Perpetuum mobile, the holy grail, at last.
The specified DF of 400 you gave in posting #1, mostly specified for 8R LS loads, means a 20mOhm output resistance.
For stereo amps, gain between channels within 0.1dB or 1% is o.k. and below hearing threshold.
However, when paralleling them, at 30Volt output there is 1% difference between them or 300mV.
This 300mV over 2 times 20mOhm means 7,5Amp current flowing from one amp into the other without a LS connected.
With an 8R speaker a simple calculation learns that one amp is supplying 9.4A, the LS is taking 3,8A and the second amp is sourcing 5.6A.
So instead of making life easier for the paralleled amps, one has to supply almost two times more current as without paralleling and the other has to sink an enormous amount of current, way above its SOA and will most likely explode.
So yes, that´s the reason why you will need serial resistances.
Putting resistances at the output of the amps, will make that both amps are now supplying current.
For instance, 1R resistances will let one amp supply 1.9A and the second one 1,6A, together supplying now 3.5A to the LS instead of 3.75A with one amp.
So, your current output has dropped by a factor 3.5/3.75 = 0.93 or 0.86 in terms of power.
Instead of getting more power you now have 14% less power and a DF that went from 400 for one amp to 16.
Is this what you had in mind ?
Hans
Last edited:
You simply don’t want to parallel them at all, if they are that badly mismatched. Take steps to get them better balanced first, where a reasonable 0.1 or 0.2 ohm ballast resistor works without excessive circulating current. If that means hand matching your feedback resistors and input pairs, so be it. The concept works, you just need to be certain your implementation is actually doing what you think it should. If it doesn’t, you’ll just make a big heater.
(1) Thanks, yes it would seem like the solutions for paralleling amps affect the damping factor... that's why I wondered if using a restricted LF speaker ( mini monitor ) would help.
(2) An aside... how does the output impedance change with frequency? Meaning, how does it affect damping factor over frequency? And naturally then, as drivers get lighter with frequency ( less mass ) do we care about damping factor at the midrange and above?
(3) Humor... sense of humor.
No, not really. I was just thinking that adding a second amp would help with high power transients as two -hopefully matched amps- would be able to drive more current.
Looks like the output resistors burn up a lot of power.
So, without opening the amps, I could just sweep an input signal at different levels and measure the output. Doing this for both channels would allow me measure how well matched they are.
But, even opening the amp, it doesn't seem like they can be that easily adjusted.
I'm beginning to wonder how useful this is all gonna be, at least I do have a stack... and I just found a fire extinguisher under the kitchen sink.
Difference between channels is not constant. It changes with aging, temperature, humidity, etc. Let's suppose a transistor in one channel dies. It will immediately make huge disbalance and probably destroy the other amp.
Uh
It won’t be easy to getting below 0.1 dB synchronization from 10Hz to 20kHz.
😂
Hans
One will need to compensate over the whole audio range and not just with the easy part of matching feedback resistors but also of all caps contributing to the final gain.
It won’t be easy to getting below 0.1 dB synchronization from 10Hz to 20kHz.
😂
Hans
True, but a transistor failing would be a corner case.... and true, in a parallel configuration we'd be running a potentially catastrophic topology.
One thing at a time... I just built up another RAID5, that was a bit more important in the queue. Will order parts within a couple of days. It looks like everything is in stock.
WIll order 1 ohm, 2 ohm and 8 ohm 100 watt wirewound resistors, and time to buy a new DMM as well. Actually two would be useful.
I think I can use an Android tablet and a USB DAC to output the test signal. Do you recommend any one Android application for that?
Now, I will have to clean up the mess on the other desk, where I got the soldering station. I haven't used that for some years now.
I was gonna link to that app note (that Fast Eddie was alluding to) last week when this thread had 20 posts and presuming we get past the "if I make more current available, it'll give more power out" thing, but it's just as well I didn't. You'd see the output resistor values and assume you could use those values in the outputs of YOUR amplifiers, then wonder why they go into overload protection (presuming they have such protection). The gain setting resistors in your amps are probably 5 percent tolerance, or 1 or 2 percent if you're lucky, but that's still not close enough.
Hmm, please give me a little bit more credit.
My first post was to calculate the output impedance based on the damping factor, so that I could calculate the value of those output resistors. I do believe I did note that I couldn't find an equation that gave me the value. I assumed, yeah assume is never that good a thing, that at the very least the output resistors would have to be an order of magnitude greater than the output impedance of the amp... yet low enough not to waste a lot of power.
I also made another assumption... that using the channels in the same amp chassis would give me two closely matched amplifiers so that one wouldn't drive into the other too much, so that a very low value of the output resistor would be sufficient.
I think we're all making too many assumptions... regardless.. somewhere after the noise and vitriol, data started to come out, it seems that at least a 1 ohm resistor will be needed and that is just not going to cut it: affects the damping factor and dissipates too much heat.
So, running two power amplifiers in parallel is a lot harder than just wiring it up with some resistor and wires... seems like the amplifiers have to be designed for that.
Both amplifiers rely on their negative feedback. If the output differs from the intended regulated value, the amplifier will correct that. With two amplifiers in parallel, like clocks, both are different (even at a minimal fraction), 'sense' the fault on the (voltage-) output level and start adjusting according... with the open loop gain. Whatever you try, they run dead.
Series resistors to equalise can be calculated, #202 gives an example, but to avoid failures you should apply corresponding values for these resistors as shown in the application note (think in this case about 22 ohms, not 0.2 ohms). And that's about opamps, which are limited near voltage sources intended for small signal and low current running in class A.
