There is also an combined bridged/parallel solution:
http://www.m1n1.de/Download/AN-1192.pdf
circuit: http://www.m1n1.de/Download/MonsterAmp_V1_4_sch.pdf
@tomchr
Thanks for your reply, looking forward to your high-power version 🙂
The app note is available from TI as well: National AN-1192.
While the chipamps certainly are convenient, I do firmly believe there are better ways to implement a high-power amp than by putting umpteen chipamps in parallel/bridge. The main attraction of the chip amps in that application is that they don't require any bias adjustment, hence, are very beginner friendly. The fact that they can operate on relatively low voltages is attractive as well.
That said, the design complexity in a bridge-parallel design is pretty high to start with. Then you wrap a composite loop around it and the complexity is taken up another couple of notches. Guaranteeing that the complete circuit works well even with the part-to-part variation of the LM3886es and the various quirks of the LM3886 would make me lose sleep.
I think better solutions can be found in a mostly discrete design with a couple of carefully chosen ICs in it. Drawbacks: More holes to drill in the heat sink, bias adjustment needed, higher supply rails.
Tom
While the chipamps certainly are convenient, I do firmly believe there are better ways to implement a high-power amp than by putting umpteen chipamps in parallel/bridge. The main attraction of the chip amps in that application is that they don't require any bias adjustment, hence, are very beginner friendly. The fact that they can operate on relatively low voltages is attractive as well.
That said, the design complexity in a bridge-parallel design is pretty high to start with. Then you wrap a composite loop around it and the complexity is taken up another couple of notches. Guaranteeing that the complete circuit works well even with the part-to-part variation of the LM3886es and the various quirks of the LM3886 would make me lose sleep.
I think better solutions can be found in a mostly discrete design with a couple of carefully chosen ICs in it. Drawbacks: More holes to drill in the heat sink, bias adjustment needed, higher supply rails.
Tom
Instead of making a composite amp out of the BPA, which may be difficult as you say, why not make each leg of the BPA a composite amp, i.e. make a BPA out of 4 composite amps. Higher parts count, of course, but the trade off might be worthwhile.
It shows an amplifier circuit that's theoretically capable of delivering over 200 watts into an eight ohm load, being driven by a +/- 15 volt @ one amp power supply for starters. And the regulator chips are being supplied the same input voltage that they are supposed to output, there will be no regulation there with that setup. Further, the current rating of those regulator chips is inadequate, they probably wouldn't last very long.
I could go on, but why bother, that design is a disaster.
Mike
Last edited:
Micheal, I think you are misinterpreting the schematic. The supply voltages are +/-42V. The regulators supply the opamps only.
Yep, you're right, I missed the PS connections on the left, I guess that's what I get for staying up way passed my bed time. 🙄
But! There's still things wrong with the amp circuit, the worst of which is the lack of output Zobel and Thiele networks. Yeah, I know that many believe that a low value resistor like the .1 ohm balance resistors at the output will serve the purpose, but they don't suppress oscillation in the emitter follower and CFP halves of the LM3886 output. And a Thiele network will isolate the output from excessive capacitance just as a resistor will, but it will also suppress RF ingress at the output that the resistors won't.
I guess my biggest problem with the circuit as a whole is that while it will work (in theory-not necessarily in practice), there are much better ways to make more power than stacking chip amps one on top of the other.
And running the chips right at their maximum supply voltage is delusional at best IMHO. Thermal issues are assured.
Mike
Last edited:
Leaving aside the error in your analysis, a Zobel can be added at the output if necessary and is hardly a make/break feature. I assume you know that the design as presented in the app note is functional, and is not a theoretical circuit. I certainly would build it if I could get a PCB
This discussion is all about using the 3886 chip to get more power. There are, of course, other ways to get 200W/8 Ohms but that really is not the point, is it? After all, one can say that there are other ways to get 40 W/8 Ohms too! The advantage of the chip amp based design is that it is pretty much bullet proof and all the 'house-keeping' has already been taken care of by Nat Semi.
This discussion is all about using the 3886 chip to get more power. There are, of course, other ways to get 200W/8 Ohms but that really is not the point, is it? After all, one can say that there are other ways to get 40 W/8 Ohms too! The advantage of the chip amp based design is that it is pretty much bullet proof and all the 'house-keeping' has already been taken care of by Nat Semi.
