Problem with parallel LM3886 (PA 100)

[geraldfryjr]

Wow,Bas,that circuit is exactly what I have been looking for a SERV0_BPA-300.
I would like to try it ,if it works as good as you say it does (which I don't doubt you) if you don't mind.
Only I want to add two more output stages and make it eight.
From there I will keep adding stages up to sixteen or as needed to feed the low impedence hunger of my esl system.


abraxalito, Are you familar with circuit maker 2000 ?
If your macromodel for the 3886 could be configured for it I would like to give it a try.
I know it can be done,but I have never learned how to add models to it.
I have tried a few times but, somehow I did not do it right and they didn't work and failed. jer

 

[DJ4MC]

Wow ton of great info here. I suspect my power distribution to the boards here. It really isn't an offset issue. It is a massive oscillation issue when the outputs are connected together. I do have some pretty good test equipment (100M o-scope 6.5 digit DMM with 4wire ohm meter) Also the suggestion that the boards may not work for this is probably the most appropriate. I will get this to work though...

Appreciate all the help
Jules

 

[Sebastiaan]

Guys - some insights into paralleling problems might be found on this thread:

lm3886-x-2-parallel

Dear Abraxalito,

I just finished the whole thread, and that was very revealing and interesting.

I never thought about the cap tolerances myself, this explains why often I see topics about parallel LM3886 problems.

So again I stick to my point, omit the cap's, the DC offset pot's and go with servo's and the circuit is safe. 1% gain resistors aren't a problem with servo's as well. Since I always use servo's that explains why I never had any problem with any parallel/serial chip design, up to 12 chips!

With kind regards,
Bas

 

[Sebastiaan]

Wow,Bas,that circuit is exactly what I have been looking for a SERV0_BPA-300.
I would like to try it ,if it works as good as you say it does (which I don't doubt you) if you don't mind.
Only I want to add two more output stages and make it eight.
From there I will keep adding stages up to sixteen or as needed to feed the low impedence hunger of my esl system.

Dear Jer,

You can add up stages, but keep an eye on the input impedance the driver chip will see.

With kind regards,
Bas

 

[Sebastiaan]

No No I have nothing against ceramics used in the right places In fact I use them all the time.. But you have found the most dreadful place to put one. In a 80dB gain loop!
Oh boy you stepped on a land mine. I been blown up before by this same one too! Ive designed PLL's with a loop filter that had FM modulation sidebands by acoustic vibration. In fact also had clock recovery circuits causing bit errors big time by tapping fingers on the bench. Been there done that. Classic stuff here my Dutch friend and your fingers are in the dike.

OK ceramics are a piezoelectric generator (microphone if you will) X7R is barium titanate its one of the worst ones. Put a scope on the output of the servo and tap the PCB, you know use really slow sweep rates. you'll see 10-100s of mV.

Dear Infinia,

Thanks for your information. I will try this out, and honestly I neglected this a bit. In that case you are right, and cost will really add up for quality SMD film caps.

With kind regards,
Bas

 

[infinia]

Here some simulations.

Of course simulations are not the real world, but it gives an indication, at least about the self noise of the used components. Also, the simulated chips are not the LM3886's but the OPA549's since there isn't a spice model for the LM3886's.

OK simulate the servo w 1meg & 1uF by itself and compare the result to the LM3886 input referred noise. see that's what I'm talking about!

 

[Sebastiaan]

OK simulate the servo w 1meg & 1uF by itself and compare the result to the LM3886 input referred noise. see that's what I'm talking about!

Dear Infinia,

I never denied that an opamp generate noise. Of course it does, like any active circuit. But the same time, a ground is never ideally ininfity low in impedance either or 100% noise free. So either one of the legs of the chip amp need to refer to a not 100% silent ground, or a not 100% silent opamp.

The output noise of the servo opamp is still below the chipamps input noise. The bottom line is that in the output of the chipamp, you don't see this noise back, measured on a real product not just a simulation.

There are always trade off's in a design, but for me adding big value resistors in a inverted chip design, a chip design with un-equal input imedance legs, caps to ground in the feedbackloop, dc offset pot's are bigger tradeoffs then a good designed DC servo. Last but not least, in my ears a stable design with a DC servo simply sounds better.

With kind regards,
Bas

 

[geraldfryjr]

Thanks,Bas,plan to see reports on this somtime soon.
I have only one question, you mentioned using the NE5534 in the servo loop ,can a NE5532 or an old school TL07x or TL08x BI-fet be used as well?
If not then I will have to get some newer opamps then aswell.
Just trying to keep the cost down ,ya know? jer

 

[infinia]

Wow ton of great info here. I suspect my power distribution to the boards here. Appreciate all the help
Jules

OK Jules that's great you have some decent test gear. Sometimes folks have cheapo stuff measuring 3mV with 200mV range. LOL

My only comments are after each amp has been AC gain tested separately is the following
1) to use all the "optional" parts on the chipamp PCB DC input blocks, RC out zobels, etc
2) use <2 twisted wires for + supply> and< 2 twisted for - PS> keep short and away from inputs
3) Use twisted wires for <AC input terminations> 2 places to generator 50 ohms output impedance or lower say audio preamp or your buffer. keep away from DC supply and speaker out
4) speaker out use twisted pair make sure returns are direct to DC bulk caps common point. so then 2 output pairs connected at load/

 

[Sebastiaan]

Thanks,Bas,plan to see reports on this somtime soon.
I have only one question, you mentioned using the NE5534 in the servo loop ,can a NE5532 or an old school TL07x or TL08x BI-fet be used as well?
If not then I will have to get some newer opamps then aswell.
Just trying to keep the cost down ,ya know? jer

Dear Jer,

No, the servo opamps going to cost some $$ unfortunately (I agree here with Infinia, it ain't cheap). The Ne5532/34 have bipolar inputs and you need a FET opamp for DC servo's. The same time the offset and drift must be really low as well for the DC servo. Afterall the DC offset of the chipamps will be as low as the servo opamp.

