Author Topic: Leak Stereo 20 restoration/mods  (Read 10754 times)

Offline Ampdog

Re: Leak Stereo 20 restoration/mods
« Reply #15 on: June 06, 2015, 05:33:42 PM »
Further on the Leak Stereo 20 (Technical Warning)

Caveat: The following is not intended to belittle the design or HJ Leak. I respect that the amplifier is popular. The difficulties encountered would for the most not present as audible, but instability could be encountered with some valve spreads. These were detected with instruments and much testing, not by hearing.

There is possibility of l.f. instability round 0,2 - 1 Hz, which could result in 'pumping' visible at the loudspeaker cone or in worst case, e.g. together with a music beat round about that tempo, cause run-away damaging the loudspeaker and amplifier.

The 8778 OPT measures a rather high primary inductance of some 190H (at 5V, 50 Hz), plus leakage inductance of round 15 mH (1KHz, secondary shorted over 15 ohm connection. These figures are in themselves superior to other OPTs - that far that good. But it does lend to somewhat higher internal capacitance, causing mild reaction in the 90KHz area. This cannot be totally cancelled by the usual measures but to my reckoning should not cause spurious h.f. oscillation with any loudspeaker system. Nevertheless, the universal load resistor is added over the secondary so as to calm matters somewhat in the presence of very capacitive loads.
A 1K.ohm 5W resistor is sufficient and may be wire-wound; slight component inductance will not count. I have mentioned before that such a resistor was universally used in valve days. Negligible power consumption and will render many valve amplifiers (including this one) stable even in case of open load. (The Stereo 20 could oscillate at 40 - 70 KHz without this, open load.)

The l.f. stability was questionable as at present. A sudden pulse displayed a 1 Hz 2 cycle jumping tendency on rails before stabilising. Not to bore, but simply put, a stable amplifier needs to maintain loop phase angle to under 90 until the loop gain drops to below unity in order to obviate unstable response. In practice this means that all phase shifting R.Cs except one, need to be large enough to keep the phase angle out of going positive until the NFB has gone low enough, by that one RC, not to cause enough feedback for oscillation. (Fellow technologists will pardon some gross simplification, please!)

In the case of the Stereo 20 this could only be achieved by a very large C5 of 470nF. It would appear that some effort toward this was made by previously making R11.C5 the small R,C by way of C5 = 20nF. In some models C5 was found as 100nF. The problem here is that a low C5 would accomplish the purpose of diminishing loop gain to the desired figure, but with a penalty. A small C5 will cause one leg of the LTP phase inverter to go away from 180 out-of-phase at l.f., causing additional problems, so one 'chickens out' of that measure. With the very high inductance of the OPT, C4.R13 and C6.R12 now become the dominant phase-shifting time constant, assisted by the time constants of C7.R14 and C8.R15 - the role of those are important at l.f. but falls outside this discussion.

By the way, R12.R13 = 1 meg.ohm initially, gave such comebacks because of run-away EL84s that Leak lowered them to 470K. Not his fault although he was using maximum allowed values.

I will ease up on the theory of the R.C in parallel with R5, just to mention that the purpose is the same as above for h.f.: To lower loop gain in time to obviate h.f. oscillations because of instability. But for full audio band effectivity of NFB, one must not start with this so 'early' as to diminish NFB within the audio band (and thereby hangs a lot of stridency experienced with many amplifiers: Inability to maintain full NFB up to 20 KHz. But that is another story.). To this end the best values I found were R = 4K7 (as in most Stereo 20s) but decreased C = 120pF instead of the installed 200pF. By the way, in some models R5 was reduced to 47K - not sure why. If to reduce gain it was misguided as 47K places the ECC83 operating point at an undesired position.

Finally, Leak's designs reaches the famous 0,1% distortion level by some 26 dB of NFB - a high value, where designers must start to be mindful of the basics. In fact, with R18 = 12K, I found the NFB to be a very high 29dB (26 dB is a factor of 20x, in the models I tested this was almost 28 - 30 x). I therefore increased R18 to 15K-18K - enough trouble as it is.

This then the explanation behind the changes I suggest.  Still I hesitate to proclaim this as a universal remedy though. The spread in characteristics of valves these days as also evidenced by the many tests of Charles, leaves one nervous. Perhaps to relent to 0,2% distortion in the interest of universal calm.

Gratitude for patience from those hopefully gaining by this lengthy exposition. Because of the popularity of the amplifier many suggestions are to be found as to the 'best' adaptations. The above are mine, not to alter the basic circuit significantly rather than to improve stability.   
« Last Edit: June 06, 2015, 05:38:43 PM by Ampdog »
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline skollie

  • Trade Count: (+5)
  • AVForums Grandmaster
  • *****
  • Posts: 5,915
  • Total likes: 0
  • Cableth make no differenth. Ith all in the plugth
    • Blue Angel
Re: Leak Stereo 20 restoration/mods
« Reply #16 on: June 06, 2015, 07:49:30 PM »
Whew!! So I've been sitting on a ticking time-bomb all these years.

