JLH 10 Watt class A amplifier

If you use 18V rail voltage and lowish bias try to use one pair of power transistors.
Please take a look at the bias settings table, those power transistors to perform in Class A, to bring out the sweet point they need certain bias level. Sometimes better to use only one pair than two. For example, if the transformer only can handle low power etc.


Prasi: the 2SA970 transistors need to be rotated 180 degrees?
Thank you.:)

Here is my progress, please do not laugh at my input cap solution.
I ordered Clarity caps but they are huge so I use these Russian type, the same problem with the 1000uF 16V Elna Silmic they just too big. :eek:
 

Attachments

  • JLH power rating.JPG
    JLH power rating.JPG
    101 KB · Views: 796
  • IMG_0001.JPG
    IMG_0001.JPG
    114.8 KB · Views: 818
If you use 18V rail voltage and lowish bias try to use one pair of power transistors.
Please take a look at the bias settings table, those power transistors to perform in Class A, to bring out the sweet point they need certain bias level. Sometimes better to use only one pair than two. For example, if the transformer only can handle low power etc.


Prasi: the 2SA970 transistors need to be rotated 180 degrees?
Thank you.:)

Here is my progress, please do not laugh at my input cap solution.
I ordered Clarity caps but they are huge so I use these Russian type, the same problem with the 1000uF 16V Elna Silmic they just too big. :eek:

Hello Gabor,
yes, the 2sa970 transistor need 180 deg. rotation, like you shared the pic.

i/p cap mounting looks perfectly alright, it couldnt have been better. I have seen people use 2u2 jupiter caps, so big and expensive Copper Foil Paper & Wax Capacitors | Jupiter Condenser :eek:

regards
Prasi
 
Someone said total harmonic distortion or TDH is not a good measurement of an amplifier. In work this week I was asked a similar question and told my boss it's the square root of the sum of the harmonic squares. To test the idea I took the sum of squares to the 21 st term and got almost exactly 46% ( 0.46 ) for a known wave the squarewave. One formular for a squarewave is [(Pi^2/8)-1]^0.5 where Pi = 22/7. The answer is ( 1.2347-1 ) ^0.5 = 0.2347^0.5 = 0.484, using 355/113 for Pi would be closer to the stated value. At college 22/7 was the allowed value.

In a way we might exspect ( 2^0.5 ) - 1 to be valid. That is 1.414-1= 0.414 or 41.4%. This is related to the sine or cosine of 45 degrees being 0.7071 or 1/2^0.5.

There is another way to calculate it that gives 43.5%, 41.4, 43.5, 46, 48.3. My 46 % favours the 48.3%. Perhaps the harmonics > 21st = 2.3%.

A trianglewave is roughly a squarewave / 4. Or 12%. That makes perfect sense as it is a filtered squarewave and the Fourier series has squared values ( 3 becomes 9 , 5 becomes 25 ).

A sawtooth although having a similar RMS value as a squarewave has 2/3 more distortion. This seems to be just of the additional even harmonics. It sounds rough also. The calming influence of the even harmonics is not quite a calming as it should be. The quantity of everything being the fact I guess. Turn the volume down and a sawooth may sound nicer for a percieved volume. I have no idea and shows how subjective these things are.

To my mind this is just a convention and not fact. 41.4% seems more correct. All the same it is not far away to the 48.3% and is a standard we can follow. In real life we don't know the name of the wave and must calculate what we see. Our curve is a JLH curve. Someone said Monotonic, only roughly so although it will plateau at Ft. I haven't complicated this by saying 5th and 7th should be avoided if higher than 4th and 6 th. The JLH side steps that nicely and is it's virtue.
 
Sometimes you people make my head hurt following some of the theories, but keep it coming: I'm sucking it up like a thirsty dog! Thank you!

So, in order to drag it down to my level again, I have a question on power supply and one on DC protection.

Power supply question #1. I think I'm settling for 18v on the secondary side, so will be feeding the board with something like 26v. (details on AliExpress says 12-35v, but as a beginner, no need to push my luck). Now I've read somewhere in this here thread on a 2x20w model run off a single 160VA toroid (just, if I remember correctly), so the question is: How low can you go, not that I'm going to, but would it be possible to run a single 10w channel off a 60VA core?. 50?. Did I miss the formula for it somewhere in the thread?.
Keeping with the PSU question: How much uF is the absolute minimum in your power supply (again, not going there). My manuals tells me the 2 x 20w NAD 3020 has 2 x 4700uF. The "100w" NAD 2200PE has 4 x 10000uF but I assume Class A does things differently.

