I'm helping my friend build a P3A amp, one channel works reliably and the other's really confusing us. MJE340 and 350 were used in place of MJE15034 and MJE15035, and the output transistors are MJL4281 and MJL4302. The amp seems to be working correctly except the output stage draws 200mA at idle (this is somewhat unstable) for reasons I don't understand - bias voltage is about 1 volt which should correspond to zero bias current. In this case, 20 ohm safety resistors are connected in series with each supply.
Here are the voltage readings for the defective channel:
And the working channel, which is biased to about 80mA:
Thanks in advance for any insights !
Here are the voltage readings for the defective channel:
An externally hosted image should be here but it was not working when we last tested it.
And the working channel, which is biased to about 80mA:
An externally hosted image should be here but it was not working when we last tested it.
Thanks in advance for any insights !
I assume you are sure your devices are not counterfeit? It's just that they seem very leaky, or worse, the circuit is oscillating and sucking juice for that reason. Did you use his board? The layout is critical for complementary pair outputs. You would most likely have problems when the amp is clipping, not at rest as I assume it is.
Check for leakage on the drivers. I prefer a 100 ohm resistor across the output base/emitter junctions, or better 47 ohm.
Check for leakage on the drivers. I prefer a 100 ohm resistor across the output base/emitter junctions, or better 47 ohm.
You could take the channel that works reliably and measure the voltage a key point. Then compare this to the one that doesn't work. Start the input section constant current source, then the differential pair and just work your way forward to the output. If you find voltages that don't agree you will as least have naropwed down the search.
Consider something a simple as a bad solder joint or wrong resitor value.
Consider something a simple as a bad solder joint or wrong resitor value.
Thanks to everyone for the suggestions. I have checked all the resistors on the board (which was bought from Rod Elliott), they are all correct - and I re-soldered many joints. All the transistors are genuine and inserted correctly. The first time my friend built up the boards, he did have Q4-6 backwards. We then flipped them around on both channels.
It appears that the input stage works correctly, the bias circuit works (and is operating like it should for zero bias) but something causes the output stage to draw current. Since the base to emitter volltages on Q5 and Q6 are so low (0.43V or so), wouldn't that mean they are not conducting at all and the 3-4mA current through R11/R12 is just going to the bases of Q7 and Q8? If that is true, why is the Q7/Q8 collector current only 200mA?
By leakage, do you mean Q5 and Q6 are defective? It's strange because the junction drops measured by a multimeter are still .611 and .619 volts.
What else might I check to see if the circuit is oscillating? It seems sort of unlikely given this is ESP's board and the zobel is in place.
It appears that the input stage works correctly, the bias circuit works (and is operating like it should for zero bias) but something causes the output stage to draw current. Since the base to emitter volltages on Q5 and Q6 are so low (0.43V or so), wouldn't that mean they are not conducting at all and the 3-4mA current through R11/R12 is just going to the bases of Q7 and Q8? If that is true, why is the Q7/Q8 collector current only 200mA?
By leakage, do you mean Q5 and Q6 are defective? It's strange because the junction drops measured by a multimeter are still .611 and .619 volts.
What else might I check to see if the circuit is oscillating? It seems sort of unlikely given this is ESP's board and the zobel is in place.
Well, a fast way to see if it's oscillating is to connect a scope. Hopefully the scope you use will have enough bandwidth to see the oscillation frequency if it's present.
Also, are you sure about C4 and C6? If these are omitted, or the wrong value, the amp could easily start oscillating.
EDIT: Also, what is the voltage drop across each of the output emitter resistors? R13 and R14
Also, are you sure about C4 and C6? If these are omitted, or the wrong value, the amp could easily start oscillating.
EDIT: Also, what is the voltage drop across each of the output emitter resistors? R13 and R14
I think your Q5 and Q6 may have been damaged and became leaky if the circuit was powered up with them in backwards. Then they could conduct even if their base-emitter voltages were less than the threshold. One thing you can do is to test the bad amp on +/-12 volts. Then temporarily replace the potentially bad transistors with 2n3904 and 2n3906 and see if you can get proper operation of the bias control. Also, you may be able to hook up a junk speaker in those test conditions and compare with and without the transistors changed. Keep the volume down low so as not to make the temporary transistors conduct more than there 200ma rating, though.
I don't recomend this until your sense of frustration becomes near terminal, and only if you have one working version of the PCB already: >>> Just strip everything from the bad boad, clean up the pcb and re-build it using the good board as a visual aid and cross check. It is, I admit, a brainless and brute force solution but it has worked for me when I lost interest in discovering "the solution" and just wanted the d**n thing to work.
sam9 said:
OK, maybe this will work, but... It is an intellectual challenge to find the bad part and learn something in the process. And you can easily lift traces when you pull everything, and end up worse off.
When I have a problem like this, I try to localize. Inject a signal in the front end and follow it to the output. Find the bad area, then bad component. It is logical and systematic and generally works.
In this case, the problem is entirely local to the output stage, apparently. The original poster already posted that he replaced the drivers after having them backwards, easy to do with the bass ackward pinout of the MJ340/350 pair. It seems like the upper transistor is leaky to me after that misadventure, that is my hunch. I'd change it and test.
But if you just want to shotgun it (technical term for changing everything
), change the drivers and outputs and I'll bet it's solved too. Can't really go along with changing everything, it's a step backwards.
