• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Cathode and other types of feedback

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I was rereading Morgan Jones' book (hoping that some more would sink in) and I came across a section on feedback. Just want to make sure I've got the basic ideas right...

Cathode feedback is best used from dedicated taps built into the transformer. These are different from an "Ultralinear" connection in that you can use whatever voltage you want instead of applying the entire B+, right?

If you don't have the dedicated taps on the transformer, you can achieve a type of cathode feedback from one of the secondary taps. Do you have to have an unused tap (i.e. an extra 8 ohm tap when you have the speakers hooked to the 4 ohm) or can you source feedback from the same tap you have hooked to the speakers?

Jones mentioned in passing what he thought should be the optimal configuration, feedback from the speaker terminals. This makes the most sense to me too but I must be missing something since I have never heard of anyone actually using this technique. What would be involved in using feedback from the speakers around the output stage of an amp? I suppose taking feedback from the speakers and applying it to the input would be the very definition of a global feedback chain:) I tend to shy away from global feedback because of stability issues, I assume that it would be even more tricky if the feedback was sourced from the very end of the chain, am I right?

Thanks for any info you can send my way, still scratching my head over how some of these things work, I'm hoping to start on my first amp before too long...

Isaac
 
Isaac,

Cathode feedback is frequently specified as a % of the primary winding. It's "correct" to speak in terms of tertiary windings.

If NFB from the speaker enclosure terminals is used, I think monoblock amps and short cables (primary and NFB) are indicated. My hunch is that it's problematic. Plenty of NICE amps have all their NFB internal.
 
Using the secondary to provide cathode feedback has advantages and drawbacks. There was just a thread about this sometime in the past week. My take- useful for some pentode circuits, marginal to poor in most other applications.

Likewise, over in Solid State, there was just a thread about taking feedback directly from speaker terminals. Consensus was that it was a solution in search of a problem. Best to follow Eli's suggestions- short speaker leads and amps located close to the speakers. From what I understand, that's what MJ actually does in his own systems.
 
Hi Isaac,

The advantage of local feedback from loudspeaker terminals is often marginal. To quote figures, for a p.p. EL84 output there will be some 5 dB feedback, but for a 30W 6L6 stage less than 2 dB. (For a p.p. stage you will need a balanced speaker winding; on a 16 ohm winding with a 4 ohm tap, the latter would be a center tap.)

To answer your question, one could theoretically use given UL taps for such feedback to cathodes via isolating capacitors, but then you will still need a suitable cathode resistor which will shunt such windings, thus not parctical.

It must be realised that theoretically all such feedback is ultra-linear or distributed load. I am sorry I cannot post a diagram at the moment, with the aid of which you would have been better able to follow. But if you picture that in UL the current between the screen tap and B+ is cathode current, and the same flows in your speaker or other winding feedback, the logic would follow. Thus the optimum for such a system would be to have a tertiary winding of say 25% of the turns of the anode-B+ portion (i.e. representing a 20% tap) as a feedback winding. Thus you would be realising the advantages of UL as well as the cathode feedback. The Quad II amplifier did it this way. For a 30W 6L6 output stage there would now be a handy 11 dB of feedback.

The grid driving signal must of course then be more. To keep that within limits the Quad used equivalent 10% taps only, but I have used 20% "taps" for superb performance in a 100W p.p. 6L6 amplifier. (Only my grid signal requirement was 350Vpp; somewhat of a tall order!)

Yes, global feedback can give problems if the design is not right, but I think it is often unneccessarily vilified. One does need at least a scope and signal generator to optimise - this is a subject on its own. To repeat a basic truth which has often been mentioned on other threads: It is mainly where too much feedback has been used in an effort to make poor designs decent, that feedback problems have arisen. Hope this explains matters. If you have difficulty in following the UL-cathode feedback connection and are interested, I can go scan-and-post a diagram.

Regards.
 
SY said:
Using the secondary to provide cathode feedback has advantages and drawbacks. There was just a thread about this sometime in the past week. My take- useful for some pentode circuits, marginal to poor in most other applications.

Likewise, over in Solid State, there was just a thread about taking feedback directly from speaker terminals. Consensus was that it was a solution in search of a problem. Best to follow Eli's suggestions- short speaker leads and amps located close to the speakers. From what I understand, that's what MJ actually does in his own systems.

Thanks for the responses! I have a single ended pentode amp in mind, a 12av5 amp to be specific. Tubelab has me all excited about these tubes and he seems to be using cathode feedback to great effect with some cheapie transformers. Tertiary windings are out of the question for right now due to monetary constraints, but I would like to keep the option open since it seems to work well for his set up. Those particular Edcor transformers have only one secondary connection. I'm assuming that I can use it for feedback and connection to the speakers. Would I be able to get better performance if there were additional taps on the secondary?

