Hello,
I'm planning to build an 807 beam tetrode amp, basically the RH807 (www.tubeaudio.8m.com) but with a different power supply, other bias point for the output stage (around 370V, 60mA) and regulated screen supply à la Gordon Rankins 807 SE (Sound Practices 1992). Rankin uses a single shunt regulator with 0d3 and 0c3 in series, shared by both channels. I would like to use a separate shunt for each channel (because the power supply is already split in two for B+ and B++). I'm not sure though if the current draw of one screengrid (or is it the shunt that draws the current?) is enough for good regulation by the VRs. Rankin writes the shunt draws 20mA, my output stage draws the same current as his so both of my shunts should draw 10mA. Can anyone confirm this guesstimate? It would be a pity to do all this extra work for a good screen supply when in the end the regulators wouldn't draw enough current to do a good job. The VRs I'll be using are good from 5mA and up.
Cheers,
Simon
I'm planning to build an 807 beam tetrode amp, basically the RH807 (www.tubeaudio.8m.com) but with a different power supply, other bias point for the output stage (around 370V, 60mA) and regulated screen supply à la Gordon Rankins 807 SE (Sound Practices 1992). Rankin uses a single shunt regulator with 0d3 and 0c3 in series, shared by both channels. I would like to use a separate shunt for each channel (because the power supply is already split in two for B+ and B++). I'm not sure though if the current draw of one screengrid (or is it the shunt that draws the current?) is enough for good regulation by the VRs. Rankin writes the shunt draws 20mA, my output stage draws the same current as his so both of my shunts should draw 10mA. Can anyone confirm this guesstimate? It would be a pity to do all this extra work for a good screen supply when in the end the regulators wouldn't draw enough current to do a good job. The VRs I'll be using are good from 5mA and up.
Cheers,
Simon
Klimon,
Simple solution - you just change the value of the dropping resistor so that the total static current (that drawn by 807g2 while idling plus that drawn by the Vreg tubes) is about 25mA.
0D3 + 0C3 means that V reg = 250V
If your B+ = 370 your required voltage drop = 120V
Resistior value = 120/.025 = 4800R
4k7 should be close enough
As the g2 current demand increases on peaks, the current passing through the Vreg tubes will drop by the same amount while still maintaining the required voltage.
pm
Simple solution - you just change the value of the dropping resistor so that the total static current (that drawn by 807g2 while idling plus that drawn by the Vreg tubes) is about 25mA.
0D3 + 0C3 means that V reg = 250V
If your B+ = 370 your required voltage drop = 120V
Resistior value = 120/.025 = 4800R
4k7 should be close enough
As the g2 current demand increases on peaks, the current passing through the Vreg tubes will drop by the same amount while still maintaining the required voltage.
pm
Klimon said:
If you're going to be playing around with different Q-Points, better get this first: STC 807 Application Report
You could also use an active regulator. Anything would beat that unbypassed series dropping resistor the RH 807 uses. That's very bad.
The question is, will regulating the screen greatly improve the performance. Old Alex says in the original article that he heard very little difference between the triode mode and the pentode mode. He extensively simulated this circuit. The plate to plate feedback seems to be extremely effective in smoothing out anomilies in response.
Shoog
I think it will make a difference, a resistor-tied screen, whether it is triode or pentode connected, is a compromise. Alex didn't compare a shunt regulated screen supply with a conventional one. Only that test would be the relevant for this issue, not triode-pentode comparison.
Shoog, Arnoldc, mach1, Miles: thanks a lot for your input, I'm ready to take the plunge --- those VR tubes do look nice
Simon
Its not really possible to speculate to what extent Alex actually modelled different types of screen supply - he did tend to imply elsewhere that the setup he suggests was tried against more elegant screen supplies and it came out well.
Early on when I was putting together my parafeed version of the RH807 I tried a dedicated screen supply. I had to drop it eventually because of a number of factors. The short time I tried it the sound that was coming out was brighter. It was however working with a smaller amount of Plate to plate feedback so this issue may well have disappeared with more feedback applied.
