation.
Check it out, and let me see what do you think. Excuse me putting the image on tinypic, but it's too big for the forum software.
http://i8.tinypic.com/255i93q.gif
Power output appears to be on the high side, about 45W at 0.7% distortion. But since I don't believe that 6L6GC grids can be drive up to +30V without strong distortion, let's say this is a 30W amp
OK, I'll bite. First, the 47 ohm degeneration resistors do almost nothing. Lose 'em.
Second, you might want to reverse the resistor and the R||C in the input tube cathode.
Third, the 15k feedback resistor is awfully large. Are you trying to make the amp ultra-sensitive? I would think that 2-4k would be more appropriate for normal gain structure.
Fourth (minor drawing point), you seem to have the batteries backward.
Fifth, the 22k load on a 6SN7 is a rather heavy one. Since you have a negative rail available, you could go up to a 47k by referencing the common of that stage to -200V or so.
On the plus side, the 6L6 is a good choice of tube to use for AB2. And you did well to stagger the RC time constants for low frequency stability.
Second, you might want to reverse the resistor and the R||C in the input tube cathode.
Third, the 15k feedback resistor is awfully large. Are you trying to make the amp ultra-sensitive? I would think that 2-4k would be more appropriate for normal gain structure.
Fourth (minor drawing point), you seem to have the batteries backward.
Fifth, the 22k load on a 6SN7 is a rather heavy one. Since you have a negative rail available, you could go up to a 47k by referencing the common of that stage to -200V or so.
On the plus side, the 6L6 is a good choice of tube to use for AB2. And you did well to stagger the RC time constants for low frequency stability.
Ooooooh that's what I like to hear! 
I'm going to correct the schematic following your advices.
Thank you very much!!!
Now questions (please, forgive me )
- this amp "would" be intended for KT66. Is there (heater and rating aside) a big difference? I choose 6L6GC (I think that's the Duncan model) in SPICE because it's the model that looks more similar to the anode curves of the KT66. If you think that 6L6GC would perform better, tell me, because they're usually cheaper
And the OPT will be 5k. Should I go higher? With all the NFB I put in this thing, I think it's better to use a lower load to gain more power, but I would also like the increased damping factor that an increased anode-to-anode load would give me.
- about the time constants, thank you SY (finally I learned!!), I had a strange bump in the frequency response at ~2Hz or so, increasing the cathode bypass capacitor in the first stage moved out that resonance further down, where the amp should have practically no gain that can give problem.
- the 15k feedback resistor: it appears high to me, but to get 10dB of global feedback, that's it. Of course I will have to recalculate it, maybe I did a math error. Now it has ~0.6Vrms sensitivity, not bad since my "customer" will use it without an active preamp. I would not make sensitivity to drop under 1Vrms or so.
- those 47ohm resistors: the schematic is old, this morning I studied the new entry in the TubeCAD blog (http://www.tubecad.com/2006/08/blog0076.htm). If you go to the bottom of the page, Broskie shows how a little cathode degeneration in a differential amplifier does wonders. Since I have a relatively high voltage negative rail available, that's worth doing. I applied the technique in the EL36PP and it lowered the distortion of about 0.3% (but with lost in sensitivity of the amp) using 1k resistors under ECC83 tubes.
Here, with 6SN7, how high could I go? Is 1k enought?
- I have really to make the second stage "down" with respect to voltages. I mean increasing the anode load, and giving the grids a negative voltage reference, as you said.
- I'll invert also the R||C with the unbypassed R in the first stage, but... can I ask why? There will always be that electrolytic in the signal path...
- (about the batteries, they're reversed, but with negative voltages, so that's correct, but good point )
Thank you all!!!
Edit: nobody commented on cathode feedback. If I can make it to work good, it would be awesome. I'm still thinking "how" to say to the winder that I need 2 CFB windings. Do I need 2 of them or it's better to have one center tapped? The winding ratio with respect to the primary should be in the range of 8-10%, so the impedance ratio should be about 1:100 - 1:150, am I right?
It has been *hard* to figure out a proper distribution of inductances in the output transformer in spice, since I had not any model of OPTs. Particulary, for class A it all works well, but class AB...
I'm going to correct the schematic following your advices.
Thank you very much!!!