Isaac
 
In the case of the Edcor transformer there is only one secondary, so I used it for both CFB and the speaker. The speaker winding is also the wrong phase for CFB, so I have the hot (8 ohm terminal) grounded and the cathode connected to the ground terminal. These are all compromizes, nevertheless it works real good for an $18 transformer. If the transformer has multiple impedance taps, you hook the speaker up where it belongs, and try each tap for CFB. Pick the one that works the best in your situation. If it happens to be the same one that the speaker is on, use it.

The higher impedance taps provide more feedback, which lower the output impedance, tighten up the bass, and lower the gain. This requires more drive. As in life each choice has its advantages, and disadvantages. With the Edcor, there is only one choice CFB, or no CFB.

I have seen transformers with a CFB winding. I have also seen transformers with a seperate winding for UL. Sometimes either of these windings are called a tertiary feedback winding. Transformers with these windings are not common. In the ideal world the transformer would have a seperate winding for CFB, and another for the screen grid (UL), and multiple output taps. Sweep tubes like your 12AV5's would benefit from a seperate screen winding so you could run the plate at 500 volts and the screen at 200 and still use UL.

I have tried CFB with several transformers. I find that you get the most obvious improvement on low cost transformers. There are some transformers that just plain don't like CFB. I haven't attempted to figure out just why this is yet, but with some transformers, the frequency response gets worse with CFB.

All of these experiments were done in SE mode either triode or UL. CFB with push pull requires a transformer that is completely symmetrical, or it won't work. These are real hard to find.
 
tubelab.com said:
In the case of the Edcor transformer there is only one secondary, so I used it for both CFB and the speaker. The speaker winding is also the wrong phase for CFB, so I have the hot (8 ohm terminal) grounded and the cathode connected to the ground terminal. These are all compromizes, nevertheless it works real good for an $18 transformer. If the transformer has multiple impedance taps, you hook the speaker up where it belongs, and try each tap for CFB. Pick the one that works the best in your situation. If it happens to be the same one that the speaker is on, use it.


Ah thanks! Appreciate the straight talk... Since I don't have much in the way of measurment devices yet, is it possible to do this by ear? Are the differences that great? Will I be able to tell if there is a phase problem by ear as well?

The higher impedance taps provide more feedback, which lower the output impedance, tighten up the bass, and lower the gain. This requires more drive. As in life each choice has its advantages, and disadvantages. With the Edcor, there is only one choice CFB, or no CFB.

OK I see. With your powerdrive circuit, extra gain doesn't seem to be much of a problem. It seems as though the powerdrive plus local feedback around the output stage is a good combo, just need to figure out what transformer to use I suppose.

I have seen transformers with a CFB winding. I have also seen transformers with a seperate winding for UL. Sometimes either of these windings are called a tertiary feedback winding. Transformers with these windings are not common. In the ideal world the transformer would have a seperate winding for CFB, and another for the screen grid (UL), and multiple output taps. Sweep tubes like your 12AV5's would benefit from a seperate screen winding so you could run the plate at 500 volts and the screen at 200 and still use UL.

Like so many other things, custom made would probably be best... Any ballpark what something like that would run? I'm betting more than $18 a piece:) If the price wasn't outrageous, I might consider ordering some...

Thanks again, I am going to build this single ended amp soon. Hope to have my 3 channel(!) power supply done by the end of next month and then the experimenting can begin in ernest!:)

Isaac
 
tubelab.com said:
so I have the hot (8 ohm terminal) grounded and the cathode connected to the ground terminal.

Hello George,

could you explain me a thing? From your words I understand that CFB works only with fixed bias, because you're grounding the cathode from a DC point of view. It's possibile to use cathode bias?

Diagram of what I've understand attached (sorry for using Paint):
 

Attachments

  • immagine.gif
    immagine.gif
    7.9 KB · Views: 2,296
Ex-Moderator
Joined 2003
You can use cathode feedback with cathode bias, but rather than putting the cathode resistor below the winding (as in your drawing) put it above. You now have one of your loudspeaker terminals grounded. There will be a small DC offset at the other loudspeaker terminal due to the DC voltage drop across the transformer's secondary resistance, but it's not usually a problem. You almost certainly want to bypass the cathode resistor with a capacitor.
 
Thank you very much.