Just though you may have been interested in that snippet of info.
Shoog
Early on when I was putting together my parafeed version of the RH807 I tried a dedicated screen supply. I had to drop it eventually because of a number of factors. The short time I tried it the sound that was coming out was brighter. It was however working with a smaller amount of Plate to plate feedback so this issue may well have disappeared with more feedback applied.
Just though you may have been interested in that snippet of info.
Shoog
Shoog said:
Given the single-ended, Class A1, topology, it may not make that much difference. However, with AB1, and a screen current that varies between 5.0mA at no signal and 16mA at max output, I find it difficult to believe that letting the screen voltage wander all over the place would be good for the sonics, even with local NFB. Still, with just a series dropping resistor, any pentode will start to display a remote cutoff type characteristic. Unless it's a low level stage where the current doesn't vary all that much, or something like a Class C unmodulated driver/frequency multiplier stage that draws a consistant current (and where distortion doesn't matter anyway) then I do like to nail down that screen voltage so that it doesn't wander. In the case of the RH 807 design, use a voltage divider, at least.
"Alex says in the original article that he heard very little difference between the triode mode and the pentode mode."
Well, that's the whole point behind local NFB: to make the dynamic plate characteristic more triode-like, and less pentode-like, as well as reducing the effective r(p) seen by the primary of the output xfmr.
"The plate to plate feedback seems to be extremely effective in smoothing out anomilies in response."
Not a part of my design philosophy. I want to minimize anomilies in response before the NFB (global or local) is connected. Include just enough NFB to improve an already good sounding design instead of covering up for your open loop design mistakes. If going the extra mile by including screen voltage regulation is part of that, then, so be it. I'm not designing for the "Big Box" people, and have no need to make the design as cheap as possible to produce.
I'll mail Aleksandar to ask him about his screen experiments. BTW: how does your rh807 parafeed perfom? Did you compare with the regular version?
I never compared it to the regular version. I did it this way because I had suitable toroidal transformers as outputs.
Mine sounds extremely good, with excellent bass response and all the SE characture you could want. I was running it with a 700V supply and a TT21(KT88) CCS plate load. At the moment it is in semi retirement as I built a 6080 PP amp with transformer phase splitter which out performs it by a long way. Also I just couldn't live with the terrible efficiency. I am currently thinking about rebuilding it with plate chokes, for efficiencies sake. I will also include a cathode CCS (LM317 based) and a few other improvements. I will probably bypass the screen supply myself as this becomes a more realistic option with the lower supply voltage.
Don't get me wrong on the screen supply - it is obviously sensible to do the best you can to get the best result, and common sense suggests that the results will benefit. My point is that its not always best to second guess the designer on these details. It would be an interesting experiment to build the shunt screen supply and then substitute a resistor for comparison. Wouldn't be difficult and would be educational for yourself if no one else. The output Tubes behaviour is completely transformed by the use of plate to plate feedback and it cannot be absolutely assumed that what works with a triode or pentode will work in a RH style circuit. Experimentation is required.
Shoog
Mine sounds extremely good, with excellent bass response and all the SE characture you could want. I was running it with a 700V supply and a TT21(KT88) CCS plate load. At the moment it is in semi retirement as I built a 6080 PP amp with transformer phase splitter which out performs it by a long way. Also I just couldn't live with the terrible efficiency. I am currently thinking about rebuilding it with plate chokes, for efficiencies sake. I will also include a cathode CCS (LM317 based) and a few other improvements. I will probably bypass the screen supply myself as this becomes a more realistic option with the lower supply voltage.
Don't get me wrong on the screen supply - it is obviously sensible to do the best you can to get the best result, and common sense suggests that the results will benefit. My point is that its not always best to second guess the designer on these details. It would be an interesting experiment to build the shunt screen supply and then substitute a resistor for comparison. Wouldn't be difficult and would be educational for yourself if no one else. The output Tubes behaviour is completely transformed by the use of plate to plate feedback and it cannot be absolutely assumed that what works with a triode or pentode will work in a RH style circuit. Experimentation is required.
