Distortion in the Leak TL/50 plus amplifier

[GrimJosef]

At the risk of starting another long thread concerning the idiosyncracies of the Leak TL/50 Plus amp, I was struck by Terry's comment from this thread https://www.vintage-radio.net/forum/...=156071&page=3 at post #52:

Originally Posted by Valvepower

... I've always struggled to get the quoted THD figures for the TL50 – from memory I don't think I ever have

I've taken one of these amps (the one mentioned in the other thread which doesn't have the prone-to-oscillate output transformer), populated it with new old stock EF86, ECC83 and GZ34 valves (all Mullard) and KT88s (both GEC) and run it at 1kHz CW using an HP8903A audio analyser as well as an HP3561A spectrum analyser. The amp had previously been rebuilt using a mix of new and original (tested) capacitors and resistors. It was operated at maximum gain with the output transformer plugged for an 8ohm load and with an 8ohm resistor bank acting as that load. So the test conditions should have been close to perfect.

The HP3561A let me look at the output from the amp, broken down into the 1kHz fundamental and its higher order harmonics. The results are shown in the attached table for different measured output powers. The figures in the centre of the table are in dB. They show the sizes of the 2nd to the 9th harmonics (H2 - H9) relative to the fundamental. The right hand column shows the total percentage harmonic distortion calculated by adding H2 to H9 together. These results were in close agreement with the total distortion plus noise measured by the HP8903A.

The figures in black were produced when the amp was running normally. The ones in red were produced when it was running with its internal negative feedback loop disconnected. The bold figures are the differences between with and without feedback. The feedback level in the amp's specification sheet is 23dB. If all of the distortion was being created within the forward-going part of the feedback loop, and if the loop was correcting it effectively then we would expect all of the bold figures to be 23dB.

The other important specification for this amplifier is that at 45W output the harmonic distortion should be 0.1% or better. We can see from the results that this spec is far from being met. In fact the harmonic distortion is already above 0.1% at 10W output. Until the very highest powers are reached, the total distortion is completely dominated by H2 and H3.

A distinctive feature is that while the negative feedback is correcting the third harmonic by something like the hoped for 23dB, it is not having nearly such a large effect on the other harmonics - the even order ones H2 and H4 in particular. If it had been able to suppress H2 by 20dB or so relative to the red figures then the 0.1% spec would have been met, more-or-less, all the way up to 20W output.

To be fair, H2 and H3 are not very offensive audibly, so the sound quality from this amplifier would probably be perfectly acceptable if it was feeding just a single pair of speakers (above 10W through a single pair of speakers the listener's ears might well not be very linear either). But a spec is a spec and I'd be curious to know how much effort it might take to deliver 0.1% at 45W from a pair of KT88s. I see that in GEC's relatively short book An Approach to Audio Frequency Amplifier Design (1957) their 50W ultralinear amp based on these valves claims to deliver only 0.2% distortion, and the graph Fig 5-3 in the relevant section indicates rather worse performance than that (the section is number 5-2 of the book, which can be downloaded here http://www.tubebooks.org/technical_books_online.htm under Audio).

Cheers,

GJ

[G8HQP Dave]

Distortion which feedback does not improve can come from common-mode problems when the input and feeback are applied to different ports in an amplifier stage. I don't know this particular circuit, but it may use the common 'input to grid, feedback to cathode' arrangement. If the relevant valve has a poor bias point then some common-mode distortion can occur and feedback will not reduce it.

[GrimJosef]

The amplifier's circuit is here http://www.44bx.com/leak/Leak/Circuits/tl50_ecc83.JPG and yes, it does send the feedback signal back to the input EF86's cathode.

I've also come across failure of feedback to deliver a clean signal which was caused by nonlinearity in one of the two feedback divider resistors. The one which saw almost all of the output voltage tested OK with a volt or two across it but behaved quite badly when that became tens of volts (peak).

Cheers,

GJ

[Trevor]

There is another version of the circuit I am sure that used ECC81 as the phase splitter -driver it also had 600 ohm cathode resistors in order to reduce output valve dissipation
Trev

[GrimJosef]

There is indeed. The cricuit's on the same site as the one above http://www.44bx.com/leak/Leak/Circuits/tl50Plus.gif. In fact the amps I have are fitted with 600ohm KT88 cathode resistors for precisely that reason.

