JLH 10 Watt class A amplifier

Jump to Latest
#3,521
The Yamaha CA610 is a very modestly rated Class AB amp from the 1970's.

The diagram I have is a little different and shows a 27-0-27 transformer with an additional 0-14 winding for panel lights etc.

Taking this as an example and realistically you are probably looking at only being able to draw perhaps a couple of amps at most continuously from the main windings.

The power consumption on the label also seems to take into account the auxiliary AC outlets that this amp has, a switched 150 watts outlet and two unswitched of 200 watts total. .
 
#3,522
john_ellis said:
Hi Michael
You also have some interesting points. My main reason for aggressive square wave testing is just to see what happens in terms of the transient, stability and so on.
Regarding the JLH original design, 1.2A is suitable as I mentioned, but this corresponds to the base currents in the output transistors being half the required peak value. It is clear that you do not want the transistors to cut off nor saturate. Hence increasing the standing current should in principle reduce the distortion because of a less-non-linear excursion of the current. Unfortunately the older transistors will have less gain at higher currents so there could well be an optimum around 1.2A.
I have doubts even whether a 1 uF capacitor is useful other than to check the performance. The problem with the JLH 10 W is that if the current is limited to 1.2A and this leads to a peak output in the range of 1.6A, the slew rate in a 1uF capacitor will be only 1.6MV/s, beyond which we will get transient distortion because one or more transistors will be cut off. As you say a 2uF capacitor will see these effects at 10kHz.
Your 2N3055 design simulates with a similar result to my simulations.
This brings me to consider the transient intermodulation distortion issue again, but this time without a capacitive load which will cause it as we have seen.
With 2N3055 (3MHz type) the basic JLH amplifier starts to struggle at 50kHz sinewave input. This is acceptable (probably) for most signal sources, in practice. The driver transistor appears to cut off temporarily during the negative excursion (at 12.6V peak output and 27V rail) while the input transistor current ranges from 100uA to 800uA and would be generating quite a lot of second harmonic.
The faster variant I described earlier with MJL3281A's manages to reach 100kHz before showing the same symptom (without the RC Zobel) to minimise loading at high frequencies).
Your "fast" JLH with the emitter follower VAS buffer also suffers from internal clipping at high frequencies, with a 2uF load. Worryingly it shows a rising transient, which is admittedly faster than the original circuit, without ringing but a falling edge which has a significant ring to it. In the falling edge transient the emitter follower current is cut off.
And this is with JHL's 4.7k-470pF input filter.
Once again, this is probably too strict a test but a 2uF capacitor is also a too heavy load for this amplifier.
Without the capacitive load the amplifier performs quite well.

In the meantime I also explored an alternative design using a complementary output stage. Rather than using an emitter follower buffer, I used a Darlington VAS. I then increased the base resistor of the new Darlington transistor to 22k and increased the PNP input current to 2mA to increase its linearity by reducing the current variation it sees. The feedback resistors were changed to 1.2k and 100 ohms, but then a 1k degeneration resistor was added. This allows the use of a 33pF Miller capacitor to be used to stabilise the amplifier, which normally I would not use but in this arrangement, the frequency response is well controlled and the amplifier highly stable.
At 20kHz square wave into a resistive load no transistors cut off nor saturate for either signal edge. The distortion at 20kHz sinewave is 0.02%. This would probably address most of my concerns and at the same time could be retro-adaptable on most existing original JLH's, but some components might need to be squeezed in.
There is still an issue with a 2uF load but unless the output current can be made much larger (when it would move into normal driver+output design territory) the original will probably never be able to drive too much capacitive loading. As most listening levels for music tend to be around 1-2W, it is probably not a big issue for this design, which, as you say, was competing with valve amplifiers of about the same power when JLH designed it.
Click to expand...

Hi John,

I am taking advantage of some better weather to be in the outdoors. I will return to this interesting discussion.

Michael
 
#3,527
I have a question about the gain of the amp.
I need about 10W and my main source is 2V (dac) I do not intend to use phono, etc. So I need less gain that the amp has.
I look at the 96 version, is it possible to lower the gain eg. changing R6 or does it have any bad impact of the amp?
 
