I have used this technique on several types of tubes, but the EL-84 was not tried. I have not noticed any reduction in available output power, provided there is enough drive to offset the feedback.
True the DC bias does not change, but now there is an AC voltage on the cathode. During a positive going signal excursion the grid will be driven in the positive direction, and the cathode will also be driven in the positive direction by the feedback. The relationship between the grid and the cathode remains the same, and the instantaneous voltage (both DC and AC) from the grid to the cathode is the same as it always was. The driver however, must supply a voltage equal to the grid to cathode voltage PLUS the cathode to ground voltage.
This is the same effect as an unbypassed cathode, except the feedback voltasge (and current) is now derived from the secondary of the transformer, so the feedback can help to mellow out a crummy sounding transformer. If the transformer is really bad (significant phase shift) the amp can oscillate.
This technique can really improve the bass and transient response of some amps, and it seems to do nothing on other amps. If the transformer has multiple impedance secondaries, it is possible to have the speaker on one tap and the feedback on another. I have the speaker on the 8 ohm tap, and the cathode on the 16 ohm tap. Swap the primary leads if the amp oscillates.
This trick will work on a push pull amp ONLY if the transformer is symmetrically wound. If not, the feedback will cause some really ugly sounding intermodulation distortion. Some of the old UTC's work OK. Audio Research did this on their P-P amps, like the D-76. They used special transformers.
True the DC bias does not change, but now there is an AC voltage on the cathode. During a positive going signal excursion the grid will be driven in the positive direction, and the cathode will also be driven in the positive direction by the feedback. The relationship between the grid and the cathode remains the same, and the instantaneous voltage (both DC and AC) from the grid to the cathode is the same as it always was. The driver however, must supply a voltage equal to the grid to cathode voltage PLUS the cathode to ground voltage.
This is the same effect as an unbypassed cathode, except the feedback voltasge (and current) is now derived from the secondary of the transformer, so the feedback can help to mellow out a crummy sounding transformer. If the transformer is really bad (significant phase shift) the amp can oscillate.
This technique can really improve the bass and transient response of some amps, and it seems to do nothing on other amps. If the transformer has multiple impedance secondaries, it is possible to have the speaker on one tap and the feedback on another. I have the speaker on the 8 ohm tap, and the cathode on the 16 ohm tap. Swap the primary leads if the amp oscillates.
This trick will work on a push pull amp ONLY if the transformer is symmetrically wound. If not, the feedback will cause some really ugly sounding intermodulation distortion. Some of the old UTC's work OK. Audio Research did this on their P-P amps, like the D-76. They used special transformers.
"I have never heard of a parafeed cap in the order of 100uF being used for any application. Have a look at the Valve magazine website for an article on the theoretical derivation of parafeed cap size."
Let me point one out to you.
http://www.alphalink.com.au/~cambie/EL36.htm#Intro
Scroll down to near the bottom. The unusual feature of this amp is the use of the PL36 which has a very low Rp and hence reuires a 100uf parafeed cap.
This is an interesting and cheap project to explore the ins and outs of the SEPP concept.
Shoog
Let me point one out to you.
http://www.alphalink.com.au/~cambie/EL36.htm#Intro
Scroll down to near the bottom. The unusual feature of this amp is the use of the PL36 which has a very low Rp and hence reuires a 100uf parafeed cap.
This is an interesting and cheap project to explore the ins and outs of the SEPP concept.
Shoog
tubelab.com said:
Thanks again tubelab. I really wish I'ld stop asking such obvious-in-hindsight questions! The driver won't have an issue. This is bumped up on the to-try list.
Hey nothing is obvious until you have tried it. Believe me, I have spent plenty of time poking around in amplifiers with scopes and other equipment. I just wish that I had documented my earlier experiments better so that I could easily put them on the web site.
Tubelab:
Good to hear that more people have positive experiences of cathode feedback.
A couple of months ago I made a pair of OPT´s designed for this kind of operation, where about 25% of the primary winding was in the cathode circuit together with the secondary winding.
Sounded darn good in a prototype (triode wired) PL504 amp.
It needs a good driver though, I used a choke (Hammond 156C) loaded PC86.
Good to hear that more people have positive experiences of cathode feedback.
A couple of months ago I made a pair of OPT´s designed for this kind of operation, where about 25% of the primary winding was in the cathode circuit together with the secondary winding.
Sounded darn good in a prototype (triode wired) PL504 amp.
It needs a good driver though, I used a choke (Hammond 156C) loaded PC86.
A side note, why is the feedback coupled from plate to plate? Wouldn't the same effect be reached by coupling the output tube plate to the driver tube's cathode by way of a voltage divider resistor between driver Rk and ground? Or would the high Rp of the output tube see problems running into the lower resistance of the 12ax7 cathode.. or something like that?
Shoog,
Have a look at this link for a mathematical derivation of parafeed capacitor size:
http://www.valvediy.com/ramseypg1.html
Although I don't think it is the final word on the subject, it offers some very useful insight into the subject and empirical data to back up its assertions.
The circuit you refer to is known as a Murray amplifier (see http://www.wdehaan.demon.nl/tubeamps/murray/ ) for a synopsis. The circuit lowers the Rp of the output tube to the point where it can directly drive a 500R load: hence the large parafeed capacitor.
The EL36 has an Rp of approx 750R in conventional triode strapped mode and would normally require a parafeed capacitor in the order of 3uF, the same as normally employed for a 2A3 (with exactly the same 750Rp) by just about every amp designer on the planet.
