In my experience there are 3 levels of bias.
1/ too little and you get crossover distortion.
2/ Just right and there is no distortion.
3/ too much and power is wasted as heat.
This is not quite as straightforward as you might hope. I've recently built a JLH precisely to compare performance between transistors.
Many have written to point out that the theoretical operating conditions for 10W 8 ohm load is 0.8A at 27V (original design). Old 2n3055's had a very steep gain:Ic curve, which meant that at peak output (1.6A) a base current of X is needed. In quiescent conditions, the base current would be X/2 but the gain is then higher, so the quiescent current will be >0.8A.
For those I recommend 1.4A, but they are well past their sell-by date. They did have a low Vce(sat) without quasi-saturation, and would have worked at 27V with no problem.
With modern 2N3055's or similar epi-base transistors the quiescent current should be around 1.2-1.3A.
But I have also noted that these transistors exhibit quasi-saturation, which means, in the circuit as it stands, the supply voltage needs to be higher than thought as the current gain falls near the saturation point. In short, 30V is more like the rail voltage needed.
If using linear-gain transistors, then the current can be reduced to the near ideal, such as 1A, but I have measured fairly high quasi-saturation voltages which pushed the power supply voltage requirement up to 32-33V.
If folks find that their JLH has higher than expected distortion near or at full power, either the voltage or current or both might need to be increased.
There is no hard rule for this, as it all depends on the output transistors. I'd say 30V and 1.3A is a reasonable compromise, but can be optimised if you are able to measure the characteristics of the transistors you will be using.
Many have written to point out that the theoretical operating conditions for 10W 8 ohm load is 0.8A at 27V (original design). Old 2n3055's had a very steep gain:Ic curve, which meant that at peak output (1.6A) a base current of X is needed. In quiescent conditions, the base current would be X/2 but the gain is then higher, so the quiescent current will be >0.8A.
For those I recommend 1.4A, but they are well past their sell-by date. They did have a low Vce(sat) without quasi-saturation, and would have worked at 27V with no problem.
With modern 2N3055's or similar epi-base transistors the quiescent current should be around 1.2-1.3A.
But I have also noted that these transistors exhibit quasi-saturation, which means, in the circuit as it stands, the supply voltage needs to be higher than thought as the current gain falls near the saturation point. In short, 30V is more like the rail voltage needed.
If using linear-gain transistors, then the current can be reduced to the near ideal, such as 1A, but I have measured fairly high quasi-saturation voltages which pushed the power supply voltage requirement up to 32-33V.
If folks find that their JLH has higher than expected distortion near or at full power, either the voltage or current or both might need to be increased.
There is no hard rule for this, as it all depends on the output transistors. I'd say 30V and 1.3A is a reasonable compromise, but can be optimised if you are able to measure the characteristics of the transistors you will be using.
Last edited:
If I will find some time over the weekend, I may measure my amp at 0.7A, then crank it up to 1A, re measure, and go up to 1.2A if I dare and measure again.
I do not have much to measure with, just regular soundcard and RightMark audio software.
I do not care about distortion at full power, I never listen loud. I would measure at 1 watt.
I do not have much to measure with, just regular soundcard and RightMark audio software.
I do not care about distortion at full power, I never listen loud. I would measure at 1 watt.
Hi Folks,
this might help you when setting up your JLH and is what I do when experimenting.
Measuring the current
Put a 1 ohm or 0.1 Ohm 1% resistor in line with the supply cable and use one of those cheap battery powered 5-segment LED voltmeters that cost about a fiver from Ebay accross it. The voltage sense wire (yellow) towards the power supply and the negative sense wire (Black) at the amp side of the resistor. The red wire (power supply) is taken with the black wire to a 9 volt battery. This gives you a direct reading of the current(with the 1 ohm) or 1/10 of the current (with the 0.1 ohm)
Setting the current
Connect a dummy load of your choice (I use a 8 Ohm 50 watt resistor on a heatsink) to the output and a sine wave signal generator to the input.