Shoog
I agree; you can expect a little comparison when the time has come to build the amp (these things can take a while but it's the first project on my list
). I've recently swapped the resistor in my triode-strapped el84's to a bypassed zener-string with very good results; that's why I'm interested in a more 'refined' handling of the screen grids even if I only understand it on a very abstract level. Andrea Ciuffoli has built a kt88 with trafo phase splitter. Looks like a very promising topology (KISS); weird though that it's so rare... But then again... this thread has brought one RH807 and one Rankin 807 out of the dark and there are probably much more people *silently* enjoying their homebuilt gear
Simon
The reason that interstage transformers are so rarely implemented is because they often cost more than output transformers. This is because unlike output transformers they are driving a high impedence load with no step down. This makes them very difficult to build without capacitance becoming the limiting factor. Mine uses cheap and simple toroidal mains transformers in a 1:1+1 setup. Unaided response is only to 8Khz, but I use a ECL82 with plate to plate feedback to lower their output impedence and hence improve their drive capability (exactly the same principle as the RH designs), I also use a bit of GNF around the interstage. I get response flat to 35khz with no overall global feedback.
A simpler approach is to step down in the interstage, which eliminates the capacitance issues.
My research lead me to believe that the very best results were achieved with interstage splitting PP designs.
A simpler approach is to step down in the interstage, which eliminates the capacitance issues.
My research lead me to believe that the very best results were achieved with interstage splitting PP designs.
Going pentode/tetrode tilts the response away from flat with a rise in high frequency response. Therefore by going tetrode you have moved away from the ideal of zero or low feedback because feeback becomes essential to flatten the response. By your definition of the ideal isn't triode the only way to go ?
Shoog
Shoog said:
Not really. I never said that not using NFB was an "ideal" because I don't buy into the latest audiophile fad that says it is. NFB is useful, but easy to abuse. In all too many cases where it results in poor performance and sound-alike amps, you often find that it's being used to correct for fundamental design errors (SS designers are especially guilty of this, but VT designers aren't immune to it either.) or to beat the numbers game. There's more to good sound than an "impressive" THD spec.
As with everything else, there are always design trade-offs. Go with tetrodes, and you have the problem of excessive high frequency response that gives a "strident" or "gritty" or "grainy" or overly "bright" sound at the high end with poor damping at the lower. This can be cleaned up with the judicious use of local feedback in one form or another: parallel, cathode, ultralinear, etc. OTOH, with low u audio or RF power triodes, you solve those problems, yet bring others to the table. The low u triode will certainly require greater, sometimes much greater, drive voltage than will a tetrode producing the same Po. A front end that swings a greater voltage is also more likely to generate more harmonic distortion. It will also have the additional difficulties of sourcing current to charge up the much higher Ci + Cmiller of the power triode, and source grid current under grid-positive overdrive conditions. In either case, it's best to get that distortion down when running open loop and add just enough gNFB to correct for residual front end distortion and, of course, OPT nonlinearities.
Hi all,
I just thought you might be interested in how I spent my Sunday. I have just completed the circuit part of the conversion of my parafeed RH807 from CCS loaded to Choke loaded. Its still on the test bench with crap speakers, so its to early to say how it sounds really.
I used microwave oven transformers as my plate chokes. Surprisingly enough they work fine. I had a few worries for a bit because I was losing response from about 300hz down. I tried putting another choke in series with the MOT to see if it was lack of plate inductance, but that didn't clear it up at all. I messed about with the plate to plate feedback to no effect. Finally I remembered that my output transformers (modified mains toroidals) need to be wired up with the primaries in one direction. So I reversed the connection and hey presto the response is nearly flat down to 10hz. Response is good up to 35khz where it starts to trail off - which if my memory serves me well was how it performed with the CCS load.
I have the screen of the 807 with a voltage divider and a 100uf bypass.