Now questions (please, forgive me
)- this amp "would" be intended for KT66. Is there (heater and rating aside) a big difference? I choose 6L6GC (I think that's the Duncan model) in SPICE because it's the model that looks more similar to the anode curves of the KT66. If you think that 6L6GC would perform better, tell me, because they're usually cheaper
And the OPT will be 5k. Should I go higher? With all the NFB I put in this thing, I think it's better to use a lower load to gain more power, but I would also like the increased damping factor that an increased anode-to-anode load would give me.
- about the time constants, thank you SY (finally I learned!!), I had a strange bump in the frequency response at ~2Hz or so, increasing the cathode bypass capacitor in the first stage moved out that resonance further down, where the amp should have practically no gain that can give problem.
- the 15k feedback resistor: it appears high to me, but to get 10dB of global feedback, that's it. Of course I will have to recalculate it, maybe I did a math error. Now it has ~0.6Vrms sensitivity, not bad since my "customer" will use it without an active preamp. I would not make sensitivity to drop under 1Vrms or so.
- those 47ohm resistors: the schematic is old, this morning I studied the new entry in the TubeCAD blog (http://www.tubecad.com/2006/08/blog0076.htm). If you go to the bottom of the page, Broskie shows how a little cathode degeneration in a differential amplifier does wonders. Since I have a relatively high voltage negative rail available, that's worth doing. I applied the technique in the EL36PP and it lowered the distortion of about 0.3% (but with lost in sensitivity of the amp) using 1k resistors under ECC83 tubes.
Here, with 6SN7, how high could I go? Is 1k enought?
- I have really to make the second stage "down" with respect to voltages. I mean increasing the anode load, and giving the grids a negative voltage reference, as you said.
- I'll invert also the R||C with the unbypassed R in the first stage, but... can I ask why? There will always be that electrolytic in the signal path...
- (about the batteries, they're reversed, but with negative voltages, so that's correct, but good point
)Thank you all!!!
Edit: nobody commented on cathode feedback. If I can make it to work good, it would be awesome. I'm still thinking "how" to say to the winder that I need 2 CFB windings. Do I need 2 of them or it's better to have one center tapped? The winding ratio with respect to the primary should be in the range of 8-10%, so the impedance ratio should be about 1:100 - 1:150, am I right?
It has been *hard* to figure out a proper distribution of inductances in the output transformer in spice, since I had not any model of OPTs. Particulary, for class A it all works well, but class AB...
One winding, CT, for cathode feedback.
I think the emitter degeneration resistors don't really need to be there. Broskie's model is interesting, but is incomplete in a system sense. I'm in a bit of a hurry now (getting ready to leave on a trip), but will try to expand on that later if someone else doesn't.
I've never used KT66, so can't comment on their suitability for AB2. I have used various 6L6 types (6L6, 7027, EL37) and they worked great. EC8010 (who has a lot of experience with KT66) tells me that 2A3 is a great AB2 candidate as well. I'd follow the datasheet recommendations on load for that class of operation.
I think the emitter degeneration resistors don't really need to be there. Broskie's model is interesting, but is incomplete in a system sense. I'm in a bit of a hurry now (getting ready to leave on a trip), but will try to expand on that later if someone else doesn't.
I've never used KT66, so can't comment on their suitability for AB2. I have used various 6L6 types (6L6, 7027, EL37) and they worked great. EC8010 (who has a lot of experience with KT66) tells me that 2A3 is a great AB2 candidate as well. I'd follow the datasheet recommendations on load for that class of operation.
Here it is. Feedback increased (both global, now ~12dB, and cathode, now it's 20%), it had a bit too much gain.
http://img55.imageshack.us/img55/144/amplixpierluigimkiiyp9.gif
http://img55.imageshack.us/img55/144/amplixpierluigimkiiyp9.gif
Hello Brian,
good point, maybe I should have been clearer on that subject.
What you can gain from an IRF820 in place of a 5687, is:
- you can swing up to the positive rail, so the dedicated power supply can be lower in voltage, thus cheaper and safer;
- much lower output impedance, and a single IRF820 could pass amperes, not mAs, before any sign of stress or distortion;
- less heat, less filament power, a couple of sockets less.
If there's something I missed, I'll let you know
good point, maybe I should have been clearer on that subject.
What you can gain from an IRF820 in place of a 5687, is:
- you can swing up to the positive rail, so the dedicated power supply can be lower in voltage, thus cheaper and safer;
- much lower output impedance, and a single IRF820 could pass amperes, not mAs, before any sign of stress or distortion;
- less heat, less filament power, a couple of sockets less.
If there's something I missed, I'll let you know
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.