Yeah, JLH needs to "work" first. Though last you wrote,
was already working but fighting overheat. This will help.
2n2222 only chosen to show that any lame junkbox NPN
would probably do. No special stress upon that component.
VCE never more than 1.4V, current never more than 200mA.
Only critical spec is base to emitter threshold of roughly
0.66VBE, that goes down as it gets warmer. Almost any
silicon transistor satisfies this.
You would have to hunt down some 0.33R's for sure, if you
wish for precise 1A quiescent. But the transistor spec is not
critical at all. Choice not limited exclusively to 2n2222.
was already working but fighting overheat. This will help.
2n2222 only chosen to show that any lame junkbox NPN
would probably do. No special stress upon that component.
VCE never more than 1.4V, current never more than 200mA.
Only critical spec is base to emitter threshold of roughly
0.66VBE, that goes down as it gets warmer. Almost any
silicon transistor satisfies this.
You would have to hunt down some 0.33R's for sure, if you
wish for precise 1A quiescent. But the transistor spec is not
critical at all. Choice not limited exclusively to 2n2222.
Last edited:
Hi Ken,
I think we've misunderstood each other.
I'm not fighting overheat. With my larger heatsink (the bare aluminium one with both MJ15003s - the second photo), it takes a good half hour to 45 minutes to heat up - it is hot then, but can be touched. So, I guess its 55-60 C. I expect the heatsinks to get pretty hot. It seems to be reasonably stable and the DC offset drifts quite slowly so I don't think there's a problem for now.
I do want to improve it - I need to build a chassis, and I need to at least double my PSU capacitance for both channels, and rethink my wiring a bit. But I need to do that before I embark on other changes.
Having said that, your suggestions are interesting and worth noting for future reference.
So, does the 2n2222 current limit Q1, thus saving a bit of heat?
I bought several 5W 0.33 ohm resistors off of Ebay actually - the 1996 uses one so you can calculate Iq based on the voltage across it (and I had to buy a replacement because I fried one in a previous comedy attempt several years ago).
I managed to get the second circuit board up and running tonight (using TIP3055s), so I need to knock up the second heatsink with MJ15003s.
I think we've misunderstood each other.
I'm not fighting overheat. With my larger heatsink (the bare aluminium one with both MJ15003s - the second photo), it takes a good half hour to 45 minutes to heat up - it is hot then, but can be touched. So, I guess its 55-60 C. I expect the heatsinks to get pretty hot. It seems to be reasonably stable and the DC offset drifts quite slowly so I don't think there's a problem for now.
I do want to improve it - I need to build a chassis, and I need to at least double my PSU capacitance for both channels, and rethink my wiring a bit. But I need to do that before I embark on other changes.
Having said that, your suggestions are interesting and worth noting for future reference.
So, does the 2n2222 current limit Q1, thus saving a bit of heat?
I bought several 5W 0.33 ohm resistors off of Ebay actually - the 1996 uses one so you can calculate Iq based on the voltage across it (and I had to buy a replacement because I fried one in a previous comedy attempt several years ago).
I managed to get the second circuit board up and running tonight (using TIP3055s), so I need to knock up the second heatsink with MJ15003s.
"So, does the 2n2222 current limit Q1, thus saving a bit of heat?"
Q3 shapes a perfectly curved bias current total for both Q1 and Q2.
That is then steered by JLH toward the appropriate output device.
Output currents through resistors R1 and R2 will be linear, equal,
and complimentary. Regardless that drive currents may never be.
Current through output transistors are not expected to be exactly the
same. Q2 will always have to work slightly harder than Q1. Because Q2's
emitter does not face toward the load, but faces a negative power rail.
Base current of Q2 is wasted without direct effect upon the load, and
it's collector has to work exactly that much harder to make up the loss.
Thieving also has to bleed out through Q2, adding another few mA.
No big deal, the load will be fooled to see perfect resistor currents as-if
the resistors have become virtually matched output devices. Any sloppy
mismatching of transistors Q1 and Q2 is tolerated and corrected as well.
If I had to find the best place to mount Q3 on the heatsink, I would look
for a spot closer to Q2, as this end of JLH expects to work slightly hotter
than the other. Theiving will make both transistors work cooler overall.
Q3 shapes a perfectly curved bias current total for both Q1 and Q2.
That is then steered by JLH toward the appropriate output device.
Output currents through resistors R1 and R2 will be linear, equal,
and complimentary. Regardless that drive currents may never be.
Current through output transistors are not expected to be exactly the
same. Q2 will always have to work slightly harder than Q1. Because Q2's
emitter does not face toward the load, but faces a negative power rail.
Base current of Q2 is wasted without direct effect upon the load, and
it's collector has to work exactly that much harder to make up the loss.
Thieving also has to bleed out through Q2, adding another few mA.
No big deal, the load will be fooled to see perfect resistor currents as-if
the resistors have become virtually matched output devices. Any sloppy
mismatching of transistors Q1 and Q2 is tolerated and corrected as well.