And what about a PP output stage (fixed or cathode bias I see it's unimportant)? If I use a common cathode resistor, can I put the secondary between that resistor and ground? What sonic effects should I have? I'm expecting that it cannot work, since I ideally need 2 separate out-of-phase secondaries, one for each "side" of the PP tubes (if only 2 tubes, one for each tube). Maybe using a 0-8-16ohm speaker taps, you can ground the 8ohm one, use the 0 as the positive for you speakers and use the 0 and 16 as two separate cathode windings.

But this way you'll have speaker back EMF feed back to only one of the output tubes, or am I wrong?
 
EC8010 is right, just put the resistor (with a cap across it) between the cathode and the transformer. Many pentodes are terribly non linear in the region near positive grid bias, so PowerDrive may not be a serious advantage here. I am using it in the current prototype because I wired 6AV5's into my 300B amp. I will do a lot more testing when I return from an upcomming trip (mid April).

If you try CFB with push pull you would ground the 4 ohm tap, and use the 0 and 16 ohm taps for the cathode connections. In a transformer the impedance ratio is the square of the turns ratio, so the 4 ohm tap is the CT of the 0 to 16 ohm winding. Most P-P transformers wind the 4 ohm winding with heavier wire than the 8 and 16 ohm windings, so they have different DC resistances. This is often the case with the 2 halves of the primary as well. Many P-P transformers have different amounts of inductance in each winding. This causes different amounts of feedback to each tube, and the whole effect varies with frequency. The tube to tube balance that you are trying to achieve gets upset with imbalanced feedback.

Audio Research made P-P amplifiers that used CFB successfully. The schematics for their amps can still be found here:

http://www.arcdb.ws/

Look at the schematic for the D76. This idea obviously works well with their transformers. I tried it with several generic transformers, and the results were not pretty.
 
Two quick questions. Why put the cathode resistor in series with the speaker? It's standard practice but currently my amp has the speaker in series with the cathode bypass cap and leaves the cathode resistor from cathode to ground. No DC through the speaker and a more direct connection between speaker and cathode. I realize this places the cathode resistor in parallel with the speaker, in the case of this EL84 though it's 270 ohms and I can make an argument it has a benefit in midly taming speakers with wild impedance variations. It measures well, any idea why it's not used?

The other question relates to a Rowe/AMI jukebox amp I bought for iron. The P-P OPTs have 70 volt windings, anyone every try using something similar for cathode feedback? Thx!
 
Ex-Moderator
Joined 2003
rdf said:
It measures well, any idea why it's not used?

How about, "Because nobody else thought of it"? Good idea.

As Tubelab says, cathode feedback in push-pull circuits does rather assume symmetry in the output transformer, and that's not usually the case. However, it costs nothing to try it, and sometimes the results are well worthwhile. In my experience, you trade HF ringing and stability against distortion. 3-6dB of cfb seems to work well, but more than that can be a problem. Further, the lower mu valves don't seem to respond well to cfb (even if the windings have been adjusted to give 6dB of cfb). If you've got a 70V winding, try it and see...
 
Thx EC8010, there's just no way I'ld ever believe that in field as throughly plowed as tube circuits I came up with anything combining originality and value. Easier to assume the technique saw use and was abandoned in the distant past for perfomance or economic reasons. You mentioned before low mu tubes caused more problems, did you ever determine the cause? Or were you applying large amounts of cathode feedback via extra windings? It just seems somewhat counterintuitive a circuit with less open loop gain could be more sensitive to the application of moderate feedback. My transformers (Rowe R-4359A amp) were meant for 7868, no ultralinear winding, so eventually I'll try JJ 7591A both pentode and triode mode using the 70 volt taps for cathode feedback to hammer down the output impedance. I'm canvassing as much prior experience as possible to avoid releasing magic $$$moke beforehand though.

Yup Jane, almost exactly like that. Significantly different front end and bias point for the EL84. I see that design uses 470uF for the cahtode cap. It's the one potential 'gotcha' I found. It's possible to achieve decpetively flat response into a load resisitor by tailoring the cap with small values at the expense of seriously reduced damping factor. My amp was dead flat to 20 Hz using 22uF, at the same frequency the DF was under 1. At least now I know what that sounds like. Currently it's 220uF but I've yet to remeasure the Zout.
 
Ex-Moderator
Joined 2003
Hello rdf, no, I'm afraid I didn't find any explanation as to why low mu valves didn't seem to benefit from cfb as much as higher mu valves. I have direct experience with EL84 and N78 (triode and pentode), 6S4A, 2A3, all in push-pull, and 6528 in single-ended, also anecdotal of triode-strapped 813. The measured (and listening) results of adding cfb to push-pull UL N78 were that it was an absolute winner. 6528 and 813 were well worthwhile. 6S4A and EL84 weren't bad, but not really worth the degradation of 10kHz square wave response. However, 2A3, with a specially wound symmetrical transformer to give 6dB cfb, was a disaster and if it hadn't been for the reduced gain and output resistance, I'd have thought I applied the feedback incorrectly (I even swapped it to make sure). Wherever possible, I used more than one type of output transformer in my tests.
 