In my tests I thought that the bass roll off might be attributable to the pentode connection, so I wired it up for triode strapped with no series resistance, it helped a tad but not much. There was a difference in sound though. The triode strapped version sounded a bit smoother, though not much.
With the choke load I am able to use more plate to plate feedback. Before I had 150K, any more produced a resonant hump in the sub 100hz bass response. I am now able to use 100K without any hump.
Overall I am very happy with the result so far. I will report some more when I have finished the case and auditioned in the main system.
Shoog
I just thought you might be interested in how I spent my Sunday. I have just completed the circuit part of the conversion of my parafeed RH807 from CCS loaded to Choke loaded. Its still on the test bench with crap speakers, so its to early to say how it sounds really.
I used microwave oven transformers as my plate chokes. Surprisingly enough they work fine. I had a few worries for a bit because I was losing response from about 300hz down. I tried putting another choke in series with the MOT to see if it was lack of plate inductance, but that didn't clear it up at all. I messed about with the plate to plate feedback to no effect. Finally I remembered that my output transformers (modified mains toroidals) need to be wired up with the primaries in one direction. So I reversed the connection and hey presto the response is nearly flat down to 10hz. Response is good up to 35khz where it starts to trail off - which if my memory serves me well was how it performed with the CCS load.
I have the screen of the 807 with a voltage divider and a 100uf bypass.
In my tests I thought that the bass roll off might be attributable to the pentode connection, so I wired it up for triode strapped with no series resistance, it helped a tad but not much. There was a difference in sound though. The triode strapped version sounded a bit smoother, though not much.
With the choke load I am able to use more plate to plate feedback. Before I had 150K, any more produced a resonant hump in the sub 100hz bass response. I am now able to use 100K without any hump.
Overall I am very happy with the result so far. I will report some more when I have finished the case and auditioned in the main system.
Shoog
Hi again,
Got the amp crated up and playing in the main system. I had to switch back to the 150K plate to plate because there obviously were some issues with resonance in the low 100's hz range.
Sound wise things are overall very good. Its different to my Class A PP amp in all the ways you might expect. The RH807 has great midrange presentation - as if a spotlight were been beamed onto the vocal range. The highs are musical but not outstanding, and the bass is strong but lacking in control compared to the push pull.
Overall There seems to be less detail resolution than the PP amp, though the sound seems more integrated somehow.
Overall I would say that the PP amp wins because of its better bass control and detail resolution, but its a very close run thing. I have yet to do an A/B comparison so things may change with time.
Shoog
Got the amp crated up and playing in the main system. I had to switch back to the 150K plate to plate because there obviously were some issues with resonance in the low 100's hz range.
Sound wise things are overall very good. Its different to my Class A PP amp in all the ways you might expect. The RH807 has great midrange presentation - as if a spotlight were been beamed onto the vocal range. The highs are musical but not outstanding, and the bass is strong but lacking in control compared to the push pull.
Overall There seems to be less detail resolution than the PP amp, though the sound seems more integrated somehow.
Overall I would say that the PP amp wins because of its better bass control and detail resolution, but its a very close run thing. I have yet to do an A/B comparison so things may change with time.
Shoog
Took the amp back down to the test bench to do some more measurements.
The sound of the amp is very full on. More so than I remember it been with the CCS plate load, though that was triode strapped as well. I wanted to see if there was any hint of reasonance between the choke/transformer/parafeed cap. A sine wave sweep showed flat response down to about 15hz and up to about 30khz. I then switched to square wave response. Overall the square wave comes through very cleanly. However careful examination showed that there was overshoot of the leading edge at frequencies bellow about 200hz, this seemed to be about half the signal level over. Seems to suggest that on hard transients the choke/transformer has relatively poor control over the signal (poor damping). Without the plate to plate feedback this manifested as a large spike of at least twice the signal amplitude, with the feedback this was much reduced.
In one instance this seemed enough to send the 807 into self oscillation, but this went away and I couldn't reproduce the conditions. This could have been a lose top cap connection - I don't really know.