However, If you wish to stick huge caps on the outputs of your parafeed amps, that's fine by me. I hope you enjoy the results.
regards
Have a look at this link for a mathematical derivation of parafeed capacitor size:
http://www.valvediy.com/ramseypg1.html
Although I don't think it is the final word on the subject, it offers some very useful insight into the subject and empirical data to back up its assertions.
The circuit you refer to is known as a Murray amplifier (see http://www.wdehaan.demon.nl/tubeamps/murray/ ) for a synopsis. The circuit lowers the Rp of the output tube to the point where it can directly drive a 500R load: hence the large parafeed capacitor.
The EL36 has an Rp of approx 750R in conventional triode strapped mode and would normally require a parafeed capacitor in the order of 3uF, the same as normally employed for a 2A3 (with exactly the same 750Rp) by just about every amp designer on the planet.
However, If you wish to stick huge caps on the outputs of your parafeed amps, that's fine by me. I hope you enjoy the results.
regards
Applying feedback to the cathode of the 12AX7 could be done - that would be global feedback across two stages, applying it at the plate is just local feedback in the output stage.
The feedback as applied actually acts primarily on the output tube and as small amounts of feedback are generally all that is required the feedback resistor can be quite large relative to the thevenin equivalent resistance of the input tube's rp and plate load resistor, and this results in minimal upset in the dc conditions on the driver tube, which in any case could be compensated for by slightly increasing the plate load resistance.
Note that this technique will result in a considerable reduction in source impedance of the output stage with better damped response in the output transformer, generally extends bandwidth and improves linearity as well.
Hopefully this is not too confusing, from an ac standpoint think of the triode as a perfect voltage generator in series with the thevenin equivalent of rp and the plate load resistor - the thevenin resistance then looks like Rin of a typical feedback network with the output tube as the amplifying element in that network. Rfb is the resistor between the two plates.
This is a serious simplification of what is going on. Hope it is a little helpful.
I suspect this technique works best with beam tetrodes and pentodes although I have seen it applied to triodes as well.
The technique was sometimes employed in single ended consule amplifiers to reduce the effects of cheap output transformers.
Kevin
The feedback as applied actually acts primarily on the output tube and as small amounts of feedback are generally all that is required the feedback resistor can be quite large relative to the thevenin equivalent resistance of the input tube's rp and plate load resistor, and this results in minimal upset in the dc conditions on the driver tube, which in any case could be compensated for by slightly increasing the plate load resistance.
Note that this technique will result in a considerable reduction in source impedance of the output stage with better damped response in the output transformer, generally extends bandwidth and improves linearity as well.
Hopefully this is not too confusing, from an ac standpoint think of the triode as a perfect voltage generator in series with the thevenin equivalent of rp and the plate load resistor - the thevenin resistance then looks like Rin of a typical feedback network with the output tube as the amplifying element in that network. Rfb is the resistor between the two plates.
This is a serious simplification of what is going on. Hope it is a little helpful.
I suspect this technique works best with beam tetrodes and pentodes although I have seen it applied to triodes as well.
The technique was sometimes employed in single ended consule amplifiers to reduce the effects of cheap output transformers.
Kevin
kevinkr said:
I think I see.... Feedback to the 12ax7 cathode would be taked from the transformer rather than the plate of the output tube... is that what you mean by global, or are you referring to something else by that term?
You are referring to the plate to plate feedback with that? You had lost me there since I had never heard the term "thevenin resistance", but I looked up the math and it is very familiar, I just had never heard that term. What you are saying is that the triode's plate resistor and internal resistance are summed, and that is what the feedback resistor sees as it's input impedance. So feedback applied to the cathode would see the Rk and cathode impedance as it's input... fair enough. I suppose that is the simplest way.
Usually there are frequency compensation networks... or at least a capacitor bypassing Rfb, I suppose that could apply here as well.
Actually you can take the feedback from the plate of the output tube or the transformer secondary and feed it back to the cathode. Either can be done and has been done.
In referring to thevenin's theorem it was just another way of saying that the rp of the 12AX7 and the plate resistor appear in parallel. (It is the equivalent parallel resistance.) The triode can be modeled as a "perfect voltage source" feeding that resistance, which is the summation point for the feedback coming from the plate of the output tube. It's not quite a perfect approximation, but it does work. The capacitor you mention is for lag compensation and would usually be used with feedback from the transformer secondary, and less frequently in the RH style configuration discussed.
Intestingly the Citation II used several nested feedback loops which serve the same purpose but were implemented differently. (I don't have the Citation 2 schematic in front of me unfortunately - that filing cabinet now lives in the garage since the failure of my business.. LOL)
Kevin
In referring to thevenin's theorem it was just another way of saying that the rp of the 12AX7 and the plate resistor appear in parallel. (It is the equivalent parallel resistance.) The triode can be modeled as a "perfect voltage source" feeding that resistance, which is the summation point for the feedback coming from the plate of the output tube. It's not quite a perfect approximation, but it does work. The capacitor you mention is for lag compensation and would usually be used with feedback from the transformer secondary, and less frequently in the RH style configuration discussed.
Intestingly the Citation II used several nested feedback loops which serve the same purpose but were implemented differently. (I don't have the Citation 2 schematic in front of me unfortunately - that filing cabinet now lives in the garage since the failure of my business.. LOL)
Kevin
Hey, thanx Kevin. That helps.
I have the Citation schematics, and I gave them closer look... WOW! Feedback loops all over.
I think I'll try the plate to cathode feedback since I am the most familiar with that configuration, and it gives a nice little place for a voltage divider too.
I have the Citation schematics, and I gave them closer look... WOW! Feedback loops all over.
I think I'll try the plate to cathode feedback since I am the most familiar with that configuration, and it gives a nice little place for a voltage divider too.
- Status
- Not open for further replies.