Do I have only 1 chip "connected" at a time? Is this done by "disconnecting" the output resistors?
Thanks!
Thanks!
Why would you do that?
Parallel amp circuits can drive lower impedances, so functional test is to load it with minimum rated load for the full board, say 4 ohms, drive it to clipping and check whether it can meet rated specs or not.
Removing one of said amps will guarantee it will not.
Side note: usually the "second" amplifier is slaved to the first one so when you remove the first, the second stops working.
On the other side, removing the slave will still let the master one working, albeit at reduced current capacity.
There are other configurations too.
Side note: apparently the words "master" and "slave" are not to be used any more 🙄
If this trend continues we won't be able to speak Electronics 😡
Technology/Engineering are as neutral as can be (as Math/Physics/Chemistry etc. are, by definition) , they should not be corrupted by Politics.
Oh well.
Parallel amp circuits can drive lower impedances, so functional test is to load it with minimum rated load for the full board, say 4 ohms, drive it to clipping and check whether it can meet rated specs or not.
Removing one of said amps will guarantee it will not.
Side note: usually the "second" amplifier is slaved to the first one so when you remove the first, the second stops working.
On the other side, removing the slave will still let the master one working, albeit at reduced current capacity.
There are other configurations too.
Side note: apparently the words "master" and "slave" are not to be used any more 🙄
If this trend continues we won't be able to speak Electronics 😡
Technology/Engineering are as neutral as can be (as Math/Physics/Chemistry etc. are, by definition) , they should not be corrupted by Politics.
Oh well.
To avoid blowing everything up at once?
Some parallel LM3886 are slaved like you describe, but most that I've seen run the LM3886es in parallel. If that's the case in OP's circuit, you can certainly test the circuit with one LM3886 populated and add the others once you're confident in the circuit. You can also pull the ballast resistors and test each LM3886 individually.
Tom
Oh, in general I find paralleling voltage sources a bad idea.
Except transformer windings where you can guarantee exact same voltages.
Slaving amps is relatively safe because "one is thinking and the other is blindly following"
Now paralleling amps where each one has its own NFB and its own idea about what output voltage must be and doggedly tries to set it, in my view is an accident waiting to happen.
Even with 1% precision NFB components, I don't like it.
Adding output resistors in series is a crude method , again personal opinion.
No problem at all with bridged outputs because they are not in parallel but in series and outputs are separated by the full load impedance, not just a pitiful 0.1 ohm resistor.
Oh well.
This poor idea was pushed by chip manufacturers, who double sales, so their advice is not exactly neutral.
Back to the OP, I'd still like to hear his motivation.
Not sure whether he's repairing a once functioning parallel output amp or testing a fresh build.
Not the same.
Or maybe he bought a batch of possibly counterfeit chipamps and wants to test them.one by one 🤷🏻
Last edited:
Can you post the schematic? The question is whether they're truly in parallel, or if there's a "primary" and "secondary" chip (the politically correct terms).
No discrete output stages with multiple devices in parallel for you then! 🙂
Except you can't guarantee the exact same voltage. Even two strands in a bifilar winding will experience slightly different electromagnetic flux and develop slightly different voltages. For two separate transformer windings you could easily have one wound on top of the other and/or have a slight difference in the way the wire exits the bobbin so you end up with slightly different voltages. The difference is likely inconsequential in the vast majority of cases, but I guarantee you that you can measure a difference if your meter has enough resolution.
Yeah, but then you'll have a slight difference in phase between the master and the slave, so you still won't get the exact same voltage at the same time.
Precision is hard... 🙂
±1% was hot 30 years ago. I use ±0.1% tolerance resistors to set the gain of the individual LM3886. Even good ±0.1% tolerance resistors like the Susumu RG-series are about five cents in reel quantities. They aren't hand-carved from solid blocks of unobtanium anymore. 🙂
The ballast resistors in series with the output of each LM3886 balances out the output currents similar to how emitter resistors balance the current between output devices. You will get some standing current through those resistors. How much depends on your circuit design and also on the resistance of those ballast resistors. If the resistance is too low the amps will fight too much and performance will suffer. If it's too high you burn most of the output power in the ballast resistors. 0.10-0.22 Ω is a good compromise.
You can see the performance of a parallel LM3886 approach here: https://neurochrome.com/products/modulus-286. It's hard to argue that it doesn't provide good performance.
That sounds like a conspiracy theory to me. I won't blame a company for trying to sell more of their products. Using more LM3886es in parallel is one way to allow for low-impedance loads even at the full specified output voltage, which is something a lone LM3886 can't do.
Tom
If you are trying to see if the lm3886 chips are working individually, YES, removing the output/ballast resistors will be the right way.
Not sure how to parallel lm3886 in slave/master mode.
I know you can parallel Tda7293 chips slave/master mode, but not the lm3886s
The TDA allows you to connect directly to its output stage. Handy for paralleling in such a master/slave configuration.
Come to think about it, you're right about the LM3886. It would be hard to turn it into a master/slave configuration because it would have to be configured as a buffer. Great, except the minimum gain is 10 V/V (20 dB) for the LM3886. With some creative compensation it may be possible, but that approach seems sketchy. It's commonly done with opamps, though.
Tom
WHO said that?
Multiple transistors in parallel in a single amp are controlled by one brain : its input differential pair (in any of its variations) and ONE NFB net.
No conflict there.