But power amplifiers change to class b very soon, and are susceptable for reactive loads then. As soon as the feedback of a trailing amplifier starts to correct for a to high output voltage, it must drop the voltage as fast as possible, even if neccessary closing the driving semi and opening the other, 'shorting' it to the other, negative rail. The leading amplifier corrects for this and dumps as much current into this load as possible from the positive rail as a result. A lot of heat will be generated during this proces. When the signal swings to the other polarity, the same happens in reverse with the remaining semi's. If the applied frequency is 400 Hz (2.5 msec), I expect this proces to last for 5 msec in total.
Series resistors to equalise can be calculated, #202 gives an example, but to avoid failures you should apply corresponding values for these resistors as shown in the application note (think in this case about 22 ohms, not 0.2 ohms). And that's about opamps, which are limited near voltage sources intended for small signal and low current running in class A.
But power amplifiers change to class b very soon, and are susceptable for reactive loads then. As soon as the feedback of a trailing amplifier starts to correct for a to high output voltage, it must drop the voltage as fast as possible, even if neccessary closing the driving semi and opening the other, 'shorting' it to the other, negative rail. The leading amplifier corrects for this and dumps as much current into this load as possible from the positive rail as a result. A lot of heat will be generated during this proces. When the signal swings to the other polarity, the same happens in reverse with the remaining semi's. If the applied frequency is 400 Hz (2.5 msec), I expect this proces to last for 5 msec in total.
(1) You know... I was pulling his leg about the feedback.... he came through a bit, ahem, "the thing that sets the amplification (or something like that)" condescending.
(2) I must admit I've been thinking in terms of class A operation... class B adds a interesting wrinkle. I suppose that if the amplifier channels are not identical, the cross over from A to B will be different (from an external reference point of view). I wonder what this will do to the distortion? Is it then easier to parallel class A than classA/B? I suppose we shouldn't even try class D with their integrating output circuits?
(3) Another question. What do you mean leading and trailing amplifiers? Do you mean both sides of a single channel (leading +. trailing -) or either channel of a parallel configuration? With the leading channel being the one dumping current into the trailing amp. So that I'd suspect the leading channel will transition into B before the trailing channel...
Last edited:
There are two issues which cannot be solved when two nfb poweramps are in tandem: the nfb itself and the class b operation.
The nfb of the amps will fight each other, the 'leading' wants upwards, the trailing downwards (&vv), shorting each other asap.
Other then shown in the opamp AN (opamps run in class A only), in class b the nfb can only correct in one direction.
It has nothing to do with the switching itself nor the a-b distortion. Only parallel low power (op-)amps according to the AN are doable.
Current can NOT flow into a switched off bipolar transistor: above Vbe-rev (approx 6V) it breaks.
Take two identical clocks and fix both clappers together, give it a firm swing to start ticking. It's a second order equation with two unavoidable poles, the receipt for oscillation. Amplifiers are not oscillators, they break after one swing. The clocks will stop immidiately.
Two identical nfb amplifiers in tandem is not a disaster, it is impossible with these two poles.
Bonus: a thermal equilibrium is also impossible (the third... and fourth pole in the 2nd O eq.).
The nfb of the amps will fight each other, the 'leading' wants upwards, the trailing downwards (&vv), shorting each other asap.
Other then shown in the opamp AN (opamps run in class A only), in class b the nfb can only correct in one direction.
It has nothing to do with the switching itself nor the a-b distortion. Only parallel low power (op-)amps according to the AN are doable.
Current can NOT flow into a switched off bipolar transistor: above Vbe-rev (approx 6V) it breaks.
Take two identical clocks and fix both clappers together, give it a firm swing to start ticking. It's a second order equation with two unavoidable poles, the receipt for oscillation. Amplifiers are not oscillators, they break after one swing. The clocks will stop immidiately.
Two identical nfb amplifiers in tandem is not a disaster, it is impossible with these two poles.
Bonus: a thermal equilibrium is also impossible (the third... and fourth pole in the 2nd O eq.).
A peer review is not something you make in 5 minutes just to get a complimentary one in exchange,even if both are "peers", that´s a caricature of what peer review means.
Actually, "nobody needs to write anything at all" (what you imply), you simply post some idea or experiment or deduction in some Scientific quality public or semi public site, maybe a Scientific or Academic Magazine, and wait.
If everything is fine, nobody needs to specifically state so, congratulate you there, whatever.
IF NOT, somebody will rise to correct or rebuke your presumed mistake.
Risking that, in principle works like self-restricting filter; and the more you deviate from accepted knowledge (which you may very well do), the more proof you need to stack behind it.
Again, not sure you understand how Peer Review works.
I see I might have unignored you too soon. This has nothing to do with running two amps in parallel, now does it? Would it be possible to stick to the subject and not try to ridicule or engage in ad hominem attacks?
But, to answer your accusation:
Please look at the Agile development environment, specifically how JIRAs are reviewed.
I've been doing peer reviews since 1979, btw.
Last edited:
So, in essence, you are saying that unless both A/B amplifier channels have identical characteristics there will be a point during the crossover (*) under class B where their feedback will fight each other. Ergo, there will be some pyrotechnics at worst and some smoke at best.
BTW, did you mean to use "tandem" in lieu of "parallel"? "In Tandem" means sequential, as in one behind the other.
(*) OK, here the wording gets tricky.. see if I can explain my use. Both channels of a parallel topology run on A/B. Because they are not quite enough identical, one channel is likely to transition from A to B at a different time than the other. At this point we will be see an issue where on channel is trying to dump current into the other. I figured the +/- switching while both are in class B should not be an issue as they both should be in phase by then.
Did I get this right?
Last edited:
You're a perfect example of the system you built yourself around to support your own weaknesses...
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8216989/
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6327091/
https://embassy.science/wiki/Theme:...g,considered a questionable research practice.
Last edited:
- Status
- Not open for further replies.