If the plate NFBKs couple back to the 12AX7 (or some other input driver tube choice) cathodes (use un-bypassed cathode degen. resistors there) instead of the input stage plates, the impedance issue will work out better. (But will require more grid input voltage swing.) I would also go with Miles' Mosfet or tube source/cath. followers to drive the outputs. The 12AX7 plate loads can then be cap./bootstrapped from these followers too for better linearity and gain. (This gets fairly close to the Wollcott design then.)
The incusion of an extra driver stage with cross coupled plate NFBKs mentioned by Ray is another good way (used in the Citation II).
My one quibble with this cross-coupled approach is that the output transformer inversion is IN the loop, and for class AB outputs, may cause some stability problems. (The Citation II used a very good OT xfmr. with its cross coupling approach)
Don
The incusion of an extra driver stage with cross coupled plate NFBKs mentioned by Ray is another good way (used in the Citation II).
My one quibble with this cross-coupled approach is that the output transformer inversion is IN the loop, and for class AB outputs, may cause some stability problems. (The Citation II used a very good OT xfmr. with its cross coupling approach)
Don
The ECC81 is a good choice for driver valve and is used in the RH807 and RH84 design. There are very few drivers which will do this job well, and the choice of ECC83/12ax7 is probably a very bad one.
The RH807 designs get around many of the issues encountered with this concept by having seperate anode load resistors and feedback resistors. This simplifies the design approach no end. Combining the two roles makes the job hugely more tricky.
I would say that you are trying to reinvent the wheel when there is a very good workable design already out there. All you need to do is change the driver tube.
Shoog
The RH807 designs get around many of the issues encountered with this concept by having seperate anode load resistors and feedback resistors. This simplifies the design approach no end. Combining the two roles makes the job hugely more tricky.
I would say that you are trying to reinvent the wheel when there is a very good workable design already out there. All you need to do is change the driver tube.
Shoog
Hi guys,
You instilated doubts in my mind, so I bring the amp back on the bench and spend the afternoon doing some measurements.
This schematic:
http://www.dissident-audio.com/PP_ECL86/Schema.gif
http://www.dissident-audio.com/PP_ECL86/THD0dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-3dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-6dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-9dB.pdf
http://www.dissident-audio.com/PP_ECL86/BWandDF.pdf
http://www.dissident-audio.com/PP_ECL86/Square_20Hz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_40Hz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_1KHz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_10KHz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_50KHz.pdf
The OPT cannot pass the 20Hz square at full level (see scales).
I know that the measures alone can't tell the whole story, but they may help.
Yves.
You instilated doubts in my mind, so I bring the amp back on the bench and spend the afternoon doing some measurements.
This schematic:
http://www.dissident-audio.com/PP_ECL86/Schema.gif
http://www.dissident-audio.com/PP_ECL86/THD0dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-3dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-6dB.pdf
http://www.dissident-audio.com/PP_ECL86/THD-9dB.pdf
http://www.dissident-audio.com/PP_ECL86/BWandDF.pdf
http://www.dissident-audio.com/PP_ECL86/Square_20Hz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_40Hz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_1KHz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_10KHz.pdf
http://www.dissident-audio.com/PP_ECL86/Square_50KHz.pdf
The OPT cannot pass the 20Hz square at full level (see scales).
I know that the measures alone can't tell the whole story, but they may help.
Yves.
hey-Hey!!!,
As noted, the partial FB arrangement works as a voltage to current stage up front. Instead of using separate loads and FB resistors, pick a suitable fraction of voltage FB from the finals and eliminate a B+ stage by connecting the plate loads to the taps on the primary of the OPT.
See Pete Millett's description of a SE E-Linear amp. He did it SE, and roughly in parallel( I think about a month two ahead ), I did it for PP. Unfortunately I don't find the typical U-L percentage useful. So we need a custom output TX with taps located to take advantage of the circuit.