Cheers,

GJ

[Trevor]

I have been lucky and over the last 50 years have owned lots of quality amplifiers ! But the ones I really wanted a pair of TL50's were always just out of reach . This explains my interest in these units . I have serviced them as mono units used for sound re-enforcement so I know the units quite well They were always powerful but never had the critical acclaim of the TL12 and TL25
I think that the phase splitter did not have enough drive for a pair of KT88 in ultra linear

[Trevor]

Have you tried swapping the output valves around as I honestly believe that the lack of ac balance drive can increase distortion if the output valves are not an accurate match
Please try this Also can you check the grid drive to the output valves
Trev

[GrimJosef]

I have swapped the valves around, within reason (they are original GECs and the amp's owner only has four of them, so we are where we are with that, although my 'cooking' pair of modern Svetlanas give much the same result). They are well-balanced for quiescent Ia - only about 2% different - so there's good rejection of power rail hum in the output transformer. I can't easily measure gm in situ but it is quite often correlated with Ia so the gm mismatch might not be too bad. I also tried two or three different ECC83s since a mismatch in gm can be compensated to an extent by a balancing mismatch in drive to the two valves. The results above were taken with pretty much the best combination of valves from the ones that I had, bearing in mind that I needed the other pair of KT88s to be close enough to use in the other amp.

I disconnected the negative feedback and put a high impedance probe (3M3 in series with the 91k input resistance of the test gear, so 37:1 voltage attenuator) across one of the 330k KT88 grid leaks. This showed me the drive signal to one KT88. Running at 20W the second and third harmonics out of the whole amp were 52dB and 37dB below the fundamental respectively. At the KT88 grid my monitor showed they were 73dB and 53dB down respetively. So practically all the distortion at that power seemed to be arising in the output stage. Of course a more thorough measurement would have checked the other ECC83 triode as well. It's theoretically possible that the triode I tested was near perfect and the other one was horrible, although they are cross-coupled at their cathodes, with one driving the other there, so I would have had to have been unusually unlucky both to pick on the good triode and also for it to be the input valve of the pair, not the 'second' one. I should repeat these measurements more carefully though. As far as level goes, the drive signal at one KT88 grid was 21.5V RMS.

The last thing worth saying is that while it is possible to improve distortion performance by carefully selecting valves while measuring the distortion, that's not the sort of thing that a 'normal' owner would be able to do. So back in the day the manufacturers argued, when they could, that their specs were not only impressive but also that they could be achieved by the man in the street. To quote Harold Leak from one of his brochures for the TL/12 Point One amp

It should be noted that the TL/12 is a design not dependent on "matched" valves for its fine performance, nor does the amplifier rely on "balancing" controls. The use of either of these devices in small power amplifiers is indicative of design deficiencies ...

Cheers,

GJ

[Valvepower]

Hi GJ,

I'll keep the pair of TL50's the pair on the bench a short wile longer just in case you need do some comparative tests.

These use ECC81 phase splitter and 560R metal clad resistors mounted above chassis - not that elegant, but they do the job.

Terry

[GrimJosef]

I would be interested to know any distortion figures from yours Terry. Do you know if any of the reputable mags (HiFi News, Gramophone, maybe even WW etc) reviewed the TL/50 back in the day ? The measurements they made were often quite thorough and independent. I wonder sometimes if the absence of a commercial amp from the review pages meant that the manufacturer knew it wouldn't do well in someone else's hands and so didn't submit it.

Cheers,

GJ

[trobbins]

I'd suggest you disconnect feedback and attempt to adjust a few key variables, and try and identify if a distortion minima is identifiable.

Adjusting AC gain - perhaps by varying either R9 or R13.

Adjusting PP balance - perhaps by varying either R19 or R20.

Swapping KT88 (and repeating above).

Somewhat similarly, the raw distortion of the Williamson is dominated by the output stage, and tube rolling and related adjustments can make very noticeable differences. The 0.1% at 45W may well have been by just new (back then) tube rolling, or with 6550, so you may not achieve that now, or with just 4 KT88 samples.

The spec doesn't identify a particular load - perhaps there is some distortion difference when loading a different output tap.

[Richard]

Hi

I have a pair of TL50+ amps in my "to do" pile, they have been there about 18 years, doubt I will ever get around to doing them :-(

From memory of many hours searching the web there is much criticism of the EF86 input stage, and various mods, some just to reduce the gain, some using alternative valves, sadly I lost all this info in a recent computer failure, but I expect the sites are still out there.

Richard

[Valvepower]

Hi,

I meant to post the THD and power output graphs for the TL50's in this thread, but I posted it in the wrong place and I posted it in the TL50 output transformer thread

It's here.... post 114....
https://www.vintage-radio.net/forum/...=156071&page=6

Terry

[GrimJosef]

Thanks for that Terry, very informative. Yours are significantly better than the one I've looked at - 0.1% at 20W rather than 0.1% at ~10W in my case, although they get more similar at the highest powers. I was running with the 8ohm tap into an 8ohm load, rather than the 16ohm one. I might try the latter since that wouldn't leave any unused secondary turns on the transformer. It would also take the NFB direct from the speaker terminals, not from a different tap on the secondary. You wouldn't expect those things to make a large difference though (would you ?).