2 parallel amps have TWO "brains, each with its own idea of what output voltage will be.
C'mon Tom, you are better than that 😄
Not surprised by a similar answer from other Forum.members but you are a Pro.
A microscopic difference?
No doubt.
But if properly wound, it will stay that way.
I commercially wind transformers and am careful to start and end windings at the same point in the circumference around the core.
Meaning I not only match turns count but also to fractions of a turn.
Which is not NASA technology but very easy to achieve.
There are other Pro winders in the Forum (Tony Tecson?) who can confirm that.
Arguing bifilar windings is a joke 😄
True at a Theoretical level, please calculate phase difference and current imbalance in a real world amplifier.
I seem to remember you are a Physicist, so maybe Parts per Million are important there; I am an Engineer, to boot a Production focused Engineer, si in our world precision is defined as "close enough"
Otherwise no saleable product would ever cross the Factory door 😫
Is the OP using 0.1% precision?
If so, more power to him 👍🏻
FWIW we have not heard of him any more..
Even worse, I still do not know what is meant by "testing", go figure.
Is it troubleshooting?
Is it performance checking?
Is it academical curiosity?
Do you know what is his goal?
I don't yet.
No.
Ballast resistors equalize currents all going "one way".
Worst case imbalance will mean one will pass more current than other, will dissipate more, but never ever will one inject current into another.
While 2 parallel amplifiers will happily inject current one into the other any day of the week.
Again, Tom, I didn't expect this kind of reasoning from you.
You bet and that's my point.
Won't argue Neurochrome, I am trying to help the OP, not to Evangelize anybody.
Just curious, do you sell some kind of parallel Chipamp board or kit? 🤔
Surprised by the fierce defense.
Not sure what are you talking about.
Never ever mentioned Neurochrome, don't know what they sell, even less offered any "explanation" or theory (conspiracy? ... Serious?) about them.
Just based on Physics "general principles" (as usual) and considering that in general parallel voltage sources , specially if each will FIGHT to keep a certain out voltage, are "poor Engineering".
That kludges are needed precisely show that
Oh well 😄
Tom offered the parallel86(lm4780) a few years back, now he has the modulus 286, 2 lm3886 chips in parallel and modulus686, I believe 3 lm3886 chips in parallel bridged with other 3 lm3886 in parallel.
His amps are proven to be very reliable, I am curious if Tom tested the protections of his amps for in the worst situations, short circuit when fully blasting, discrete transistors amps would suffer catastrophic damages.
His amps are proven to be very reliable, I am curious if Tom tested the protections of his amps for in the worst situations, short circuit when fully blasting, discrete transistors amps would suffer catastrophic damages.
My way of testing such combination was using 0.1% resistors in feedback network, after that measuring voltage difference on outputs of the chip (before current share resistors, if you are looking on the signal path), if there is no big difference during program (music playing), only few mV, I would say that amp will work fine.
Built 3 stereo parallel amps and I am satisfied with.
So important thing is using 0.1% resistors, when you use 1% resistors the results were not so good, lot of dissipation on the chips because of some difference in the amplified signal and probably one of the chip will sink current.
Built 3 stereo parallel amps and I am satisfied with.
So important thing is using 0.1% resistors, when you use 1% resistors the results were not so good, lot of dissipation on the chips because of some difference in the amplified signal and probably one of the chip will sink current.
@JMFahey - Wow. Now I'm the one who's a bit blown back by the force of your response here. I was simply pointing out the contradiction in your statements. I wasn't defending anything other than my point of view; that you can, with sufficient care, connect voltage sources in parallel.
The only difference I see is that emitter followers can't sink current so you won't have a standing current between two emitter followers. But as I also pointed out, you can minimize the standing current to the point where it doesn't affect performance (even at the ppm THD level). If it doesn't impact circuit performance, is it still a problem? If a tree falls in the forest...
The only issue I see is that each amp 'slice' will have to provide both the output current and the standing current, so if the standing current is too large the output current available for the load becomes too small. The solution to this is easy: First minimize the offsets (both DC and AC) between the amp 'slices'. Then set the ballast resistors such that the standing current is sufficiently small. The standing current in my designs is around 10 mA, typical. Should 10 mA out of an output current of 7 A be much of a concern?
Tom
You did. You said that you weren't a fan of connecting voltage sources in parallel. An emitter follower is a voltage source. An output stage with multiple emitter followers in parallel has multiple voltage sources in parallel. So if you're not a fan of connecting voltage sources in parallel, you're not a fan of emitter follower output stages with multiple devices in parallel.
But there is a conflict. Each output device will be different from the others, so it will take a slightly different share of the output current and have slightly different gain. Recall the local feedback loop in an emitter follower. Basically, each emitter follower has its own brain (to use your terminology).
The only difference I see is that emitter followers can't sink current so you won't have a standing current between two emitter followers. But as I also pointed out, you can minimize the standing current to the point where it doesn't affect performance (even at the ppm THD level). If it doesn't impact circuit performance, is it still a problem? If a tree falls in the forest...
And how is this a problem?
The only issue I see is that each amp 'slice' will have to provide both the output current and the standing current, so if the standing current is too large the output current available for the load becomes too small. The solution to this is easy: First minimize the offsets (both DC and AC) between the amp 'slices'. Then set the ballast resistors such that the standing current is sufficiently small. The standing current in my designs is around 10 mA, typical. Should 10 mA out of an output current of 7 A be much of a concern?
Tom