Hi All,
I am refurbishing a Leak Stereo 20 for someone, and have a few questions.
1. More than one circuit diagram (component, valuewise) appear to exist. The amp in my possession seems to have several changes not listed on any diagram. These components appear to be the original as mounted in the factory. (At this stage I must apologise for not succeeding in copying the diagram onto this post. I will refer to component designations as per the diagram mostly found.)
2. My main problem is that R5, indicated as 100K on the diagram has been lowered to 47K in the amp. With the indicated cathode resistor of R3 (2K2), the input triode works at a rather starved point on the Ip/Vp curves, giving at least 2x the distortion it would have with R5 = 100K. (The plate signal swing is very low, but still; preferred R3 for R5=47K would be 1K2.) Can anybody recall having seen R5 as 47K on any mod?
3. Going to R3=100K increases the loop gain by some 1,7x, bringing the amp on the verge of instability with a then feedback factor of 25 dB. For this one can lower R4 (100R) but that moves it away from the original again. What was the position?
4. In my amp C5 = 100nF, but the diagram indicates 20nF. This presumably for some stability improvement at low frequencies (which would have been better served by keeping C5 high and rather decreasing C3). What did other members encounter?
5. The capacitor C1=1nF (if I read correctly) in parallel with R4=100R, presumably for some h.f. stability, has a -3 dB (45 degree) effect above 1.5 MHz only, where there certainly is no loop gain left to cause any problems; the loop gain passes trough 1 at some 200+ KHz. Is C1=1nF correct?
I may add that all components have been replaced, so encountered stability effects are not as a result of aged components.
Any comments would be greatly appreciated.
Regards.
I am refurbishing a Leak Stereo 20 for someone, and have a few questions.
1. More than one circuit diagram (component, valuewise) appear to exist. The amp in my possession seems to have several changes not listed on any diagram. These components appear to be the original as mounted in the factory. (At this stage I must apologise for not succeeding in copying the diagram onto this post. I will refer to component designations as per the diagram mostly found.)
2. My main problem is that R5, indicated as 100K on the diagram has been lowered to 47K in the amp. With the indicated cathode resistor of R3 (2K2), the input triode works at a rather starved point on the Ip/Vp curves, giving at least 2x the distortion it would have with R5 = 100K. (The plate signal swing is very low, but still; preferred R3 for R5=47K would be 1K2.) Can anybody recall having seen R5 as 47K on any mod?
3. Going to R3=100K increases the loop gain by some 1,7x, bringing the amp on the verge of instability with a then feedback factor of 25 dB. For this one can lower R4 (100R) but that moves it away from the original again. What was the position?
4. In my amp C5 = 100nF, but the diagram indicates 20nF. This presumably for some stability improvement at low frequencies (which would have been better served by keeping C5 high and rather decreasing C3). What did other members encounter?
5. The capacitor C1=1nF (if I read correctly) in parallel with R4=100R, presumably for some h.f. stability, has a -3 dB (45 degree) effect above 1.5 MHz only, where there certainly is no loop gain left to cause any problems; the loop gain passes trough 1 at some 200+ KHz. Is C1=1nF correct?
I may add that all components have been replaced, so encountered stability effects are not as a result of aged components.
Any comments would be greatly appreciated.
Regards.
Hi,
I give a link for the schema, hope it helps for the other members to answer your questions:
http://werple.net.au/~kiewavly/Leak/Circuits/Stereo20.gif
Best regards,
I give a link for the schema, hope it helps for the other members to answer your questions:
http://werple.net.au/~kiewavly/Leak/Circuits/Stereo20.gif
Best regards,
Thanks TJ. Thus far I could not find a different circuit diagram, but the power transformer measures 280 - 0 - 280V and does not get too hot. There is no indication anywhere of the vintage.