The AN1192 has a flaw in my opinion by tying the positive leg of the inverting servo direct to ground. This way the servo chip itself has a fairly amount of offset by impedance mismatch from the servo itself (even though it is a FET inputstage), and the chipamps will have an offset around 1mV. With balancing the inverting servo (as I did) u can reach a DC offset as low as 6uV per chip, and this keeps the system really stable and cool with no need for 0.1% gain matching resistors. However if u go all the way differential as in my last example, those 0.1% resistors strongly improve the CMMR of each chip.

I will provide a more accurate schematic in the near future and might provide complete boards, so hold on. I don't recommend to build a complete DC coupled amplifier (like my schematic) without any protection. My designs do have DC protection. Currently I am doing some final experiments with various chip designs to come to better solutions. If things could be done simpler I prefer that. For now I stick to the DC servo solution.

With kind regards,
Bas

 

[Sebastiaan]

Dear Infinia,

You got me there But the 1M value is a typo in my schematic, since I stick to 2M for slower time constant. 1M isn't a good value for an amplifier design. My schematic has more errors I see. C8 from the differential driver stage must connected to ground and not to the DC servo.

I also admit, if you lower the 1uF to 470nF like the national app note, noise increase as well and WILL be visible in the output.

Thanks again for pointing me on the ceramics. Even though I never encounter any problems with them in servo's and I see them all the time in commercial products with DC servo's you do really have a very valid point there. Instead I try (expensive ) SMD film cap's. Wonder if they exist in those values, never really searched for them. My goal is to stick as much as possible with SMD's

With kind regards,
Bas

 

[infinia]

I see them all the time in commercial stuff like set top box TV tuners. Go ahead smacked it and see the picture freeze up LOL.
WIMA 50V has 2,2uF

 

[geraldfryjr]

Thanks again,Bas,right now I have five LM4780's to work with and I have everything to get one going except for the servo opamps,so I will wait for your recomendation before I order any chips.
As I have enough things to get finished up for the system right now,Thanks again! jer

 

[Sebastiaan]

WIMA 50V has 2,2uF

Thanks. I take a look! must be very expensive... (u know what they say.. Dutch are greedy..)

With kind regards,
Bas

 

[Sebastiaan]

Thanks again,Bas,right now I have five LM4780's to work with and I have everything to get one going except for the servo opamps,so I will wait for your recomendation before I order any chips.
As I have enough things to get finished up for the system right now,Thanks again! jer

Your welcome Jer. Later this year I will visit Chicago. Not far from Michigan. maybe by that time I can provide some help

With kind regards,
Bas

 

[infinia]

Thanks. I take a look! must be very expensive... (u know what they say.. Dutch are greedy..)

With kind regards,
Bas

Dutch are economical, Americans are greedy
I thought WIMA was Deutsch? SMD up 6,8uF http://www.wima.com/EN/smdpet.htm

It;s DIY audio so we can be the best!

 

[geraldfryjr]

Cool,meanwhile I will experiment with the ones I have and by then I'll be ready to build the final larger version,thanks. jer

 

[DJ4MC]

Sebastian, I would be very interested should you finalize a board layout for paralleling these with DC servos on the inputs. I am going to redo my wiring on the DC supplies. I braided the ground and +/- wires and each board connects to the main Supply at the filter caps. I am not however twisting my speaker outputs with ground from the boards. I have the speaker returns going straight to the star ground. All line level audio is using small shielded coax that is grounded at one end only. I will go over things with a fine tooth comb and post my results.

Thank you.

Jules

 

[abraxalito]

abraxalito, Are you familar with circuit maker 2000 ?
If your macromodel for the 3886 could be configured for it I would like to give it a try.
I know it can be done,but I have never learned how to add models to it.
I have tried a few times but, somehow I did not do it right and they didn't work and failed. jer

Tried googling for that but found some posts that said its no longer available. Did it get bought up by Altium - the first link Google gave me was to that. If CM2000 runs a standard version of Spice then I'm sure it could be got to work. My model is based on this paper so if you know anything about the variant of Spice your CM2000 uses, you might compare that with the examples here: ADI AN-138

 

[abraxalito]

So again I stick to my point, omit the cap's, the DC offset pot's and go with servo's and the circuit is safe. 1% gain resistors aren't a problem with servo's as well. Since I always use servo's that explains why I never had any problem with any parallel/serial chip design, up to 12 chips!

So Bas, how do servos make paralleled chips work with only 1% resistors? You use fairly large sharing resistors and don't mind an amp or so flowing between chips? How much power output did you get with your 12 chip design? Very curious about that as I'm currently doing a mere 4 chip (8 amps) one. Oh, and there have been others saying that paralleled amps just don't sound as good as single ones - have you heard any degradation between a single and paralleled design with your servos?