JLH 10 Watt class A amplifier

[paulb]

I recently resurrected my JLH and am enjoying it immensely again.

 

[yue1970]

Paulb,

I happen as well to restart my work on 2 monoblocks of JLH - post #1995
I'm planning now to introduce XLR sockets and interconnect both of them with a DAC.

At the time, I'd been however troubleshooting an earthing issue before other matters took me away from this project.
Seems like I'm using the same PCB than yours, hence I'll be curious to see what overall earthing schema you're using if it's ok for you to share the info.

Cheers,
Jean-Francois

 

[paulb]

My power supply is in a separate enclosure, actually on a chassis. This makes it easy to have a clean ground at the amp. It's quite a brute.

 

[yue1970]

Thanks Paulb.

I'll try to put the power supply outside as well during my next investigations to see if it helps.

Cheers,
Jean-Francois

 

[nautibuoy]

I wonder if I can borrow from the technically savvy amongst you.

I working on a JLH amp build using the SiliconRay PCBs. Based on information from several years ago in this thread, I plan to use MJL21194 transistors for the output devices - some of the common recommendation for the output devices (MJ15003, 2SC2830, etc) are getting hard to find nowadays but I know I can lay my hands on the MJL21194s.

I just have a sneaking suspicion that these devices might oscillate?

Can anyone confirm that the MJL21194 transistors are suitable.

Thanks

Ray

 

[Nisbeth]

IMHO they should be suitable, the MJL21194 aren't that much faster than the other transistors normally recommended here. Of course it could be that the amp oscillates, but had you substituted some other MJLs with a 30-odd MHz Ft I think that would have been much more likely.

/U.

 

[john_ellis]

I agree. A simulation suggests that it will not oscillate. As I pointed out several replies/years ago the issue with these transistors is that the rise time is limited by the ft. The MJ21194 is better than the 2N3055 (4 MHz vs 2.5MHz) but for really good performance you need higher speed devices (e.g. 2sc5200 etc). I would recommend these but they will oscillate unless you add a 47pF compensation capacitor across the feedback resistor.

John

 

[Bigun]

The MJ21194 is better than the 2N3055 (4 MHz vs 2.5MHz)...

I'll be hoping that the 2N3055's are useable

http://www.diyaudio.com/forums/solid-state/268846-tgm9-my-version-jlh-69-class-amplifier.html

 

[nautibuoy]

Thanks chaps.

...but for really good performance you need higher speed devices (e.g. 2sc5200 etc). I would recommend these but they will oscillate unless you add a 47pF compensation capacitor across the feedback resistor.

John

that would be a 47pF cap in parallel with R8 on the attached schematic?

An externally hosted image should be here but it was not working when we last tested it.

Ray

 

[john_ellis]

Yes to both the last questions!
The current 2N3055 is OK for an output transistor but works better with a driver. When used in a bootstrap circuit as in JLH design, it can only pull up at its own speed. The epi 3055 is certainly better than the original RCA for this circuit!

Yes the 47pF capacitor should be wired across R8.

John

 

[john_ellis]

oop[s - I was thinking of the original circuit. BUt the difference in the rise times are small. THe attached diagram shows simulations of the amp with three output stages (a) using MJL3281A (2SC5200); (b) 2n3055 with current source and (c) 2N3055 with bootstrap. These are without the input capacitor, and cause overloading in the input and/or driver stages. With the input capacitor they seem fine but the rise time is equivalent to the slower signal. SHould sound OK unless the bandwidth is a limitation.

JOhn

 

[gannaji]

In a "letters to the editor" post a month or two after original article is published, JLLH offered solutions for instability.

 

[Bigun]

When used in a bootstrap circuit as in JLH design, it can only pull up at its own speed.

what does this mean? the upper output device would be pulling up the output when the phase splitter transistor current is decreasing, alowing its collector to become more positive and hence the base of the output device becomes more positive. Why is there a speed limitation ? I assume it is because there is an RC time constant somewhere that is associated with the output device. What R and what C do you refer to ? Do you mean the base capacitance is too high ? I thought that JLH was able to get good square waves at well beyond the audio range with his outdated devices.

With class AB amplifiers I might want to have lots of feedback at very high frequency (well beyond audio range) to tame crossover artifacts, but Is it so important with class A.

 

[nautibuoy]

Yes the 47pF capacitor should be wired across R8.John

Thanks John, just have to decide which way to jump now!

Ray

 

[Sailing]

JLH 2005, First measurements.
(I moved this from Vender Forum).
I have now done my first measurements on the JLH2005 class A amplifer.
Only the first channel, so far and with Power from lab Powersupply.

The result so far is very good I think ! PDF attached.