If I had to find the best place to mount Q3 on the heatsink, I would look
for a spot closer to Q2, as this end of JLH expects to work slightly hotter
than the other. Theiving will make both transistors work cooler overall.
Last edited:
If you have 60Hz hum that worsens with Iq its probably supply. I now avoid regulated supplies in favour of a cap multiplier. It drops about 1.0v. Since instal, my '69 JLH has no audible hum at speakers or headphones. We all worry about our speakers and I have no qualms about using a large electrolytic (bypass switcheable) for output,even with a split supply. I can't tell which way the switch is set by listening! Switch in the cap for testing or expensive speakers.Thanks for the advice mooly, much appreciated.
I'm happy with the hum level, it is low really (I'm used to class B amps with virtually no hum at all), but I will investigate. I hadn't thought of lowering the bias (although I was keenly aware that I'd get more ripple when I up Iq).
Now running with MJ15003s on a bigger heatsink. Takes longer to heat up, but still gets rather hot. Can't really tell if it's better (possibly is) than TIP3055, as I'm listening in mono, but I can play with increasing Iq.
After an initial settling time of a couple of minutes, and setting the DC offset to close to zero, the DC offset rose by only about 2 hundredths of a volt - so does seem quite stable (I wonder if it will ever completely settle? I guess so).
Thanks for the link, I shall have a read...
Last edited:
I'm building the second heatsink (basically drilling some aluminium sheet and bolting it to a heatsink) for the second channel, then I'm going to redo the PSU.
I need to double the capacitance used, so it will be a good idea to investigate the PSU entirely. Hopefully will sort it out.
Interesting that the cap multiplier drops only a volt. Thanks for that. Might save some money on more caps. I think I can justify the loss of two volts between rails.
I need to double the capacitance used, so it will be a good idea to investigate the PSU entirely. Hopefully will sort it out.
Interesting that the cap multiplier drops only a volt. Thanks for that. Might save some money on more caps. I think I can justify the loss of two volts between rails.
Who did it first?
Hi,
Are there any patents on the JLH amplifier topology, other than this one Low output impedance feedback power amplifier ?
Thank you.
Hi,
Are there any patents on the JLH amplifier topology, other than this one Low output impedance feedback power amplifier ?
Thank you.
Hi and sorry for the old reference but I wanted to ask something concerning the post #990 of Graham. I recently build a JLH (2005 UPDATE) with an enormous capacitor banK but I can not rid off the small ripple. Thats why I found this post (#990) and I want it ask if these solutions are applicable to the 2005 UPDATE circuit (+- rails) with the same value of components, also when the post says the first transistor means the input transistor or the first output transistor. Sorry if this is outdated for anyone, but i couldnot find the answer for my questions.
Thanks.
Another way to reduce the ripple voltage is to use a toroidal inductor between the bridge rectifier and the reservoir capacitor. I've yet to try it, but I've noted recently that some of the better hifi amplifier manufacturers talking about doing it in their equipment. They usually call it a power supply choke.
You also need to distinguish inherent hum from hum induced on the input wires. If I remove the input wires from the PCB, there is no noticeable hum at all. When I connect the wires from the PCB to the input sockets on the back of the chassis, I then get a little bit of hum. Using twisted pair signal/earth, taking the earth to the star point at the PCB end, reduces this to very acceptable limits in my case.
You also need to distinguish inherent hum from hum induced on the input wires. If I remove the input wires from the PCB, there is no noticeable hum at all. When I connect the wires from the PCB to the input sockets on the back of the chassis, I then get a little bit of hum. Using twisted pair signal/earth, taking the earth to the star point at the PCB end, reduces this to very acceptable limits in my case.
Currently no with accuracy due to scope failure but I measured yesterday the voltage between the chassis and the signal ground to be about 14mV ac. I would like to mention here that the chassis is connected to power ground via a 10 ohm resisto and a cap in parallel. Is that normal? Is maybe the cause of the hum. Also as it is obvious the chassis is connected to the outlet ground of the house.
Member
Joined 2009
Paid Member
If you don't need to run at high power, reducing the idle current will lower both ripple and temperature. Output devices will spend less time being driven into beta droop - where their current gain falls steeply as current flow increases. Of course this only works if you can accept less output power.
Shielded input cables are a great idea - although if you don't have suitable wire to hand it can be expensive to find. A twisted pair might work better than a single wire.
You could bring out a potentiometer to the front panel and have the idle current adjustable, or switcheable. I believe somebody (Mooly ?) once suggested adding an extra gang to the volume potentiometer and have the idle current track general listening levels.
Shielded input cables are a great idea - although if you don't have suitable wire to hand it can be expensive to find. A twisted pair might work better than a single wire.
You could bring out a potentiometer to the front panel and have the idle current adjustable, or switcheable. I believe somebody (Mooly ?) once suggested adding an extra gang to the volume potentiometer and have the idle current track general listening levels.