Thx EC8010. I wasn't familiar with the N78. Wow, what a nice tube! I can understand why it responded so well to cathode feedback. Very linear and the Marconi spec sheet shows 4 watts out (about 5.5 volt 0-peak into 8 ohms) with 5 volts swing at the grid in pentode. That's a pretty good return. Surprised about the EL84 though. I used the technique on the EL84 family in triode with two different transformers. The first a James SE in fixed bias, currently a Hammond 1628se as shown in Jane's schematic. It worked well both times but the James exceptionally so sonically. I really can't say much about square wave response, the Hammond's a defective design and the James donating parts for a KT100 (or 814) SE so no longer available for testing. I recall the James rings pretty good at 60-80 kHz with an EL84 in any case.
 
tubelab.com said:
Most P-P transformers wind the 4 ohm winding with heavier wire than the 8 and 16 ohm windings, so they have different DC resistances. This is often the case with the 2 halves of the primary as well. Many P-P transformers have different amounts of inductance in each winding. This causes different amounts of feedback to each tube, and the whole effect varies with frequency. The tube to tube balance that you are trying to achieve gets upset with imbalanced feedback.

Tubelab,
I have some difficulty here. I do not see how different dc resistances in windings will influence the degree of feedback, which is dependant on winding ratio (apart from secondary issues like capacitance, but I do not think you bring that in here). Likewise it is unclear how P-P transformers can have differing inductances in each winding, if the turns ratio is equal (as surely it is supposed to be). I have never found that, although I have probably measured fewer transformers than you. Lastly, will unbalanced feedback not tend to minimise the influence of tube differences rather than upset it? (I take it you are talking generally here).

As an aside, are there output transformers having a physical tap on a 16 ohm winding for 4 ohms, as compared to bringing out the 2 equal windings and wiring either in serie or in parallel? What a waste and unbalanced arrangement!

Then a remark about using tertiary windings for feedback purposes (especially global): If these are unloaded, leakage reactance plays a very diminished role, making higher degrees of NFB possible without stability problems (depending of course on transformer design). Several such circuits existed long ago, making a pentode output stage almost equivalent to a triode one. A 70V winding might serve this purpose.

The drop in frequency response at the loudspeaker winding itself will still be there as this is not included in the feedback loop, but the total effect appeared to have had an advantage over feedback including the (loaded) loudpseaker winding. Excellent square wave response was displayed.

Regards.
 
The different DC resistances are just another symptom of the underlying problem. Low cost output transformers are often wound in this fashion. One half of the primary goes on first, then the secondary, then the second half of the primary. Better quality transformers will interleave the secondary into the two primaries, but often the two primary halves are wound consecutively. This leads to differences in inductance and distributed capacitance between the two halves. I have taken several transformers appart and this seems to be a common practice. Since the winding that goes on last has a longer winding path, there is more wire in that section. The DC resistance is one clue that this is the case, however I have seen many transformers where this (longer) winding is wound with thicker wire.

In the low cost transformers that I have examined the secondary is one continuous winding with taps for 4 and 8 ohms. Often each section is wound with progressively smaller wire. Since the 0 to 4 ohm section carries the most current is is wound with the thickest wire. The wire is spliced to a thinner wire, and the winding continues. The splice is brought out to form the 4 ohm tap. The winding is spliced again to an even smaller wire and this splice forms the 8 ohm tap. The (thinner) winding continues until the end of the secondary which is the 16 ohm tap. Yes this is unbalanced, but it is common practice.

Transformers wound in this fashion do not work well for CFB using a circuit similar to the Audio research designs mentioned earlier. I have a large quantity of P-P transformers that are wound as described above. They work reasonable well in standard P-P circuits with no feedback (my 300B P-P amp for example) or in circuits that use a single feedback loop from the amp output to the driver or input stage. Any attempts at balanced feedback results in lumpy frequency response above 5KHz and they sound bad.

I don't know how the unbalanced transformer reacts with the unbalance caused by the tubes, but I tried several different transformers in this amp (P-P 6L6) , and found that each one behaves differently. This experiment wes done several years ago, so I don't remember all of the little details. I remember that the only transformers that I could get to work were some old UTC's and some mains toroids. I don't have many P-P transformers to try though.

There are good transformers with bifilar primaries, and multiple secondaries that can be interconnected to get the desired impedance ratio. These tend to cost more than I am willing to spend though.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.