• 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.

Negative Feedback Amplifier with embedded Positive Feedback Booster

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Disabled Account
Joined 2013
Dear DIYer, Friends,

I wish to share this OTL amp with you all. It uses low gain tubes and embedded positive feedback.
For more theory you can read this article, this amp can fit in the mechanism descript here:​
Ref: POSITIVE FEEDBACK GAIN-ENHANCEMENT TECHNIQUES FOR AMPLIFIER DESIGN


It is possible to use many different types of low gain tubes include Triode connection Pentode which I have not tried. As for now my best listening result is 5687 and 6H9C/6SN7 and 8*6C19p output tube. It 's also possible to build smaller version as headphones amp with reduced driver voltages. 2 output tubes. There are too many tasks for one person to do this is why I want to share it instead. If you know SPICE, you input the information to see what happened, many would perhaps like to know more than what I already being observed. If you wanted more info about the driver PCB I will send the one that I use and sound clip if possible.



Let me know what you think, thank you in advance.





 
AX tech editor
Joined 2002
Paid Member
It is a useful and effective technique. I have used it in solid state amps with success.
There is one thing you need to be aware of. The effective feedback is still negative, otherwise it would not be stable. But depending on your circuit, you can get the situation that when the amp clips, the nfb becomes nonoperative, while the pos feedback still works. In such a case the amp hangs on to one of the power rails and that's the end of your speakers (and possibly the amp).

So you must make sure that even when clipping, the effective fb remains negative.

Jan
 
Disabled Account
Joined 2013
Dear Jan,

Thank you very much for the good comment:clipping is one thing I did not mention in my presentation. But I just do a quick test just now the clipping is just gradual squaring of top and bottom waveform, nothing else happened. I have used the amp for about 3 years it still going...as to what you said, it is certainly true to some extend.
As I mentioned even with total loss of negative feedback (by simply removed the NFB link), the amp would just went into oscillation (motor boating, square wave) and survive many times. I have speaker protection circuit in the output, response time about 1ms. The DC you mentioned is not really a big problem, one way or other you can have DC in the output though, especially serious when direct coupled and solid state. It CAN be protected by good design as you pointed out, thank again.
 
Thanks for posting this patent and the other links and info. The positive feedback with phase correcting to match the negative feedback is most interesting.

There is a later Wolcott patent or two, 3328711 and 3111630, with a similar motif. A thread on it here:
http://www.diyaudio.com/forums/tubes-valves/58352-wollcott-cross-coupled-circuit.html

The negative impedance (current sensing output tap) feature in your design can also be applied to the primary side of a conventional OT'd amp to remove the fixed winding resistance.
 

Attachments

  • Miller_Pos_Fdbk1.bmp
    128.5 KB · Views: 64
  • Wollcott-1.gif
    Wollcott-1.gif
    21.4 KB · Views: 423
Last edited:
Disabled Account
Joined 2013
Yes, but my OTL is asymmetrical totem pole, do you think the PF in my design also provides additional function balance equalization of top and bottom tube. I did say something in that respect in my presentation, but I hope to hear some clear explanation, remember there were no bootstrap or any other mechanism involved.

(I have speaker protection in output, it hardly got activated except critical PSU failure like +ve or -v rail blown fuse, the fuse is 1A, it is 25W hardly blown. Little or no DC at full load, as well observed from speaker cone. With the same gain and NFB only in another design , it it quite different story.)
 
Looks like the driver circuitry in your design is providing equal drive voltages to both the output tube grids. Usually there is a feedback bootstrap (from the output) to one of the V2 driver plates, or cathodes (to the load resistors) to compensate the drive levels for the cathode follower and grounded cathode output tube drives.

The VR4 adjustment looks like it would allow adjusting the relative output impedances of the top versus bottom output tubes. So that could compensate some, but would depend on the load Z. The overall NFB of course tries to equalize the output swings. I would guess that VR4 needs to be re-adjusted if the speaker Z is changed. How high is the 2nd harmonic for this thing?

OTL Bootstrap driver from Tubecad site:
 

Attachments

  • OTL_C.gif
    OTL_C.gif
    6.9 KB · Views: 397
Last edited:
Disabled Account
Joined 2013
Thank you for prompt reply. 2nd harmonic is fairly high from waveform observed estimated 2%. I got better symmetry when critical damping is approached (adjust VR1) i.e when the output neither raised or lowered when load is applied for both the top and bottom waveform. I guess that is the point where 2nd harmonic is smallest. Does this mean anything?

Yes VR4 is adjusted for best symmetry I would keep it fixed unless symmetry is unobtainable using damping control VR1,. Thanks, that remind me of speaker Z change I guess the output waveform needs to be observed or 2nd harmonic measured first. My HP audio spectrum broken down years ago. I bought another one but no working yet. I only got THD meter and scope now. Upper tubes clipped earlier than the bottom tubes though. I will make change to VR4 later to see what else happened.