Pentodes and triode/MOSFET cascodes do quite well. Near-flat/high plate resistance constructions are the Way...
cheers,
Douglas
As noted, the partial FB arrangement works as a voltage to current stage up front. Instead of using separate loads and FB resistors, pick a suitable fraction of voltage FB from the finals and eliminate a B+ stage by connecting the plate loads to the taps on the primary of the OPT.
See Pete Millett's description of a SE E-Linear amp. He did it SE, and roughly in parallel( I think about a month two ahead ), I did it for PP. Unfortunately I don't find the typical U-L percentage useful. So we need a custom output TX with taps located to take advantage of the circuit.
Pentodes and triode/MOSFET cascodes do quite well. Near-flat/high plate resistance constructions are the Way...
cheers,
Douglas
Bandersnatch said:. . .
Pentodes and triode/MOSFET cascodes do quite well. Near-flat/high plate resistance constructions are the Way...
cheers,
Douglas
My goal was to use ECL86/6GW8, that's why !
But I agree that penthodes are a better choice:
The beefiest ad fastest driver I've ever built for a big DH triode.
Yves.
Yvesm said:
I never doubted your original measurements as published on the schematic.
Some measurements on published designs are always appreciated - even if they do not tell the whole story.
Your good measurements may seem to be a result of carefule balancing of modest inner feedback, and a little more of global feedback.
My goal was to try if the local "partial" feedback could reduce the output impedance to a level where global feedback could be an option. In theory it should be possible to acheive triode like output impedance with EL84 anode-grid feedback.
My findings so far are however that with 12AX7 as driver, the local feedback does more to degenerate the output of the driver tube than it helps the EL84.
SveinB.
I would use one twin opamp TL072 to drive output toobs directly; they can drive very low input impedance so any OPT would be greatly linearized due to deep local feedback. However, tubes must have high transconductance, like EL32, EL84, or even higher, to demand less output voltage swing from opamps.
Why?
Because anyway there are lot of opamps in signal path from studio microphones to output amplifiers.
As I said before, this "partial feedback" is a naming nonsense: actually it is a parallel feedback by voltage. What is good in it, feedback and source signals are applied to the same point, so non-linearities between grid and cathode like in case of a serial feedback when feedback itself adds distortions are eliminated from the equation.
Why?
Because anyway there are lot of opamps in signal path from studio microphones to output amplifiers.
As I said before, this "partial feedback" is a naming nonsense: actually it is a parallel feedback by voltage. What is good in it, feedback and source signals are applied to the same point, so non-linearities between grid and cathode like in case of a serial feedback when feedback itself adds distortions are eliminated from the equation.
"I would use one twin opamp TL072 to drive output toobs directly"
You don't have to be restricted by the voltage compliance of op amps anymore, see:
http://www.diyaudio.com/forums/showthread.php?threadid=113444
+/- 100 V rails for LM4702 and LME49810 (but are a bit pricey yet)
Anatoliy, you make a good point about applying the feedback to the same point as the input signal for low distortion. The cross coupled (Citation II) scheme with an extra driver stage does this. Also, the (non cross coupled) Wollcott like approach can be made to do this too, if the feedbacks AND input signals are connected to the driver stage cathodes.
But for pure convenience, the nicest feedback point is to the driver screen grids, same voltage level as the output plates and it provides additional attenuation of the feedbacks for free. I don't think I have seen this done, but it sure looks enticing.
Don
You don't have to be restricted by the voltage compliance of op amps anymore, see:
http://www.diyaudio.com/forums/showthread.php?threadid=113444
+/- 100 V rails for LM4702 and LME49810 (but are a bit pricey yet)
Anatoliy, you make a good point about applying the feedback to the same point as the input signal for low distortion. The cross coupled (Citation II) scheme with an extra driver stage does this. Also, the (non cross coupled) Wollcott like approach can be made to do this too, if the feedbacks AND input signals are connected to the driver stage cathodes.
But for pure convenience, the nicest feedback point is to the driver screen grids, same voltage level as the output plates and it provides additional attenuation of the feedbacks for free. I don't think I have seen this done, but it sure looks enticing.
Don
Guys,
For LM4702 use in a power amp see AudioExpress Jan 2008 in your shops right now.