Finally, DC Protection, would it be a wise choice for a beginner like myself to put in one of these boards with delayed start and DC protection or would that be a waste of money?. https://www.aliexpress.com/item/2pc...on-C1237-UPC1237-Mono-DIY-Kit/1666082471.html

Again, thank you guys for making my head spin ;)
 
JLH output adventure

I will post here some measurements using different output transistors.
First using the long black Prasi's board.
Power supply=+/-18v
One pair out transistors.
picture 8 is the MJ15024 BOARD
picture 9 is the TTC5200 BOARD
 

Attachments

  • TTC5200 2.8V.PNG
    TTC5200 2.8V.PNG
    83.8 KB · Views: 289
  • 2.8V 6R.PNG
    2.8V 6R.PNG
    81.6 KB · Views: 180
  • STEPS 2.8V.PNG
    STEPS 2.8V.PNG
    71.9 KB · Views: 170
  • 6V.PNG
    6V.PNG
    84.8 KB · Views: 154
  • TTC5200 6V.PNG
    TTC5200 6V.PNG
    83 KB · Views: 157
  • STEPS 2.5A BIAS 6V RMS 6R OUT.PNG
    STEPS 2.5A BIAS 6V RMS 6R OUT.PNG
    65.9 KB · Views: 143
  • STEPS 6V.PNG
    STEPS 6V.PNG
    58.1 KB · Views: 131
  • IMG_20181112_122728.jpg
    IMG_20181112_122728.jpg
    993.1 KB · Views: 181
  • IMG_20181112_140108.jpg
    IMG_20181112_140108.jpg
    853.8 KB · Views: 173
  • STEPS 2.5A BIAS 2.8V RMS 6R.PNG
    STEPS 2.5A BIAS 2.8V RMS 6R.PNG
    62.6 KB · Views: 151
Last edited:
The 19 kHz might be something I wouldn't like.

One thing I forgot to say is if no AM band oscillation is happening the JLH should show the same distortion curve as another JLH clone. If gain is optimum the overal THD will be lower. The final harmonics should be lower in a more ideal way. A lot of that may be lost due to the need for stability. Self argues external capacitance is better if needing stability of a VAS ( cb capacitor of the driver ). That feels to be correct.

One way we can make THD look better is to remove second harmonic distortion. With a JLH this would be easy and a disaster. Why? The distortion is very low and as is rare the harmonics are in an ideal order. Douglas Self " seems " to say, if distortion is very very low any discussion is wrong. Whilst unhappy with that I side with him a bit. As he rightly says if there is a problem something of that problem will show in simple tests. Music is nothing like that testing. The JLH should not cause these problems from any point of view. Ironically the Self test works well with the JLH. Anything that seems wrong is wrong and most likely is layout, grounding, decoupling, excessive HF gain. One problem can be where a speaker is returned to the PSU and how. In very rare cases the obvious place will not work. The Quad 303 causes this problem as a performance upgrade. A very odd and sage design. The 405 isn't my cup of tea. There is a lot of absolute nonesense said about valve and transistor designs. Both Quad and Leak went the extra mile to make their transistor designs good. They were not stupid and did realise transitor amps were not easy to design. Even though they did hear things they were very unwilling to say so. JLH was the one who had no trouble saying it. He was also respected as technically able to state facts. I never knew anyone laugh at him, that's unknown for most.
 
19 kHz

That 19kHz is a little curious, as it does not seem to change with output (much). I would suspect it might be pick up from somewhere?
Perhaps an old FM radio nearby?
Yes agree,just to be more clear look this measurement when amplifier is off.
Last picture a sound card loop test.
 

Attachments

  • JLH OFF.PNG
    JLH OFF.PNG
    75.4 KB · Views: 650
  • LOOP.PNG
    LOOP.PNG
    75.3 KB · Views: 650
More about THD. There was said to be a transistor sound in the early 1960's. The Quad 303 being one that didn't have it. Thus when people got most excited about it the battle was already won. Both the Quad 303 and JLH have ideal distortion. Both designs are low distortion by any sensible measurement with the 303 being exceptional. Ironically both the Quad and the JLH could have been made lower by using a long tail pair input pair. In the case of the JLH I speculate that it could be <0.01% THD below 5 watts. Many designers then and now would not be able to say no to that idea. Some who were caught up in the transistor sound debate I suspect knew to remove second harmonic was not a great idea. Someone said to me two amplifers sound good. One with JLH distortion ( Jean Hiraga Le Monstre also ). The other with no distortion. I like that idea. Hiraga was the person to state this preference as a Hi Fi religion. He was always seeking correct science to back him up. I believed it from the first thing I read. A beautiful mathamatical curve was given and I saw it must be true. The Coral 777EX low cost moving coil said by Hiraga to have the ideal curve. The Supex 900 slightly less good. A bit like a Naim amp. That is 3rd and 5 th slightly higher up than would be ideal. Mark I say slightly. It gives a fast sound in the talk of the day.

My first amp was a Teleton GA202 which was mildly like a Leak or Tobey Dinsdale. I plan to make one for fun. Some say a Mitsubishi product. Could be as it was nicely made. Very British which makes me doubt it. It was very cheap.
 