For now I believe this is bootstrap+ i.e bootstrap with gain greater than unity, please comment as will, thank.
 
Last edited:
Disabled Account
Joined 2013
.....
The VR4 adjustment looks like it would allow adjusting the relative output impedance of the top versus bottom output tubes. So that could compensate some, but would depend on the load Z. ... :

I recalled now at about 15W during the test, there is little waveform difference s between 8 ohms Z and internal Z about 2K, so VR4 is not adjusted for Z change. VR4 initially I intended for gain variation in output tube, to enable clipping the same time for top and bottom waveforms. The amp is stable with no termination at output and input except internal value. However I still need to works on the front end stage to improve symmetry. For driver stage I have load resistors change to 1/2 of 33K about 16K, making sound faster and sharper.
 
Disabled Account
Joined 2013
Today I work on the front end stage to improve symmetry. I double one of V1 cathode value of R26 to about 7K and settled on 10K. This reduced shunt loss to ground for the positive feedback drive thus restored the correct drive for the lower bank of output tube. V1 cathode has raised and plate voltage is 90V. The waveform now appears more symmetry estimate d<0.5% 2nd harmonic distortion at full output. Perhaps this not ideal route for injecting PF, perhaps inject into grid is better choice without too much loss but that is best I can do at this moment.

Clipping is now at same time for top and bottom output wave form. I also take a look at the driver voltage during clipping. While the output waveform is sharply squared , the driver output waveform is gradual, with rounding (a shallow arc) across the top and bottom. I do not quite fully understand the full recovery process at this moment but there is nothing else happened . If you do please comment.
 
Last edited:
I'm puzzled about your doubling the value of R18 to increase the grid drive of the bottom output tube. One would expect the upper tube (in cathode follower mode) to need the additional grid drive instead. Is the diagram correct at W4/S4 for straight through connections?

It seems to me that you are using the net feedback to correct the inherent symmetry of the output grid drives (to get the required drive asymmetry). (V3 needing more grid drive in CF mode) Unbalancing VR4 unbalances the negative feedback versus output polarity current. (however, giving unbalanced N feedback, versus polarity, ONLY when there is a load current. ie. R34, R35) Presumably the neg. feedback loop tries to reduce the relative output from the bottom V2 (for driving V4) and increase the relative output of the top V2 for driving V3 to maintain output symmetry. That then reduces the positive feedback (at VR1) when V2 plate signal is reduced during V4 conduction, and increases PF during V3 conduction. So the NET neg. feedback is varying in a way to enforce the symmetry restoration.

So, if I'm understanding this correctly, the NET negative feedback will always be trying to restore symmetry, but it will have MORE loop gain during V3 conduction. (Pos Fdbk signal from bottom V2 plate larger during V3 conduction)
I think this means that V3 will get its drive boosted relatively more, than V2 will get its drive reduced (during applicable drive phases, when VR4 is off center, and output loaded). Getting this process tuned to a high level of output symmetry sounds like it would be quite sensitive to the various potentiometer adjustments.

This is certainly not the conventional solution. But it may be workable with some tweaking (specifiic to a certain load Z).

I assume you are familiar with the conventional solution for totem pole (asymmetric) drives (to get symmetric outputs). See diagram in post #10.
The bootstrapping of one driver load resistor off the output means the output grid drive signal derived for the top follower stage is relative to its cathode, so it gets the SAME grid-cathode drive signal as seen at the bottom grounded cathode stage. Basically, fixing the asymmetry, BEFORE the neg. feedback is applied. This scheme likely works much better with pentode drivers, so the voltage offset at the one bootstrapped driver plate does not influence the tube gain. Its a fairly elegant solution. One could also use an interstage xfmr to drive the top follower relative to its cathode.

A triode driver however might give more 2nd harmonic output (in the bootstrap solution), due to the bootstrap effect on plate Rp, if thats desired.
 
Last edited:
Since the bootstrapped driver load in the conventional solution can be viewed as positive feedback itself, it might clarify things to contrast the two approaches. (at least how I see them presently) The conventional solution is using positive feedback ONLY for the top driver/follower output to symmetrize the final output, and is a load current independent fix. (only using voltages) (the top follower needing more drive signal than the bottom grounded cathode output) Neg. feedback is then applied conventionally to the full amplifier.

Your solution is using positive feedback (in addition to the usual global neg. Fdbk) for BOTH sides (top CF, and bottom GC outputs), but VR4 allows the NET feedback (actually affecting the Neg. Fdbk path directly) to be dynamically varied versus output conduction phase. So in theory I suspect these could achieve the same effect, except the VR4 gain shift effect is load current sensitive. (ie, the NET feedback is effectively dynamically offset between the + and - output phases, similar in effect to the fixed built-in NET feedback offset in the conventional solution)

Unless I'm reading your circuit wrong somewhere.
 