I've put on goggles, ear protectors and donned my bomb disposal suit and unplugged it from the mains. Now busy filling sandbags for safe storage in my garage :ROFLMAO:

I vaguely remember a story (likely an urban legend) of Harold J surreptitiously exiting via fire escape stairs during some or other UK audio show. I'm not sure if it was during an occasion when he set up a Stereo 20 for the public to listen to but it could have been.

FWIW, I have never experienced 'pumping' drivers while using the S/20.

Regds

skollie
« Last Edit: June 06, 2015, 08:17:52 PM by Skollie »
If you don't like my principles, I have others - Groucho Marx

Offline charles

Re: Leak Stereo 20 restoration/mods
« Reply #17 on: June 06, 2015, 10:33:47 PM »
"FWIW, I have never experienced 'pumping' drivers while using the S/20."

I agree with Andre

Offline Ampdog

Re: Leak Stereo 20 restoration/mods
« Reply #18 on: June 07, 2015, 03:17:23 AM »
Gentlemen,

I said there was the 'possibility of',  under certain conditions. Not every model would have exhibited this. Showed up at tests as a very mild signal, some 10 mV on output in the present model, but could be excited by a certain rythm, particularly with a high loudspeaker impedance at l.f..  I do not think HJ would have released it if it occurred with any regularity. Nyquist stability analysis reveals the possibility of, but cumbersome to calculate for worst-case scenarios. E.g. high-gain ECC83s would have exacerbated this; most however test lower than published (see Charles's previous post on this). See other Leak designs of similar topology, where the relevant cap was often 250NF.  But no quibble; those were my findings for unconditional stability. Enough maths for now.   
 
Charles, Skollie, others: What value for C5 does your amplifiers have? My schematic shows 20 NF, amplifiers that came my way had 100 NFs. Can anybody point me to a schematic specifically stating the use of the 3125 OPT?  (Why oh why do designers often have so many different designs for the same thing?)

Skollie,

Put away your protection gear, or rather have it more ready for the weather you are having these days!
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline skollie

  • Trade Count: (+5)
  • AVForums Grandmaster
  • *****
  • Posts: 5,915
  • Total likes: 0
  • Cableth make no differenth. Ith all in the plugth
    • Blue Angel
Re: Leak Stereo 20 restoration/mods
« Reply #19 on: June 07, 2015, 02:30:10 PM »
Hello Ampdog

Apologies for my silly post. I bow to your superior knowledge about valve amplifiers and still have difficulties distinguishing between capacitors and resistors.

I have a Stereo 20 schematic and will have a look later to see what I have at C5.

The other thing which bothers me is that resistor which is supposed to desolder itself in case of trouble. How does it do that, even if it gets so hot that the solder melts? If the resistor overheats and causes the solder to melt, it will obviously not desolder if the wire ends are bent over at the tags.

IMO it will also not desolder even when the wire ends are kept straight and pushed into the tags without bending them.

I learned late last night from a fellow forum member that a wellknown Jhb restorer we all know also uses a fuse in this place for better protection.

Regds

skollie
« Last Edit: June 07, 2015, 02:38:18 PM by Skollie »
If you don't like my principles, I have others - Groucho Marx

Offline charles

Re: Leak Stereo 20 restoration/mods
« Reply #20 on: June 07, 2015, 08:48:34 PM »
Dear Andre,

Maybe this will help. Many years ago I obtained the ST 20 that I still have.

When I got it, I switched it on and just played it for a couple hours.  This was my first real experience with a Leak valve amplifier. Then It stopped playing. I looked under the chassis and saw the big green resistor desoldered itself. The wire ends of the resistor were straight.

I replaced all the coupling capacitors and reservoir electrolytic ones. Soldered the the green resistor in its place. Switched it on and the resistor stayed in its place.   

Offline skollie

  • Trade Count: (+5)
  • AVForums Grandmaster
  • *****
  • Posts: 5,915
  • Total likes: 0
  • Cableth make no differenth. Ith all in the plugth
    • Blue Angel
Re: Leak Stereo 20 restoration/mods
« Reply #21 on: June 07, 2015, 08:56:52 PM »
Thanks again, Charles :)

That is just old-fashioned magicke. I've got an underside pic here somewhere of my Leak and the last time I looked, it had TWO large green resistors soldered in that spot. I didn't do this - it was done by a tech gent that worked on quite a few of my things.

OT I will send you a pic and also one or two from the Pye monos so you can see what awaits me.

Regds

skollie
If you don't like my principles, I have others - Groucho Marx

Offline afroaudio

Re: Leak Stereo 20 restoration/mods
« Reply #22 on: June 08, 2015, 04:54:40 PM »
Johan I have a schematic which is apparently for the earlier version of the ST 20 (with 3921 OPTs)
C5 is listed here as 20nF.

I can send you this schematic if you're interested.

Offline afroaudio

Re: Leak Stereo 20 restoration/mods
« Reply #23 on: June 13, 2015, 07:42:34 PM »
This is the schematic I have that is apparantly for earlier Stereo 20 (3921 OPTs)

Leak Stereo 20 3921 OPT by Bregtje Cals, on Flickr

Offline Ampdog

Re: Leak Stereo 20 restoration/mods
« Reply #24 on: June 13, 2015, 10:39:24 PM »
Thanks Afroaudio,

That is the schematic in my possession, though not indicating for which OPT. Anyway, I would doubt that there would have been much difference in characteristics of the OPTs.