I've tried both ways before. I like the intellectual method.
I did remove the power transistors; one had an Hfe around 150 and the other was around 250. This doesn't seem to matter.
Anyway, I just mounted everything up again and this amplifier is working as advertised. Since I hadn't changed any components (only resoldered pretty much everything) the most likely cause of the problem was a faulty solder joint, somewhere. Oy, well, it happens. This made me think "wow, it's a miracle YOUR amps work."
Also, [thanks to janaf for reminding me] I would like to note that the schematic in the initial post, while it is publicly available, is copyrighted by Rod Elliott; we should be careful about posting this sort of thing.
Thanks again to you guys for getting me on the right foot again.
I did remove the power transistors; one had an Hfe around 150 and the other was around 250. This doesn't seem to matter.
Anyway, I just mounted everything up again and this amplifier is working as advertised. Since I hadn't changed any components (only resoldered pretty much everything) the most likely cause of the problem was a faulty solder joint, somewhere. Oy, well, it happens. This made me think "wow, it's a miracle YOUR amps work."
Also, [thanks to janaf for reminding me] I would like to note that the schematic in the initial post, while it is publicly available, is copyrighted by Rod Elliott; we should be careful about posting this sort of thing.
Thanks again to you guys for getting me on the right foot again.
Kilentra,
I'm glad there's been a happy ending...... Soldering is quite a skill and it's easy to make a small but critical mistake.
While a scope is the only definitive test for oscillation, you can easily spot the effect by closely monitoring the Zobel 10R resistor.
Oscillation is always around 500KHz and up, and if present, always warms this resistor. Often, it will cook it, and a brown ring around the body gives the game away. I even find smell a useful sense when experimenting with new designs.
BTW, you should understand the purpose of smoke. It is actually fundamental to correction operation; semiconducting smoke, at high pressure, is encased during device manufacture into the black encapsulation, which is actually a pressure vessel. (If it gets hot, the seal is naturally lost, and you know the rest......) This is what makes semis so expensive; the high pressure process is tricky. If for any reason the seal breaks, the smoke escapes, and usually shorts the device.
I despair of people who fail to understand this simple fact; the smoke must be kept in at all costs. It is pivotal to correct semiconductor behaviour.....
Cheers,
Hugh
I'm glad there's been a happy ending...... Soldering is quite a skill and it's easy to make a small but critical mistake.
While a scope is the only definitive test for oscillation, you can easily spot the effect by closely monitoring the Zobel 10R resistor.
Oscillation is always around 500KHz and up, and if present, always warms this resistor. Often, it will cook it, and a brown ring around the body gives the game away. I even find smell a useful sense when experimenting with new designs.
BTW, you should understand the purpose of smoke. It is actually fundamental to correction operation; semiconducting smoke, at high pressure, is encased during device manufacture into the black encapsulation, which is actually a pressure vessel. (If it gets hot, the seal is naturally lost, and you know the rest......) This is what makes semis so expensive; the high pressure process is tricky. If for any reason the seal breaks, the smoke escapes, and usually shorts the device.
I despair of people who fail to understand this simple fact; the smoke must be kept in at all costs. It is pivotal to correct semiconductor behaviour.....
Cheers,
Hugh
Kilentra said:
He does post this on his website for the world to see and use, so I don't think you are breaking any rules. I suspect he would be glad to have the free advertising (Any mention is a good mention).
This amp is so straightforward that I doubt anyone could claim infringement, since amps like this have been published in data books since the early 70s. It is very like the RCA data book amps, without protection.
His mosfet power amp is a better design, using mirrors and current sources for the input pair, but even it is basically a 1980s design. That's fine with me, because the circuit is simple and well proven, but personally I don't like boot strapping (you always have a turn-on thump while the bootstrap cap charges).
I also don't like mounting devices under boards, because you lose heat dissipation ability from the board side. The outputs are almost insulated thermally, never good. I would prefer it if the board is spaced away, and the outputs are mounted outside the board area for better cooling. So I don't think I will be building any of his boards soon.
One last quibble: at these low voltages, I don't see the need for the high voltage parts he specifies. MJ15030/31 and even mj15001/2 devices would serve equally well. There are plenty of other lower voltage parts that would work well and save a few bucks. I don't understand why he thinks the latest goodies from ON are needed here (no doubt they are great parts, just not needed here).
AKSA said:
Hugh,
Indeed, loss of encapulating smoke is the major reason for the typical bathtub curve defining failures in semiconductor devices.
But major developments in material science may provide vastly improved devices. A recent development adds to the currently available forms of carbon, graphite, diamond, nanotube, buckyball. This new structure (I believe an Aussie) development is a 'feathered' structure showing tremendous enclosing volume per mass along with hyper strength. A unique feature is that this structure is actually magnetic for a few minutes after formation. This feature, along with the much ballyhooed MMMDs* promise substantial benefits, not only from a reliability measure, but also increased linearity, etc....
Given that these developments are not ready for the commercial market, I've endeavored to apply my own tweeks. I've devoted much research to this and have found that from my experiments, a system can be much improved by maintaining equal partial pressues of smoke both withing and without the semiconductors. Additionally, the composition of the smoke has specific impact on the various metrics of audio. As soundstaging and imaging are particular 'hot buttons' for me, I've settled on Pall Malls'
YMMV
* MMMD, Modified Magnetic Maxwell's Daemons
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.