The conclusion I come to is despite the plate to plate feedback, this amp still suffers from relatively poor damping compared to a good PP amp. I would guess that the use of a better choke would probably improve the situation somewhat. It could however be the difference between pentode sound and triode strapped sound.
Shoog
The sound of the amp is very full on. More so than I remember it been with the CCS plate load, though that was triode strapped as well. I wanted to see if there was any hint of reasonance between the choke/transformer/parafeed cap. A sine wave sweep showed flat response down to about 15hz and up to about 30khz. I then switched to square wave response. Overall the square wave comes through very cleanly. However careful examination showed that there was overshoot of the leading edge at frequencies bellow about 200hz, this seemed to be about half the signal level over. Seems to suggest that on hard transients the choke/transformer has relatively poor control over the signal (poor damping). Without the plate to plate feedback this manifested as a large spike of at least twice the signal amplitude, with the feedback this was much reduced.
In one instance this seemed enough to send the 807 into self oscillation, but this went away and I couldn't reproduce the conditions. This could have been a lose top cap connection - I don't really know.
The conclusion I come to is despite the plate to plate feedback, this amp still suffers from relatively poor damping compared to a good PP amp. I would guess that the use of a better choke would probably improve the situation somewhat. It could however be the difference between pentode sound and triode strapped sound.
Shoog
Shoog, comparing triode strapped and pentode would be almost comparing apples and oranges. With choke load, the tube Rp is in AC parallel with the choke and can be seen as a form of damping, and this will be much lower in triode compared to pentode mode.Of course, P-to-P feedback will lower effective 'Rp' as seen by the choke, but still, this is one thing to consider that would be very different triode vs pentode. Isn't one of the first rules of successful tweaking to change one thing at a time?
Hi there again,
I just did some tests on the amp to see if different screen connection made any difference.
I tried the voltage divider (47K - 8K6) with 100uf of smoothing.
I tried a series resistor of 8K6.
I tried a pure triode strapped with no series resistance at all.
I tested it with both sine wave and square wave, with an output of about 1watt. I got useable response from all three up to 120Khz (down by half power in all cases) and down to 10hz (though the square wave was triangular at this point) with good square wave down to about 40hz. I didn't do any harmonic distortion measurements - because I can't.
And the result was - they all measured near enough exactly the same. Subjectively there might have been a very slight softening of the sound on the pure triode strapped version - but if it was there it was small. I didn't put it into my main system because I couldn't face lugging its huge bulk up and down to the shed.
So I can only conclude that the extra effort of doing good screen regulation just isn't worth it, just as Mr Kitic suggested. I can only conclude that the difference in bass control between my PP amp and the RH807 is an intrinsic quality of SE output stages. If you want big bass with great midrange then the RH807 is the design for you. If you want tightly controlled bass then a good PP design seems in order.
Hope that help. I would be very interested is someone else could confirm or refute my observations.
Shoog
I just did some tests on the amp to see if different screen connection made any difference.
I tried the voltage divider (47K - 8K6) with 100uf of smoothing.
I tried a series resistor of 8K6.
I tried a pure triode strapped with no series resistance at all.
I tested it with both sine wave and square wave, with an output of about 1watt. I got useable response from all three up to 120Khz (down by half power in all cases) and down to 10hz (though the square wave was triangular at this point) with good square wave down to about 40hz. I didn't do any harmonic distortion measurements - because I can't.
And the result was - they all measured near enough exactly the same. Subjectively there might have been a very slight softening of the sound on the pure triode strapped version - but if it was there it was small. I didn't put it into my main system because I couldn't face lugging its huge bulk up and down to the shed.
So I can only conclude that the extra effort of doing good screen regulation just isn't worth it, just as Mr Kitic suggested. I can only conclude that the difference in bass control between my PP amp and the RH807 is an intrinsic quality of SE output stages. If you want big bass with great midrange then the RH807 is the design for you. If you want tightly controlled bass then a good PP design seems in order.
Hope that help. I would be very interested is someone else could confirm or refute my observations.
Shoog
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.