As far as my experience with the Baby Huey variation of Yves design goes I can only offer a few hints.
About 4 to 5 dB of local feedback is all you can achieve. More than that demands too much current swing in the 12AX7 diffamp. Thats why I used it in conjunction with Ultralinear connected EL84 and even then I tossed up whether to add global feedback - in the end I ran without global feedabck but output impedance is a little high.
Some output impedance measurements were posted to the Baby Huey thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=72536&perpage=25&pagenumber=6
I've never done distortion measurements (I don't have the gear) all my tweaks and adjustments were done by ear.
Most recently I have found that zobel networks on the secondary of the output transformers helped a lot - probably by linearising (with frequency) the load impedance and hence the voltage being developed at the output tube anodes (the local shunt feedback source).
It is the simplicity and the minimal power supply requirements which attracted me to this circuit arrangement. I originally built Yves ECL86 design with cathode biasing of the outputs, ultralinear connection and a cascode bipolar transistor current source on the diff amp. That amp (which is now in the home of a Opera Buff, running her Polk R5s) was so lovely I decided to try a 12AX7 EL84 version - the prototype Baby Huey. It started life as a Morgan Jones Bevois Valley ECC88 EL84 and I found that I prefered the sound of the Baby Huey.
It is not beyond the realms of possibility that that just means that I like the sound of the particular distortion profile it produces. As I said above I've never run distortion measurements. To be honest, I'm not sure that I believe in them all that much - the power amp I absolutely detested the most in my system was a 210W per channel ROTEL with 0.001% distortion specification. It was that piece of "!@##$$%" that got me started into tube amps.
Cheers,
Ian
For LM4702 use in a power amp see AudioExpress Jan 2008 in your shops right now.
As far as my experience with the Baby Huey variation of Yves design goes I can only offer a few hints.
About 4 to 5 dB of local feedback is all you can achieve. More than that demands too much current swing in the 12AX7 diffamp. Thats why I used it in conjunction with Ultralinear connected EL84 and even then I tossed up whether to add global feedback - in the end I ran without global feedabck but output impedance is a little high.
Some output impedance measurements were posted to the Baby Huey thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=72536&perpage=25&pagenumber=6
I've never done distortion measurements (I don't have the gear) all my tweaks and adjustments were done by ear.
Most recently I have found that zobel networks on the secondary of the output transformers helped a lot - probably by linearising (with frequency) the load impedance and hence the voltage being developed at the output tube anodes (the local shunt feedback source).
It is the simplicity and the minimal power supply requirements which attracted me to this circuit arrangement. I originally built Yves ECL86 design with cathode biasing of the outputs, ultralinear connection and a cascode bipolar transistor current source on the diff amp. That amp (which is now in the home of a Opera Buff, running her Polk R5s) was so lovely I decided to try a 12AX7 EL84 version - the prototype Baby Huey. It started life as a Morgan Jones Bevois Valley ECC88 EL84 and I found that I prefered the sound of the Baby Huey.
It is not beyond the realms of possibility that that just means that I like the sound of the particular distortion profile it produces. As I said above I've never run distortion measurements. To be honest, I'm not sure that I believe in them all that much - the power amp I absolutely detested the most in my system was a 210W per channel ROTEL with 0.001% distortion specification. It was that piece of "!@##$$%" that got me started into tube amps.
Cheers,
Ian
Don;
I saw such beasts designed to drive power transistors, but TL072s are freely available in my bin.
Also, a discrete transistor amps would work great here, like this one:
I myself made inverting power amps that had line input transformers (input impedance was 600 Ohm), and signals from transformers and GNFB were summed of the grid of the 1'st lamp. It sounded very nice! But I did not try to linearize output toobs and transformers by such a way...
Using screen grids you'll get back that well known "Ultra Linear" approach, or I misunderstood you?
I saw such beasts designed to drive power transistors, but TL072s are freely available in my bin.
