JLH 10 Watt class A amplifier

As more current is demanded the gain reduces. Often if a sensible 1 amp is required the gain typically can be 100 rather than 20 implied by spec sheets. The 2SC5200 I setup with 16 mA of base current was able to give one amp. The problem is there will be sample to sample variation.There was a downside in that I would expect bias stability to be less good. When a class AB it probably won't be dramatic and is under servo control due to the vbe multiplier bias ( rubber diode ). . PNP devices often have slightly better gain if 1 amp. Some say that the PNP version sounds better.

Input impedance is directly related to gain . The input impedance of the JLH will be seemingly too low to be driven by it's driver transistor. Reality shows it can work. All the same any improvement in gain is useful. High speed transistors generally have higher gain. The problem with saying high speed transistors sound better is there must be a reason. If not it's like saying a certain medicine is working in an unrecognised way. I don't do simulations because it's a hypothetical layout. All the same better than no measurements. Intermodulation distortion is the most likely. Even so valve amplifiers prove IM distortion doesn't always mean an amplifier is unpleasant to listen to.
 
Member
Joined 2010
Paid Member
MJ15015 / 2N3055AG current gain spec from On Semi datasheet. I don't understand this.
It's a kind of shorthand for the upper and lower Hfe limits of the semi it refers to. I if you look at the full datasheet of any power transistor, you'll see that Hfe varies with power, temperature etc. according to performance and derating curves/plots. There are lots of curves that could be important in different applications so when you mess with parts, it pays to be aware of their full range of specs. rather than the few typical figures you see quoted in one-line, tabular form.
Take a look at figures 1 and 2 here: https://www.onsemi.com/pub/Collateral/2N3055A-D.PDF
Also compare figure 2 with the brief numbers you quoted.
 
If you look at 25C and 150C of that device one could say gain of 100 plus at 1amp going down to 70 plus at 2 Amps. The fact that the 25 and 150C are not too far apart helps biasing. At very high current let's say for 50 watts class AB 1 ohms the gain can drop to 5. A compound 2N3055 Darlington was used. Up to 5 amps can be tolerated into the base of the real 3055. This compromises high speed. Quite remarkable how good they did sound in a Quad 303.
 
https://www.onsemi.com/pub/Collateral/MJ15003-D.PDF

Comparing to mj15003. Gain is better with 2N3055AG across the jlh current range based on the graph of typical. Table for mj15003 says max hFE is 150 but the curve of typical doesn't come close to that. Amazing how much variation in these things. Not many industries where inter product variation would be so high. Imagine buying car that might go 70 or might go 150. You'd want to test drive a couple before buying.
 
Last edited:
Hi,
Firstly, I have been using BC337-40 for TR3 so I'm interested to understand any concerns about the power rating. Assuming the gain of the output transistors is 100 then for Iq=1.2A, Ic(TR3) is a little over 12mA, say 15mA average. Vce(TR3) is out of phase with Ic and 13.5V average, say 15V. The dissipation is therefore 225mW peak which occurs when current and voltage are at their average levels. Average dissipation is a bit lower. If the device is good for 600mW then this seems like an OK margin? I've tried BD139-16 here but the devices I got from Farnell have a gain of no more than 170 and this isn't enough. There aren't many other alternatives..
For the output transistors I agree gain > 100 is needed. I bought some multicomp mj15003 from CPC and these were hopeless, with a gain around 20 at the required current. Buying On Semi devices from Farnell was more successful but still yielded <100 about 10 % of the time.
Cheers
Jon
 
It's a very big deal isn't it. I never fully understood the JLH with this. If you took the gain at 100C that's realistic. You can see if it gets to 150C it's going to runaway and unsolder the working parts inside the device. 150C might be only 60C at the ideal heatsink. And with class A it is what it is. Class AB can be turned down when less loud.

Did you notice PNP MJ15016 is vastly faster ? They have just numbered up something I think of the parts bin.

What I am working on is a three transistor buffer in class A/AB with op amp gain stage. The buffer with it's own feedback loop using current feedback. A LM833 is a nice gain stage. The bit I would like to do is set the DC via the op amp. No feedback between op amp and output. AC-capacitor coupled. 2 watts A 30 watts AB. 60 watts 3 ohms AB. I suspect 95% of the time it would be in class A. The buffer would have a little gain if wanting everything. The PE Rondo could do this. It has a lot of problems. I did make one work. I don't like the op amp being forced to do power transistor correction.

If someone says I used 60 MHz transistors and didn't have problems it is the same as saying I have never been hit by lightning. In the right conditions you will. If you feel a tingling in those conditions throw yourself on the gound. If under a tree walk slowly away foot to foot then throw yourself on the ground. The branches usually match the roots. The foot to foot is to stop a path to your heart via the roots. You could get burnt which is the smaller risk.


My JLH was sensetive to speed ( hence the RF graph, 1.2 A 27 VDC ). Although I thought I had the best possible layout perhaps I didn't. My amp was if you like a bit turbulant until fixed. Sure it worked. It wasn't right.


Rod Elliot if memory is right says there is a myth that older designs were totally stable due to single input transistors and simplicity. Rod says no, it's due to slow transistors. I built an example and he was totally right. There is a suggestion that almost unstable amplifiers sound best. I can see why. Over compensation is to damp the amplifer or make it slow.
 
Member
Joined 2010
Paid Member
If genuine Central Semi product, it will likely perform close to the original because those guys specialise in obsolete parts but if not, you may have something different which may or may not really matter too much. The 2N1711 in my amplifier was branded with the old Motorola "bat" device but it was also brand, sparkling new and printed with ink so its a fake but what could they really be? Well, some people have tested them and guess that they most likely contain a chip similar to 2SD667A or 2SD669A, now both very common Chinese copies of the obsolete Hitachi/Renesas audio drivers. A few Chinese manufacturers offer them in different packages, quite cheaply and good quality. They can also be found, along with their PNP complements, in many good audio amplifier kits. FWIW, I've built the JLH69 with one of them and did notice differences - albeit something of an improvement.

The Hitachi/Renesas transistors are also superior as audio drivers to BD139/140, which are old favourites too but the downside of using Chinese copied semis is that you don't know for certain that every source of them is up to scratch - some could be using different, inferior silicon and you have no easy way of testing without a direct, careful comparison in your amplifier. You may find they measure fine according to a little battery operated transistor tester and few mA current but how they fare with >1A audio is a different question.
 
I'm trying to understand why very high gain input and driver devices are so important if the originally specified ones didn't have gain that high.

I started looking for a different driver (to the BD139-16 @ 170) to reduce distortion at 10kHz and above. In my experience higher gain helps achieve this in measurement and in simulation. I haven't yet figured out why.
 
It's rather depressing to have so little choice. If using PNP output MJ15016 looks good. BC327-40 the driver if so and BC337 the input. BC327/337 are very low noise as a bonus. Still made by recognized brands. One of us thought the PNP version better. I would believe that. The power PNP had a difficult development which seems to have resulted in a better overall device. 2N2955 was seen as inferior. Possibly not in a JLH.