I can only repeat that there were signs of latent l.f. instability after e.g. a momentary short overload. The amp took some two cycles to stabilise. By eliminating C5 from the 'stability algorithm' by making it large, return to stability following an overload pulse was immediate. [I have only recently received my PC with my Spice software back after service (I use a separate one disconnected from any internet access), and will do a l.f. simulation there when next I have time.]

It would then appear that C5 was 100nF in later models, but again I have no data as to with which OPT.

Charles? 
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline Ampdog

Re: Leak Stereo 20 restoration/mods
« Reply #25 on: June 13, 2015, 11:56:10 PM »
Skollie,

Regarding the green wire-wound resistor (R21). Replacing that once blown or de-soldered might be a problem. One does not know what solder was originally used or what the temperature characteristics of the resistor were. Modern multi-core solder (60% Sn, 40% Pb) melts at about 180C. I read that the maximum temperature for ordinary w.w. resistors is 300 - 400C. So fine there .... but how long will the leads take to reach say 200 at the solder point, what is their thickness and temp. conduct. factor? Again with respect to HJL, I would rather use a fuse - but then again one might have to look for a quite delayed fuse in order not to blow it too frequently. HJL might have had such a delay in mind, or just a rough "long-stop" measure. There is a main fuse.

Probably re-soldering with not too much solder and too long resistor leads one can approximate the intended condition. Is that likely to happen in repair shops? No - I doubt if anybody knows the purpose of those resistors these days. (One also gets these funny cement resistors these days. How will they behave?)
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline skollie

  • Trade Count: (+5)
  • AVForums Grandmaster
  • *****
  • Posts: 5,915
  • Total likes: 0
  • Cableth make no differenth. Ith all in the plugth
    • Blue Angel
Re: Leak Stereo 20 restoration/mods
« Reply #26 on: June 14, 2015, 10:37:31 AM »
Hello Ampdog

The Stereo 20 I have is not in constant, continuous use - maybe once or twice per week for about an hour or two and it has never given me any cause of concern.

I played it yesterday in the presence of another forum member and each time I use it, it surprises me with sheer listenability. But then, in the land of the blind and deaf...

I sometimes regret that you are located in faraway Pretoria and I'm here in the land of Kaapse skollies.

Over the next few weeks, I will see a gentleman knowledgeable about valve things so I will ask him about that dropping off resistor to ensure mine will do what it is supposed to do.

Regards

skollie
If you don't like my principles, I have others - Groucho Marx

Offline afroaudio

Re: Leak Stereo 20 restoration/mods
« Reply #27 on: June 15, 2015, 04:42:30 PM »
In case this is of value to anyone, this seems to be the other version of the Stereo 20 schematic:

Stereo 20 alt schematic by Bregtje Cals, on Flickr

Offline Ampdog

Re: Leak Stereo 20 restoration/mods
« Reply #28 on: June 24, 2015, 03:09:19 PM »
Afro,

Haven't visited here for a while - missed this now being a sticky!!  (no comment please ....)

Thanks for the schematic.  Where are we now??

Yes, I have both those schematics though not tied to OPT types.  Charles, are there more versions? Not that I could find.

It brings one to kind of the best of both.  Not rereading (so pardon repetition), I still suggest the values found by me under post #14; most of the reasons given.

But it seems I was wrong regarding the amplifier being not too sensitive regarding input signal. I found that I recorded 70 mVrms input for full output. That would be classed as sensitive in most cases. The best remedy would involve a perhaps too drastic change. That would be replacing the input valve with something of lower like ECC82, or maybe more acceptable, removing C2. That will cut down on loop gain, and restoring the 26 dB NFB would then result in perhaps a few 100 mV input signal, which might be more acceptable. I still have a Stereo 20 with me and will test that as soon as I can, for those interested.

For those interested, characteristics of the model with me :  Maximum Output: 14W under music conditions, -3 dB at 35 Hz and 22 kHz, Small signal -3 dB bandwidth: 7 Hz - 70 kHz, Damping Factor: 40 @ 1 kHz,  Output stable with both loaded and open output terminals etc.   
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline charles

Re: Leak Stereo 20 restoration/mods
« Reply #29 on: June 24, 2015, 04:04:41 PM »
Dear Ampdog,

The diagram in Stephen Spicer's book, "firsts in High Fidelity" is the same as the first diagram posted by afroaudio.

The other diagram is from the original original Leak manual I have:

"Stereo 20" POWER AMPLIFIER INSTALLATION AND MAINTENANCE INSTRUCTIONS
There is no date printed in the instructions manual.  The diagram in the instruction manual is the same as the second one posted by afroaudio.
I think the second diagram could be the earlier version as the feedback capacitors C9 are indicated as 0.0002 and on the other diagram 200p

I did not check if there are any differences between the two diagrams.