Anatech,
I was forced to change component values slightly to get an acceptable amplifier - reluctantly so. As said above the feedback was overly high to my mind, and I decreased it to 15x by changing R18 (the feedback resistor) to 18K. Furthermore I found the phase compensation by C9 totally excessive. I lowered that to 68 pF, giving an almost perfect square wave response without a load (under which condition it was then stable), and a slight "undershoot" into an 8 ohm load. This looks better with loudspeaker load, which is higher than 8 ohm at h.f.
In this process I not only varied the C9 value, but also tried a serie resistor as is sometimes found. I found it best with no resistance in serie with C9.
Then I found C1 of no use, as said before. It very slightly increased the small damped oscillation at the onset of a square wave, and I omitted it.
I also changed R5 to 100K (as shown on the diagram; see above) and R3 to 1,5K to get the anode swing into a more linear region. (R3 could actually go to 1K for still better operation.) Also I later noticed that R12, R13 in my model were 470K, not 1M as per the diagram - I kept that value.
Lastly R2=22K gave a loop response of -3 dB at only 12 KHz. I lowered that to 4,7K, which also slightly improved the square wave (onset) response. I realise that that puts the input stage somewhat more at the mercy of the feeding impedance, but at least now it will be better with a low impedance feed. (My signal generator feeds from 600 ohm.)
These are a lot of changes by someone who likes to keep originals as was, but it could not be avoided. Incidentally, this was the 2nd Leak Stereo 20 that I had to alter in this manner, which lowers the chance that I might have an odd one here.
I must admit that I just cannot fathom the reasons for the original values. I am still hanging on in case someone else comes up with wisdom.
Regards.
Anatech,
I was forced to change component values slightly to get an acceptable amplifier - reluctantly so. As said above the feedback was overly high to my mind, and I decreased it to 15x by changing R18 (the feedback resistor) to 18K. Furthermore I found the phase compensation by C9 totally excessive. I lowered that to 68 pF, giving an almost perfect square wave response without a load (under which condition it was then stable), and a slight "undershoot" into an 8 ohm load. This looks better with loudspeaker load, which is higher than 8 ohm at h.f.
In this process I not only varied the C9 value, but also tried a serie resistor as is sometimes found. I found it best with no resistance in serie with C9.
Then I found C1 of no use, as said before. It very slightly increased the small damped oscillation at the onset of a square wave, and I omitted it.
I also changed R5 to 100K (as shown on the diagram; see above) and R3 to 1,5K to get the anode swing into a more linear region. (R3 could actually go to 1K for still better operation.) Also I later noticed that R12, R13 in my model were 470K, not 1M as per the diagram - I kept that value.
Lastly R2=22K gave a loop response of -3 dB at only 12 KHz. I lowered that to 4,7K, which also slightly improved the square wave (onset) response. I realise that that puts the input stage somewhat more at the mercy of the feeding impedance, but at least now it will be better with a low impedance feed. (My signal generator feeds from 600 ohm.)
These are a lot of changes by someone who likes to keep originals as was, but it could not be avoided. Incidentally, this was the 2nd Leak Stereo 20 that I had to alter in this manner, which lowers the chance that I might have an odd one here.
I must admit that I just cannot fathom the reasons for the original values. I am still hanging on in case someone else comes up with wisdom.
Regards.
Every Leak tube amp schematic that I can remember, from 12w to 50w, has that 0.001uF (C1) in there. I have never seen an explanation of what it's for and nobody else's design seems to use it.
The only thing I can think of is, maybe it's supposed to compensate for HF roll-off caused by the R-C network across the plate load of the first stage.
Thanks everybody.
This would appear to conclude this for now - I have had no further inputs, also (not) from other sources. One or two other sites came up but a few years old - e-mail addresses not active any longer. I have since checked yet another Stereo 20 - same thing and findings. The original value choices must remain a mystery, until somewhere other wisdom comes up by chance.
Meet you on other threads.
This would appear to conclude this for now - I have had no further inputs, also (not) from other sources. One or two other sites came up but a few years old - e-mail addresses not active any longer. I have since checked yet another Stereo 20 - same thing and findings. The original value choices must remain a mystery, until somewhere other wisdom comes up by chance.