Also, a discrete transistor amps would work great here, like this one:
I myself made inverting power amps that had line input transformers (input impedance was 600 Ohm), and signals from transformers and GNFB were summed of the grid of the 1'st lamp. It sounded very nice! But I did not try to linearize output toobs and transformers by such a way...
Using screen grids you'll get back that well known "Ultra Linear" approach, or I misunderstood you?
"Using screen grids you'll get back that well known "Ultra Linear" approach, or I misunderstood you?"
Well, I meant using the screen grids on the (pentode) driver tubes for the return point of the output tube plate feedback resistors (with an additional screen to screen resistor across the diffl. pair to set feedback attenuation and maybe some DC screen bias resistor to ground too for lower DC screen volts). I guess one could call it an ultralinear feedback driver. (also, the feedbacks now have to be crossed from outputs to drivers due to the inversion between grid and plate in the drivers.)
Don
Well, I meant using the screen grids on the (pentode) driver tubes for the return point of the output tube plate feedback resistors (with an additional screen to screen resistor across the diffl. pair to set feedback attenuation and maybe some DC screen bias resistor to ground too for lower DC screen volts). I guess one could call it an ultralinear feedback driver. (also, the feedbacks now have to be crossed from outputs to drivers due to the inversion between grid and plate in the drivers.)
Don
If we use solid state buffering, maybe a MosFet source-follower ("superdrive") as used in the Tubelab.SE could be adopted.
Probably not relevant or needed for DHT triodes, and not practcial for the EL34 family due to high drive requirements, but might be useful for something like 807 as well as the EL84.
(resistor values are arbitrary values, might be OK for EL84)
Tubelab.SE
Probably not relevant or needed for DHT triodes, and not practcial for the EL34 family due to high drive requirements, but might be useful for something like 807 as well as the EL84.
(resistor values are arbitrary values, might be OK for EL84)
Tubelab.SE
Attachments
Svein_B said:
Made some tests with and without "outer loop".
With outer loop, previous published measures showed an increase of 1dB ( times 1.1) when removing the 8 Ohm load.
This means that the internal impedance of the amp is some 0.8 Ohms.
With the "inner loop" alone (I agree that local FB is not the most appropriate term), this difference rises to 5 dB ( times 1.78) indicating an internal impedance of near 6 Ohms.
The OPT having a 1167 to 1 Z ratio, this means that the apparent Rp of both power tubes in serie is 1167 * 6 = 7K, thus each tube have a apparent Rp of 7/4 = 1K75.
So so ! The same order of magnitude than an EL84 in true triode mode but probably not low enough to eliminate the need of the outer loop.
Yves.
SveinB,
If your output transformer has Ultralinear taps there is one other scheme you can try.
Instead of connecting the 220K (180K in your case) diffamp anode loads to that 2 x 47K + 15K network off the output tube anodes you can delete that network and simply run the diffamp anode loads to the Ultralinear taps. That removes some of the self balancing action of the "Baby Huey" scheme BUT will allow you to use say 120K or 150K anode loads (which means the local shunt feedback voltage is not divided down by the rp as much) and push the diff amp current up at least by a factor of 2.
This MAY also allow the use of a little lower rp tube such as the 12AT7 depending upon how much local feedback you are prepared to sacrifice.
Cheers,
Ian
If your output transformer has Ultralinear taps there is one other scheme you can try.
Instead of connecting the 220K (180K in your case) diffamp anode loads to that 2 x 47K + 15K network off the output tube anodes you can delete that network and simply run the diffamp anode loads to the Ultralinear taps. That removes some of the self balancing action of the "Baby Huey" scheme BUT will allow you to use say 120K or 150K anode loads (which means the local shunt feedback voltage is not divided down by the rp as much) and push the diff amp current up at least by a factor of 2.
This MAY also allow the use of a little lower rp tube such as the 12AT7 depending upon how much local feedback you are prepared to sacrifice.