By the way, can I ask what colour yours are - champagne (early), grey (middle) or bronze (late) ? If the ID numbers are the hand-labelled ones inside the chassis then 61 and 63 are probably the years.

Cheers,

GJ

[G8HQP Dave]

It would also take the NFB direct from the speaker terminals, not from a different tap on the secondary. You wouldn't expect those things to make a large difference though (would you ?).

It could make some difference, especially if the OPT was less than perfect. The one thing you don't need is distortion within the feedback network, because then the feedback ensures that this distortion appears in the output.

[GrimJosef]

Originally Posted by Trevor

Here is a thought what about the properties of the nearly 60 year old lamination's ? Has the core changed its structure Would heating and cooling of the lamination's change the distortion figure
All other components show sign of aging
Regards Trev

The above post appeared in the oscillation thread. But since it refers to distortion I've posted my reply here.

A few pages back there was some discussion of whether the transformer might have been 'cooked' in use. A bit later I did note that the amp this one came from had been modded to use solid-state rectification, which might have taken the HT voltage up beyond the KT88s' comfort zone. In that case overcurrent through the transformer can result. There was no darkening of the insulation layers, or evidence of all the wax having melted away though.

Occasionally I wonder whether an output transformer's lams might have picked up some permanent magnetisation, and what the consequences of that would be, and how I might check.

Cheers,

GJ

[kalee20]

Unlikely IMO. If iron alloys changed significantly in 60 years, we've got a lot more to worry about than GrimJosef's friend's TL50's - the Forth bridge, the Eiffel Tower, etc...

The magnetic properties are dependent on crystal structure as is tensile strength, Young's modulus, so physical attributes and magnetic ones tend to go hand-in-hand.

Iron does change its characteristics when it rusts, but assuming relatively rust-free it is stable. It takes heating to several hundred degrees C to do anything (unlike ferrite, which drops Bsat dramatically above 100 deg C).

It's well thought of, and certainly it doesn't pay to take anything for granted! But a change in core would be well down on the list of possibilities. Far more likely is a dodgy set of lams in the first place!

[Trevor]

sorry re posting in the wrong area i realized this 10 mins after i had done it !
I want to believe it worked correctly when new such is my respect fo Leak
TREV

[GrimJosef]

Originally Posted by G8HQP Dave

It could make some difference, especially if the OPT was less than perfect. The one thing you don't need is distortion within the feedback network, because then the feedback ensures that this distortion appears in the output.

Well somewhat to my surprise it has made a difference. I re-plugged the output transformer for 16ohms, attached a 16.3ohm resistor and re-ran the harmonic distortion measurements. The results from my amp are now essentially the same as Terry's measurements on his - 0.03% at 1W out, 0.06% at 10W and 0.10% at 20W. I can't conveniently go higher in power because I don't have a 16ohm resistor array I trust at CW powers higher than this, but comparing with my previous measurements I've no reason to believe that we'll learn any more at higher powers.

The striking thing is the second harmonic. Its behaviour was tricky to explain when I was taking the output from the 8ohm tap - the feedback didn't seem to be correcting it very effectively. But when I switched the load to the 16ohm tap things were much improved. H2 dropped 3dB at 1W out, more than 7dB at 10W and it nearly disappeared at 20W. I guess there is some anomalous behaviour on the unloaded section of the secondary between the 8ohm and 16ohm tappings which is, of course, included in the feedback path.

At least we now have three different amps in two different places (here and at Terry's) all behaving in much the same way.

Cheers,

GJ

[Valvepower]

GJ,

They are Bronze and yes, I also feel the last number in the serial number is the year of manufacture. I was told the first number is the number of the amplifier and the second is a works number and the last the year of manufacture. I’ve third TL50 with 389/37/63 dated 21/8/63.

I’ve always thought it's.... champagne (early), bronze (middle) and grey (late)?

I’d swapped the valves around to get the best THD – the ECC81 is a Brimar 6060. I used Welwyn MFR5 1% resistors in the anode load positions.

I’ve always felt 16 Ohm gives the best performance – as you said it comes directly from the output and encompasses all the windings in the output transformer.

I’ve always felt the TL50 wasn’t the best sounding Leak amplifier (although they make good Heavy Metal and Reggae amplifiers ) and didn't sound as good as the TL25, where the latter TL25 with the 8615 output transformer being my favourite of the two TL25 output transformer variants.

See the instability thread for a plot of the phase response of both UL taps on the TL50 output transformer.

Regards
Terry