Meet you on other threads.
I have to say that I'm as bemused as everyone else about that 1nF capacitor. It effectively forms a step network at RF that reduces feedback and long before that it affects the phase angle of the feedback. You can produce the same effect by adding a resistor in series with the capacitor bypassing the feedback resistor, and that's significant because the fairy fingers feedback box that I use for determining optimum feedback has a variable resistor in series with the bypass capacitor and on lots of experiments with different output transformers etc I've never yet seen adjusting R > 0 do anything but make the amplifier ring or oscillate.
That means I'm saying that Leak got it wrong. But a few things ought to be borne in mind. When the Leak valve amplifiers were designed an oscilloscope was genuinely laboratory equipment, as was the function generator (to produce the required square waves). Non-inductive dummy loads were also virtually unknown. Thus, an amplifier would have been tested using a dodgy source, driving a wirewound (inductive) dummy load, into an oscilloscope with perhaps 2MHz of bandwidth. It's no wonder that they arrived at seemingly inexplicable results. These days, a cheap and cheerful function generator is capable of producing a 10MHz fundamental so its 10kHz square waves are almost beyond reproach, 50W non-inductive metal film resistors are readily available, and it's difficult to buy an oscilloscope with a bandwidth of <20MHz.
That means I'm saying that Leak got it wrong. But a few things ought to be borne in mind. When the Leak valve amplifiers were designed an oscilloscope was genuinely laboratory equipment, as was the function generator (to produce the required square waves). Non-inductive dummy loads were also virtually unknown. Thus, an amplifier would have been tested using a dodgy source, driving a wirewound (inductive) dummy load, into an oscilloscope with perhaps 2MHz of bandwidth. It's no wonder that they arrived at seemingly inexplicable results. These days, a cheap and cheerful function generator is capable of producing a 10MHz fundamental so its 10kHz square waves are almost beyond reproach, 50W non-inductive metal film resistors are readily available, and it's difficult to buy an oscilloscope with a bandwidth of <20MHz.
Quite so, EC8010.
But then I would still have expected Harold Leak to have noticed some effect from that 1nF with whatever equipment he had, so what could it have been?
I have noticed on occasion a C9.C1 network used to take over capacitively so to speak from R18.R4 (using the Leak circuit designations), ostensibly to swamp other undesirable capacitive effects from wiring or whatever. It also seemed an early attempt at the phase-lead-input-lag kind of feedback equalisation as opposed to the Cdom kind (R.C over the first anode load). This Leak does rather suffer from the input ECC83 Miller effect, curtailing loop gain early at h.f. (which is why I am not in favour of using this tube for inputs - but that is another subject).
But all that does not work here. One can also frown at the relatively long lengths of wiring used in the low impedance side of the feedback where C1 is supposed to act. (You would not build your medium wave radio with such leads.) So ........
Regards and thanks for your input.
But then I would still have expected Harold Leak to have noticed some effect from that 1nF with whatever equipment he had, so what could it have been?
I have noticed on occasion a C9.C1 network used to take over capacitively so to speak from R18.R4 (using the Leak circuit designations), ostensibly to swamp other undesirable capacitive effects from wiring or whatever. It also seemed an early attempt at the phase-lead-input-lag kind of feedback equalisation as opposed to the Cdom kind (R.C over the first anode load). This Leak does rather suffer from the input ECC83 Miller effect, curtailing loop gain early at h.f. (which is why I am not in favour of using this tube for inputs - but that is another subject).
But all that does not work here. One can also frown at the relatively long lengths of wiring used in the low impedance side of the feedback where C1 is supposed to act. (You would not build your medium wave radio with such leads.) So ........
Regards and thanks for your input.
May I toss out a hypothesis about the 1nF based on no evidence other than the results that Johan has reported? I'd guess that it is to allow the spec sheet to show the amplifier to be flat to 20kHz. The ringing is a consequence of trying to extend the high end to make the specs be competitive. No doubt the designer felt that the sonic consequence was minimal...