Cheers,
Ian
gingertube said:
hey-Hey!!!,
I have to say that I have not yet seen great results from E-Linear FB with a triode. Pentodes and cascodes are quite another story. One further advantage to the E-Linear design is it requires a single B+ supply; everything is fed through the OPT's center tap( save for SE ).
cheers,
Douglas
I had a basic / first priciples sort of think about the EL84 PP and 12AX7 diffamp fronr end version.
Now I may be confused in my thinking here (would'nr be the first time and won't be the last) but this is how I see it.
1st consider a +ve swing on the input signal, in that 12AX7 section
=> decreased g1-k
=> increased Ia
=> more signal than strictly linear (from Ia vs Vg slope)
=> reduced rp
=> reduced shunt feedback (divided more by the lower rp) which is really opposite of what we ideally want.
then consider -ve swing on input signal
=> increased g1-k
=> decreased Ia
=> less signal than strictly linear
=> increased rp
=> increased shunt feedback (divide less by the higher rp) which again is opposite of what we ideally want.
So YES it is an "increased" distortion design. What it is doing is emphasizing the distortion profile of the diff amp. In a well balanced diff amp ( which is why we recommend a cascode current source tail) the distortion profile is almost entirely 2nd harmonic.
My assertion is that we are adding in some 2nd harmonic which helps to balance the primarily 3rd harmonic of the PP outputs. That coupled with the moderate output impedance tends to give these amps an almost single ended type of sound. I think I alluded to that in the Baby Huey thread where I stated that it sounds very similar to my 845 SET.
If you want a minimal distortion design then you need an inverting driver stage after the diff amp just as Ray_Moth stated.
If you want a poor mans 300B sound alike (well a little like) then the simple PP EL84 and 12AX7 diff amp arrangement can be made to work and sound better than it apparently measures.
One last point - ALL of the variants of this amp that I have built, including the very first ECL86 "knock off" of Yves design have used Ultralinear connected outputs which gives a good head start in getting that output impedance under control.
Also I suggest that you have another look at some EL84 / 6BQ5 data sheets and application notes. It is the ONLY tube I have come across which has lower THD in Ultralinear Mode than in Triode or Pentode Mode.
Should I jump into my flame proof vest? Nah! - seriously, if my thinking is wrong then you do me a service by pointing it out.
Cheers,
Ian
Now I may be confused in my thinking here (would'nr be the first time and won't be the last) but this is how I see it.
1st consider a +ve swing on the input signal, in that 12AX7 section
=> decreased g1-k
=> increased Ia
=> more signal than strictly linear (from Ia vs Vg slope)
=> reduced rp
=> reduced shunt feedback (divided more by the lower rp) which is really opposite of what we ideally want.
then consider -ve swing on input signal
=> increased g1-k
=> decreased Ia
=> less signal than strictly linear
=> increased rp
=> increased shunt feedback (divide less by the higher rp) which again is opposite of what we ideally want.
So YES it is an "increased" distortion design. What it is doing is emphasizing the distortion profile of the diff amp. In a well balanced diff amp ( which is why we recommend a cascode current source tail) the distortion profile is almost entirely 2nd harmonic.
My assertion is that we are adding in some 2nd harmonic which helps to balance the primarily 3rd harmonic of the PP outputs. That coupled with the moderate output impedance tends to give these amps an almost single ended type of sound. I think I alluded to that in the Baby Huey thread where I stated that it sounds very similar to my 845 SET.
If you want a minimal distortion design then you need an inverting driver stage after the diff amp just as Ray_Moth stated.
If you want a poor mans 300B sound alike (well a little like) then the simple PP EL84 and 12AX7 diff amp arrangement can be made to work and sound better than it apparently measures.
One last point - ALL of the variants of this amp that I have built, including the very first ECL86 "knock off" of Yves design have used Ultralinear connected outputs which gives a good head start in getting that output impedance under control.
Also I suggest that you have another look at some EL84 / 6BQ5 data sheets and application notes. It is the ONLY tube I have come across which has lower THD in Ultralinear Mode than in Triode or Pentode Mode.
Should I jump into my flame proof vest? Nah! - seriously, if my thinking is wrong then you do me a service by pointing it out.
Cheers,
Ian
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.