Last edited:
Disabled Account
Joined 2013
First I am sorry to confuse you , actually VR4 is not use currently and it is intended for compensating of output tube if I have to use it. So that suppose that the top bank has a lot more gain than lower one, the output would not be symmetry and upper tube will clip earlier esp, at high load and adjusting VR4 can correct it to some extent. to make thing less complicated, I use matched tubes VR4 is dispensable for now. So discussion is based on one without it, I appreciate your view on this. You can call it DC balance if you like.

"Is the diagram correct at W4/S4 for straight through connections? "
Yes, it is correct. You can trace the small phase diagram, whether they are not all correct from input to output.

"I'm puzzled about your doubling the value of R18 to increase the grid drive of the bottom output tube. One would expect the upper tube (in cathode follower mode) to need the additional grid drive instead. "
Yes, that what is expected, the amp is no exception that the top tube required more gain than bottom tube. This is the result of positive feedback loop it produces much more drive for the top tube. This partly due to shut loss at V1 cathode, due to lower gain of V1. If V1 is high gain say 12AT7, no change is necessary. So gain of V1 plays a role, change of R18 is to compensate for the loss. Even so my previous data (in the slice) shown that for 16V output the top tube grid required 80V and bottom 50V drive to confirm that. Previously ECC99 is used and it has more gain than 5687 now in used. The change may produced more gain than necessary but I think the amp got balanced out by NFB loop. But still I don't know whether this the sole cause of loss in bottom drive .
I like to leave the question opened for discussion (excluding VR4), and whether the top drive actually increased more than bottom as result of PF (may assume that no loss PF and tube gain)

"I assume you are familiar with the conventional solution for totem pole (asymmetric) drives"
Yes I am the one who posted the question to Editor TubeCad and he made changes of drawing you attached earlier. I have experimented nearly all bootstrap methods, include inverted Futherman, still have about 5-6 amp in the storeroom, they do not sound like this one. They use high gain tube, sound is not as open, and low Dumping factor using low gain tube. The bootstrap can not provide additional or enough gain to improve further.

I will pause here before I digest your other comment and reply soon.
 
Last edited:
Disabled Account
Joined 2013
I would like to point out some some differences in conventional positive feedback with gain

1) Positive feedback is mixed with input signal at grid input. NFB is at another place. This is not pure positive feedback inside NFB loop as it is not embedded inside NFB loop. PF is purely to increase gain but NFB can not have full control.

2) Positive feedback is injected in the phase splitter thereby produce summery drive out for top and bottom tube, this is ideal for symmetry PP application as they are push pull symmetric output. Therefor NFB is not interact directly so as not to cause any unbalance drive.

3) In this amp the positive feedback and negative feedback in one place as unbalance d drive is desired the negative feedback can have control over PF level during top and bottom drive. This is to similar to Global NFB, but the magnitude is amplifier or boosted in addition to original NFB. The initial gain is very low to begin with, about 70 = 7 * 10, gain expanded to maximum 450 at full load. In conventional NFB initial gain is 450 or much more to begin with and gain reduced to 70 when closed. Because maximum gain can not be exceeded, therefore this helps recovery a lot during clipping.

Do you see the differences now?
 
Last edited:
Disabled Account
Joined 2013
Another thing about the positive feedback is that its opened loop gain can not be measured directly. Because it just went into oscillation and circuit at oscillation stage is not as same normal open loop circuit. Therefore the gain of individual gain of each stage has to be measured separately. E.g. with only V1 in circuit the gain is measured about 7, then with only V2 in circuit the gain is again measured 10. Therefore the initial gain is 7*10 ~70. When in closed loop NFB takes over, the circuit is stabilized (outside of oscillation now) . The over gain is still same or lower (depends of NFB level) without signal input as there is no output at V2 therefore no PF, the gain increased is zero. PF will increase gain only when there is input sign and attained certain maximum gain in V1 only, say 45. Therefore the maximum gain is 45* 10 =~450, initial gain is 70.

What do you think?
 
Last edited:
Disabled Account
Joined 2013
The other thing I would like to look at is crossover distortion. You can read what happened Here. It is not terribly important that it is elimated, but rather is an indication of good or ideal implementation, I don't use a simple bootstrap probably I think it is major culprit causing crossover distortion, despite large amout of bias in output tube, the conducting angle is just not right esp. at higher frequency. Because I have done those model before I am awared of this problem, I notice this amp approach has made crossover distortion to disappear for certain reasons. I avoided bootstrap in the output because it (gain) causes input and output impedance to change, and can not be easily maintained at higher frequency or at different output level. So if the input and output impedance becomes variable hence the conduction angles. I choose to avoid it in the output stage.

What do you think?
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.