Use an oscilloscope (or software monitor) to monitor the output. Turn up the input slowly until the output just starts to clip on the 'scope.
Adjust the current level up and see if the clipping goes away, if it does, turn up the input a bit more and repeat - always keeping a eye on the current meter. Do this until increasing the current has no effect on the clipping of the output. You have now set the amp up for maximum output power.
Turn down the input level and you will see the current increase to its no signal value.
Alternatively, if you use ARTA or the like, set up a spectrum analyser and adjust the current until you get minimum distortion before clipping occurs - say at half the output power. You can see this happening in real time when the 2nd harmonic suddenly drops down.
When doing the above, always keep an eye on the current meter and a thermometer on the heatsink in case something is getting too hot or failing. Talking of which I have always found it good to earth the heatsink on which the output transistors are mounted.
Hope this helps
Regards
Mike
this might help you when setting up your JLH and is what I do when experimenting.
Measuring the current
Put a 1 ohm or 0.1 Ohm 1% resistor in line with the supply cable and use one of those cheap battery powered 5-segment LED voltmeters that cost about a fiver from Ebay accross it. The voltage sense wire (yellow) towards the power supply and the negative sense wire (Black) at the amp side of the resistor. The red wire (power supply) is taken with the black wire to a 9 volt battery. This gives you a direct reading of the current(with the 1 ohm) or 1/10 of the current (with the 0.1 ohm)
Setting the current
Connect a dummy load of your choice (I use a 8 Ohm 50 watt resistor on a heatsink) to the output and a sine wave signal generator to the input.
Use an oscilloscope (or software monitor) to monitor the output. Turn up the input slowly until the output just starts to clip on the 'scope.
Adjust the current level up and see if the clipping goes away, if it does, turn up the input a bit more and repeat - always keeping a eye on the current meter. Do this until increasing the current has no effect on the clipping of the output. You have now set the amp up for maximum output power.
Turn down the input level and you will see the current increase to its no signal value.
Alternatively, if you use ARTA or the like, set up a spectrum analyser and adjust the current until you get minimum distortion before clipping occurs - say at half the output power. You can see this happening in real time when the 2nd harmonic suddenly drops down.
When doing the above, always keep an eye on the current meter and a thermometer on the heatsink in case something is getting too hot or failing. Talking of which I have always found it good to earth the heatsink on which the output transistors are mounted.
Hope this helps
Regards
Mike
Hi gents, i did have some time to measure distortion vs bias current. It responds positively. The higher the current, lower the distortion.
I did not insert any resistor into power supply, i got plenty multimeters measuring current up to 10A.
My distortion numbers are plagued by some ripples in the power supply, there is clearly 60hz, 120hz and up ripples, which add to overall distortion numbers, but still, since the same power supply was used for comparison, the conclusion is valid.
I measured at 0.7, 0.8 and 1.0 A per chanel. Distortion at 2.8 volt output was 0.13%, 0.126% and 0.11%. Purely second harmonic. No wonder it sounds great.
I will beef up power supply with more caps, or do regulated supply. I will see.
Cheers.
I did not insert any resistor into power supply, i got plenty multimeters measuring current up to 10A.
My distortion numbers are plagued by some ripples in the power supply, there is clearly 60hz, 120hz and up ripples, which add to overall distortion numbers, but still, since the same power supply was used for comparison, the conclusion is valid.
I measured at 0.7, 0.8 and 1.0 A per chanel. Distortion at 2.8 volt output was 0.13%, 0.126% and 0.11%. Purely second harmonic. No wonder it sounds great.
I will beef up power supply with more caps, or do regulated supply. I will see.
Cheers.