I've just been leafing through Kevin Spicer's "Firsts in High Fidelity" book that is a history of Leak (interesting book by the way) in the hope of picking up a clue and I also reread this thread. Spicer reports that Leak had extremely long loudspeaker cables to allow stability testing and it occurs to me that they would be a good aerial. Perhaps the purpose of that 1nF capacitor was to shunt any RF picked up by the aerial that might overload the (ludicrously sensitive) input stage and be demodulated?
Regarding SY's hypothesis, I noticed that the original TL12 amplifier was expected to be flat 20Hz - 20kHz +/-0.1dB. I would expect that to have been a very tricky measurement using 50s and early 60s test equipment...
Regarding SY's hypothesis, I noticed that the original TL12 amplifier was expected to be flat 20Hz - 20kHz +/-0.1dB. I would expect that to have been a very tricky measurement using 50s and early 60s test equipment...
SY,
Ye-e-e-s - but the amp easily reaches 20 KHz although the loop gain is 3 dB down at about 14 KHz. (That can be moved up by decreasing input serie resistor R2 from 22K to 3,3K or so as said previously.) As said C1 does nothing until way out of the audio range. The time constant with R4=100 ohms is at 1,6 MHz! (although phase influence shows earlier, but still at quite over 100 KHz). Nevertheless a very valid remark.
EC8010,
I did not think of that. The amp is still here and I will try to see if I can induce any rf effects. I think your remark is probably close. Then I would very much like to have a look at the Kevin Spicer book. I would imagine it is out of print, but can you kindly give the publisher, please?
Also, thanks to both for not picking up on my error regarding the phase-lead-input-lag remark! I was staring blind at the present C1 and forgot that it goes to common. For input lag it of course needs to go cathode-grid instead (with a much lower value).
Regarding the +/- 0,1 dB: If one should measure at 5Vp output (about 1W into 15 ohm), such a difference would show an output of 5,06Vp; indeed a tricky measurement; if by scope not much more than the scope trace width those days. I presume a vacuum tube voltmeter with a big meter face could show that, but then how did you determine that the VTVM is accurate to that degree in the first place?! Paper started to be patient a long time ago.
Regards.
Ye-e-e-s - but the amp easily reaches 20 KHz although the loop gain is 3 dB down at about 14 KHz. (That can be moved up by decreasing input serie resistor R2 from 22K to 3,3K or so as said previously.) As said C1 does nothing until way out of the audio range. The time constant with R4=100 ohms is at 1,6 MHz! (although phase influence shows earlier, but still at quite over 100 KHz). Nevertheless a very valid remark.
EC8010,
I did not think of that. The amp is still here and I will try to see if I can induce any rf effects. I think your remark is probably close. Then I would very much like to have a look at the Kevin Spicer book. I would imagine it is out of print, but can you kindly give the publisher, please?
Also, thanks to both for not picking up on my error regarding the phase-lead-input-lag remark! I was staring blind at the present C1 and forgot that it goes to common. For input lag it of course needs to go cathode-grid instead (with a much lower value).
Regarding the +/- 0,1 dB: If one should measure at 5Vp output (about 1W into 15 ohm), such a difference would show an output of 5,06Vp; indeed a tricky measurement; if by scope not much more than the scope trace width those days. I presume a vacuum tube voltmeter with a big meter face could show that, but then how did you determine that the VTVM is accurate to that degree in the first place?! Paper started to be patient a long time ago.
Regards.
It would have been nice if I had got the author's name right. "Firsts in High Fidelity" Stephen Spicer. Audio Amateur Press. (2001) I looked on the audioXpress site but couldn't find it. Amazon (UK) claimed to be able to get it, but at the princely sum of £68.70! Looks like you may have to go to Alibris.
Stone Audio (www.stoneaudio.co.uk) are still advertising the book at its original price of £24.95.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.