I use the Mj15003, but will perhaps also test with 2N3055 as I have lots of them.
Also interested if someone has used other BJT !


/Christer

 

[mjona]

Yes to both the last questions!
The current 2N3055 is OK for an output transistor but works better with a driver. When used in a bootstrap circuit as in JLH design, it can only pull up at its own speed. The epi 3055 is certainly better than the original RCA for this circuit!

Yes the 47pF capacitor should be wired across R8.

John

In his book on valve and transistor amplifiers Linsley-Hood commented about the transition frequency of the output devices (MJ480) of the 1969 circuit being 4MHz whereas those used at input and phase splitter were in the 400 MHz range and he observed "the circuit has an inbuilt dominant lag in its loop characteristics. This ensures the loop gain has fallen below unity (one) before the loop phase angle reaches 180 degrees. No additional HF compensation networks are therefore necessary to ensure complete loop stability even with reactive loads".

I cannot see any advantage in substituting 30 MHz devices or lead compensation in the negative feedback loop.

 

[mjona]

what does this mean? the upper output device would be pulling up the output when the phase splitter transistor current is decreasing, alowing its collector to become more positive and hence the base of the output device becomes more positive. Why is there a speed limitation ? I assume it is because there is an RC time constant somewhere that is associated with the output device. What R and what C do you refer to ? Do you mean the base capacitance is too high ? I thought that JLH was able to get good square waves at well beyond the audio range with his outdated devices.

With class AB amplifiers I might want to have lots of feedback at very high frequency (well beyond audio range) to tame crossover artifacts, but Is it so important with class A.

The purpose of C1 in the diagram attached to post 3033 is to serve as an ac. feed in positive phase with respect to the voltage amplifier transistor Tr3 - this is to increase the apparent collector load presented on Tr3 and with that an increase in amplifier loop gain.

Whereas Tr2 is biased on from a the dc. rail via resistors, the ac. via C1 also modulates Tr2 base voltage - enabling a slightly greater ac. voltage swing and slightly more power.

Looked at that way the speed of the bootstrap is tied to that of the output stage but in Class A stages both output devices are "on" all the time. There is no handicap in using 2N3055 devices there. Those normally stocked by suppliers will be the 2.5 MHz ones.

The MJ480 devices Linsley-Hood used in his 1969 version were 4MHz and relatively high gain. High gain output was important to the results he achieved. The same can be obtained with 2N3055's if Tr3 is a high gain device.

Hood recommended replacing 2N1613 with 2N1711. If using BD139 there will be advantages in measuring for dc gain and selecting the best.

 

[Sailing]

So far so good...

My heatsinks are rated 2,2 C/W and I will stay with + - 18 V, 1 A until my final heatsinks has arrived.

18 V and 1 A is 18 W for each output transistor.

With a Tamb of 21 C the heatsink temperature should be :

Ths=21+ (18 * 2,2) = 60 C.

When doing a simple measurement with a digital kitchen term. a measured 61,1 C after 45 minutes of operation.

I hope the calculations are correct ??

 

[AndrewT]

one 2.2C/W to each output device?

If Rth s-a times Power comes to less than what the manufacturer uses to measure/specify their heatsinks, then you must apply a de-rating factor (DF).
For 18W * 2.2C/W = 39.6
Expect DF ~1.1 to 1.2, using DF=1.15 gives deltaTs-a ~ 45Cdegrees.
Heatsink temp Ts = Ta+deltaT = 21+45 ~66°C
deltaTc-s = Rth c-s * Power ~ 1C/W * 18 = 18Cdegrees. you might be able to get Rth c-s down to 0.5C/W for a good insulator. Without an insulator 0.2C/W can be just about achieved for a metal to metal contact interface.
Tc ~ 18+66 = 84°C

De-rate the transistor by applying the formula
Pderating = Pmax * {Tcmax-Tc}/{Tjmax-25°C} = Pmax {66}/{125} = Pmax * 0.528
This last calculation is the graph shown in the datasheet P vs Tc

What about summer Ta? Or do you have air conditioning on then?

 

[Sailing]

Yes one for each output device.
The heatsinks I used are old and I found them in my garage ..
They are marked by me some years ago with 2,2 C/W. They could be made by Fisher in Germany ?
I have no de-rating factor, or at what power they are specified.
Is it usual to specify this ?? I can't find any information, so I use your 1,15 factor.

Anyhow, an unexpected "precision" between calculated and measured temp on the heatsinks.

I have orderred better heatsinks. Tamb will certainly be higher and there must be a margin. Also the heatsink temp should be below 60 C so that you can put your hand on the heatsink without burning yourself.

Thanks for your information

/Christer