Has anyone tried for Q3 2SD438F or some similar transistor with high hfe instead of 2N1711 / 2SC3421 / BD139?
https://cdn-reichelt.de/documents/datenblatt/A100/2SD438_SAN.pdf
https://cdn-reichelt.de/documents/datenblatt/A100/2SD438_SAN.pdf
In complementary pdf, number in parenthesis represents the PNP. So 10pf is typical for NPN 2SD438.
I did try BC337-40. It seemed to cope. I had to fit 33 pF to get complete stability. I think if I had tried 12 pF it might have worked as the instability was only just detectable. I am in the process of moving house so will restart building a JLH when my new workshop built. For an 8 ohm version it seemed happy. BC327/337 are easier to get and seem genuinely able to to do more other common choices. I would imagine a genuine 2SD756 would be very good. BC327/337 are very low noise, high gain and good current. Distortion using BC337-40 seemed to better common examples. I think below 0.05% 5 watts is possible even at 10 kHz.
With Cob of 2.4 pf I guess an extra capacitor will be required. Gain ideally is > 200. Ft > 50 Mhz would be plenty.
This is a slightly better BD139 type of device ( Or BC639 ). The capaitance being 20 pF probably is ideal. Y spec gain > 160. Often TV types have nice audio qualities. I was told gun linearity demands the same things. This device has a very usefull 160V spec.
KSC2383YTA | ON Semi KSC2383YTA NPN Transistor, 1 A, 160 V, 3-Pin TO-92 | RS Components
This is a slightly better BD139 type of device ( Or BC639 ). The capaitance being 20 pF probably is ideal. Y spec gain > 160. Often TV types have nice audio qualities. I was told gun linearity demands the same things. This device has a very usefull 160V spec.
KSC2383YTA | ON Semi KSC2383YTA NPN Transistor, 1 A, 160 V, 3-Pin TO-92 | RS Components
I happen to have an ST 2N1711, which I did not know was in a transistor pack - one of the last items I bought from Maplin! Its Cjc measured at 18pF and Cje 50pF. The collector capacitance may drop to about 4pF at 10V, depending on the collector-base doping, so the question is what voltage is Motorola's 4pF specc'd at? Many quoted capacitance values were based on 10V bias. The capacitance changes so much with voltage it is important to specify the voltage it is measured at too (and ideally the power law value so we know its working range).
Last edited:
I have not read all the posts exhaustively so I hope I am not duplicating material but on the question of output current JLH was taken to task in the WW Letters to the Editor section after he published the original 1969 article.
A correspondent said he had set it too high (theoretically) and as such was doing the cause of Class-A amps a disservice.
In reply JLH conceded that the maths for the 'correct' theoretical calculations given by his critic were in fact correct but he (JLH) had deliberately run about 300ma more through the output stage for three practical reasons. I can't recall them all but two were (i) to allow for asymmetrical waveforms and (ii) non ideal speaker impedance. I can't remember the third.
So the 1.2 amps in the original article was what he thought was an optimum current for real world situations 'erring on the side of caution'.
But as usual JLH was actually across both the theory and practical aspects of the design.
Cheers Jonathan
A correspondent said he had set it too high (theoretically) and as such was doing the cause of Class-A amps a disservice.
In reply JLH conceded that the maths for the 'correct' theoretical calculations given by his critic were in fact correct but he (JLH) had deliberately run about 300ma more through the output stage for three practical reasons. I can't recall them all but two were (i) to allow for asymmetrical waveforms and (ii) non ideal speaker impedance. I can't remember the third.
So the 1.2 amps in the original article was what he thought was an optimum current for real world situations 'erring on the side of caution'.
But as usual JLH was actually across both the theory and practical aspects of the design.
Cheers Jonathan
Last edited:
A long thread like this is bound to go round in circles. In post 5323 I explained the reason for requiring a higher than theoretical quiescent current. Nelson Jones wrote the letter about near-ideal current but failed to take account of gain variation with current in the transistors. JLH nearly got the cause right but again did not quite manage to say so. But his suggestion was practically speaking correct.
Member
Joined 2009
Paid Member