Well, this is a little off-thread, but I would suggest that the 303 was not as good as thought. Not because of distortion so much as a limited bandwidth.
But regarding "transistor sound", there were certainly very poor designs around which were considerably worse than the 303. Some high power amplifiers had very slow output transistors (e.g. the hometaxial 2N3442, 2N3773) which would have caused slew rate distortion, and probably could not manage full power at 20kHz very well. Some circuits of the 70's were truly awful. Crossover distortion, insufficient base drive, leading to intermodulation distortion, you name it and these were probably the sort used in station announcer systems that everyone mocked at the time. Some of the criticisms were valid but it seems to me that good designs got tarred with the same brush.
The JLH is somewhat unique. Either earlier in this or another thread I mentioned that a differential input stage gave higher distortion than single-ended - therefore the initial conclusion is that distortion in the various stages are partially cancelling, not just second harmonic in the push-pull output stage. I have not delved into this in any depth, but that does seem possible.
I would not be sure about it improving a 303 either. Mainly because the open loop gain would halve unless a current mirror was added to join the halves up again, but that would require a more complicated redesign. The best improvement I made with a 303 circuit was to use higher speed epi base 2N3055's, reducing the compensation capacitors and doubling the bandwidth. Did need a trick to solve oscillation in the triplet stage.
 
But regarding "transistor sound", there were certainly very poor designs around ...
See Hamm, R. O. (1973). "Tubes Versus Transistors-is there an Audible Difference." Journal of the audio engineering society, 21(4), 267-273 which explains the "sound" differences between "well engineered" circuits quite well.

Yes, these differences only exist (in class A) when there's clipping but all Class A amps (and 99.99% of AB amps) clip.
(Unless your music is compressed to death but that's another thread altogether:D)

The JLH is somewhat unique. Either earlier in this or another thread I mentioned that a differential input stage gave higher distortion than single-ended
Yes, I recall that, and in particular the problems that much of the differential input stage is, by definition, outside of the feedback loop it creates.
 
The Quad 303 was a miracle in it's day. The Quad ELS 57 rises at 25 kHz if diven directly at the plates. Quad had the transformer wound ( Whiteley as in Stentorian ) to fall at 18kHz. The restricted bandwidth was most likely their prefered option. Often said to be a British sound.

I built a zero loop feedback valve amp using a modified VOX AC30 output transformer. It started to roll off at 7 kHz. It had no obvious treble loss. That was until feedback was added. Then it did sound dull. Ironically it was flat well above 20 kHz then. I know all the problems of feedback valve designs and they account for this.

The Quad 2/22 is a respected valve design. Quad insisted that it was impossible to know if the valve or 303 power amplifier was being used. I took the test and could not tell. As far as I know the statistics were no better than chance when some thought they could. It is very likely the 303 is super matched to the ESL 57 and some Celestion SL6 would prefer something brighter.

On the graphs above, both are good. It would be interesting to know which sounds best. The second might sound more open and faster. Neither should sound bad.
 
BTW. A 303 arrived for repair that had a very strange fault. The regulator was shorted so as to give > 80VDC instead of 67, the usual fault gives 22V. The transistors didn't kill themselves. There was a little bit of hum. On returning the amplifier the customer said " you've ruined the sound" he did agree hum was lower. Quad had a word with him and said whilst exciting it migh be a short life. I heard it myself and have to say it was a very different amplifier. It sounded much brighter and I have to say ideal.

I spoke with Julian Vereker of Naim about his NAP250. That had a very different sound compared with the NAP160. It was the power supply. Julian said the active power supply needs to be very fast if it is to be of advantage. Even the NAP250 can sound a bit dull.

The Quad 303 at 1 watt and 30 watts gives about -93 dB third harmonic and worse case - 80 db second ( better than stated ). 5 th is - 103 dB. 7 th - 108 dB.
 
Some more interesting points!
The Quad ELS 57 rises at 25 kHz if diven directly at the plates. Quad had the transformer wound ( Whiteley as in Stentorian ) to fall at 18kHz.
My conclusion is that this is just "good engineering" where the frequency performance of one part of the system is balanced by a fall in another part. It is of course difficult to get high power transformers with a high bandwidth anyway.
A 303 driving an ELS might well benefit from inherent higher treble response.
Subjective differences is difficult territory because it demands explanations. All I can say is that with a higher B/W in a "303" (upgraded) it sounded livelier than a direct comparison with an unmodified circuit using RCA 2N3055 output devices.
The 303 was often driven from a 33 pre-amp. That had quite a filter on the output (1.2k/2.2nF) which will have limited the bandwidth too.
My point being that while individually the bandwidths are over 20kHz, collectively this may have an audible limitation.
I am pretty sure I could not tell a 303 from my 20W ULA EL34 amp (which I wound the transformers for) which has a B/W of 18kHz but otherwise sounds "pleasing".
I'll have to fix the volume control on that to compare with a JLH using 2SC5200's sometime...