Last edited:
I was critiquing the circuit as presented. It is incomplete at best.
I can tell you from experience that the circuit as shown is inherently instable, it will almost certainly oscillate at MHz frequencies regardless of whatever load it's driving. And proper layout is a must for a circuit like this. Performance will suffer greatly, or it won't work at all worst case if the circuit is poorly laid out. Many (maybe most?) DIY members here lack the experience and resources to do a proper layout for a circuit like this that will work as expected.
I don't think that it's a circuit that anyone lacking a decent O'scope and/or distortion analyzer should even attempt.
Just one guy's opinion - from experience.
Mike
Also missing Cc, Rf2, and Cf. All needed for stability near clipping.
Correct. 0.1 Ω is not enough.
I completely agree. The BPA200 is a good starting point if all you want is high output power. If you also want good performance, there are better ways to achieve that.
I wholeheartedly agree.
It can, but it does no good there as adding the Zobel at the output maximizes the inductance in series with the Zobel. You want the Zobel connected right at the amp output pin and with as short a route to the local decoupling cap ground as you can.
You may want to double-check that claim against the title of the thread. 😉
Most of the housekeeping anyway. You can still screw up a single LM3886 implementation with a poor layout. This is especially true for a high-power design.
Which do you think would be the best seller:
- Design A: Composite BPA200. Mono channel. 200 W @ 8 Ω. BOM cost: $225. Board cost, say, $140 (four layers, pretty big board).
- Design B: Discrete + well-chosen opamp design. Mono channel. 200 W @ 8 Ω. BOM cost: $75. Board cost, say, $100 (four layers).
Design A would likely have measurably worse performance. Is avoiding having to turn a screwdriver to adjust a bias current worth $190 to you?
Tom
Last edited:
I would have guessed Design A until you said:
why is that?
However, is it possible to combine Modulus-86-like performance with higher output (200Wish)?
Last edited:
Because you can design an amp that is intended for delivering 200 W rather than cobbling a 200 W amp together from parts intended to provide 50 W. That's also what my simulations of a 200 W design show.
200 W into 2 Ω? Sure. 200 W into 8 Ω? Nope. Not with good performance anyway.
Tom
And the moral at least for me is? 🙄 To get more sensitive speakers or go active 😉
Back to the Modulus 86!
Best,
Anand.
Back to the Modulus 86!
Best,
Anand.
Open for pre-orders of the final board run of the Parallel-86
I have opened up for pre-orders on the absolutely final board run of the Parallel-86. You can pre-order your boards here: Parallel-86 (Last Chance to Buy)
Please pre-order by July 24th if you are interested in getting in on this last time buy. Four of the ten boards needed for the final board run to start have been pre-ordered. For the latest updates on this last board run for the Parallel-86, please see the Parallel-86 thread in the Group Buy forum.
You can read more about the Parallel-86 here: Neurochrome Parallel-86.
Please do note that the LM4780 is getting harder to source, so please make sure you have a source before placing your board order. Currently, Newark/Farnell and RS sit on the distributor stock available world wide.
Tom
I have opened up for pre-orders on the absolutely final board run of the Parallel-86. You can pre-order your boards here: Parallel-86 (Last Chance to Buy)
Please pre-order by July 24th if you are interested in getting in on this last time buy. Four of the ten boards needed for the final board run to start have been pre-ordered. For the latest updates on this last board run for the Parallel-86, please see the Parallel-86 thread in the Group Buy forum.
You can read more about the Parallel-86 here: Neurochrome Parallel-86.
Please do note that the LM4780 is getting harder to source, so please make sure you have a source before placing your board order. Currently, Newark/Farnell and RS sit on the distributor stock available world wide.
Tom
The Parallel-86 group buy is definitely happening. The minimum quantity has been met so now I just need to know how many boards to order. I'll tally it up on Saturday and get the boards ordered. If you're interested in these boards, please pre-order by 23.59 your local time tomorrow (Friday) night by following this link: Parallel-86 (Last Chance to Buy)
Tom
Tom