#3,528
The gain of the amp is pretty low by modern standards, you should be fine with it as it stands tbh.

If anything, you might feel the need for more gain if playing a quiet track. 2vrms from a DAC will refer to maximum possible output. A typical music track will be way below that level.
 
#3,529
Average voltage levels from a CDP will be around 10% (-20dB) to 30% (-10dB) of the 0dBfs signal.

If you know your CDP and other digital sources are sending out a 2Vac voltage for 0dBfs, then you should be measuring very roughly 200mVac to 600mVac as average levels.
 
#3,531
john_ellis said:
Hi Michael
You also have some interesting points. My main reason for aggressive square wave testing is just to see what happens in terms of the transient, stability and so on.

There is still an issue with a 2uF load but unless the output current can be made much larger (when it would move into normal driver+output design territory) the original will probably never be able to drive too much capacitive loading. As most listening levels for music tend to be around 1-2W, it is probably not a big issue for this design, which, as you say, was competing with valve amplifiers of about the same power when JLH designed it.
Click to expand...

Hi John,

I have simulated the 15 watt 1996 circuit - using 2N3055 and the fast MJL3281 transistors.

I used Hood's component designations except for substituting fixed resistors for RV1 and RV2 in his article.

A small change I made was the inclusion of a .22 ohm buffer resistor at the output - an improvement Hood adopted latter.

I used an initial voltage setting of -.6 and 6 - if the level is raised above this the load demands more current than the 2A capability of the output stage.

The voltage will rise a little in response to the increase in stimulus but the wave-form will lose its shape.

I increased the signal voltage setting in steps to -7 and 7. There is a small glitch at the back heel in the .raw file.

One of Hood's prescriptions for stability was a settling time of less than 20uS on square wave outputs.

I will have to send my attachments later that function is not working at present.

Michael
 
#3,534
Here are models of the fast1996 Class A and for what it is worth my variation on the 1970 Class AB - both of these using revised models for the output transistors.

The sim results are what they are - For now I am not deterred from using the 1996 Class A amplifier with 2N3055 outputs. There are quite a few models of devices that people might want to try for their own simulations.

I will think about changing to the fast outputs later.
 

Attachments

#3,535
Well I tried my first class A amp last night, a DIY JLH 1969 that I purchased from a local classifieds. I am impressed. I prefer to a couple of my other tube amps. It's an engaging sound but I felt a bit of upper mid emphasis that lead to some fatigue. Is this a trait with these amps? I will listen again tonight then try to swap in the PNP boards I purchased on ebay
 
Last edited:
#3,536
frangus said:
It's an engaging sound but I felt a bit of upper mid emphasis that lead to some fatigue. Is this a trait with these amps?
Click to expand...

"fatigue" is not something usually associated with JLH 69s.

My immediate response (which is likely wrong c.f. Mencken) is to wonder whether there's a mismatch with your speakers. Second thought is to wonder about the state or type of the output capacitors. Like parafeed amps these are in the signal path.
 
#3,537
Thanks thoglette

It was a speaker issue which was fixed with a change to some component values in the crossover. I have been enjoying the sound of this amp very much over the last few nights. I see why folks are still building it and talking about it so many years after its inception.
 
#3,538
i bought some small boards with tip41C output transistors,(1969 design) i changed them for 2sc5200 (selected for almost the same hfe. the boards are giving 10w eacht, so i would set 2A idle current @24V. (single rail, i have big heatsinks (o,28K/W) weighting 3.5kg per piece.
I would like to have 20w, so 4A idle current, is that possible? (the output transistors are sufficient and the heatsinks are capable too...
 
#3,540
but i use a big heatsink per amplifier because it seems that they become otherwise very hot. Is 48w per transistor too much if they can handle 150W? I know it's stupid, but i'm used to build tube amplifiers and i'm allergic to amplifiers with too much active and passive components because the sound will not become better.... and the jlh from 1969 seems to be cool (hot 😛) because it's still popular even after 47 years...
 
You must log in or register to reply here.
Top Bottom