Massive GC? Suggestions please

[Nicks]

Hi,

I have 2pcs 2x500VA (Total 2000VA) transformers with 24VAC secondaries. I have looked at the BPA 200 design but i think my transformers can handle a bit more. I read somewhere that Jeff Rowland used something like 12 LM3886 per channel.
Now that sounds really cool.
I think that would be overkill for me but what i want is some schematic thats bigger than the BPA 200.


My knowledge in this subject is limited. Thats why i humbly ask for this forums help.

BR

Nicks

 

[Nicks]

ok, just to make things clear.

Im looking for something a little bit bigger than the BPA 200.

Would it be better to use 3 LM3886 in parallell instead of just 2 like in the BPA 200?

What would i gain?

I need some inspiration folks!

There must be something cool to build with 2kVa of transformer power!

 

[jaudio]

More current and the capability of driving low impedance loads.

 

[paulmac]

Howdy

I'm planning on making the attached BPA200 design with six 3886s.

I wanted the extra capacity to handle loads that weren't 8ohms all the time, or another way to look at it was to just give the amp a bit of extra overhead from the published design.

My layout was derived from the usual AN1192, and the front page of the 3886 design note, and from spending far too much time reading this forum, than actually doing it...

I chose the dimensions because 150mm x 150mm PCB board is cheap and easily available.

Obviously (I hope), this is only one "side" of the BPA200 design.

I laid out the board with the taller caps and components away from the chips, so that I could mount the chips to the heatsink using a spreader bar over the chips, to help squish them closer to the sink. I reckon it looks nice and symmetrical, but the purists amongst us won't like it because I put the power supply caps too far from the chips... Perhaps mounting the 0.1uF power rail caps on the trace side could help...

I have put in components for a zobel network on each chip also.

The "zen" (or is that chi?) was ruined a bit by adding two 10k resistors and a header pin for connection to a Rod Elliot SIM board from the middle chip.

My plan was to drive the boards from a balanced output from a mixing desk directly, so I hope I won't need a buffer. That could easily be built onto an input board if needed.

My power supply will be a 300VA toroid, which I had to unwind from 30 to 24v, so your 500VA transformers will be perfect for a heavy duty 200+ watt power amp.

So, forum members, what do you think? I wish I could say I've built it, but I'm having trouble with the toner transfer method and the paper I have at home. I keep getting 70-80% transfer. I've printed it out on overhead transparency film, so that's my next move to try.

Also, instead of the 0.1r 5w resistor, what about using the LR network as detailed on pg 20 of the 3886 notes? This would provide extra amp stability. Would it achieve the same thing for load sharing?

Paul

 

[motherone]

I've been planning on doing a BPA design as well. Mine will be based off of BrianGT & Peter's LM4780 boards. I'll be using 2x4780's (essentially 4 3886's) in parallel, then bridge two of those together. I'll be doing thes as two monoblocks, with either a transformer to split the input into inverted/non-inverted, or using Digi01's DRV134 board (haven't decided yet...)

I think there's a big advantage on using the 4780, since you don't have to try to mounth 8-12 chips per amp (only 4-6)

Yes, it's crazy, but what the hell. It should be fun.....

 

[autoexec]

how do you get around the problem of the lm3886 producing less power at anything other than 4 ohms? (i think thats right...)

 

[Nicks]

Correct me if i am wrong here...

But i think you get around that issue when you connect them in parallell.

Thanks for all the inspiration!

I think the way for me is to go with 3 chips in parallell on each side and then bridge the two sides.

That should be a potent recipe for an amp!

At least i hope so.

BR

Nicks

 

[jaudio]

Bridging also doubles the output voltage. More power into 8ohms.

 

[motherone]

Check the National Semi doc.. i think it's AN-1192 (it's in other posts around here) on designed bridged-parallel amps with the 3886.

I humbly suggest going with the 4780 for ease of use if you have 4780 based boards laying around or if you're designing from scratch. You can build the simple BPA out of two chips with that, or you can build monster ones using half the chips that would normally be required (since it's two 3886's on the same package).

Regardless, it should turn out to be a fun monster amp. Just make sure you use adequate wiring to carry all the current those chips will be capable of pumping out.

 

[jaudio]

Does the lm4780 have twice the heat conducting area as 2 lm3886?

 

[Nicks]

I think (read: not sure) that the heat issue will be easier to solve with 3886.
My current GC (BrianGT 3875) runs really cool with a 7mm aluminum plate as a heatsink.
But im going to use a separate heatsink for each 3886 in this new amp. im going for the cheap but useful CPU heatsinks with a copper core and aluminum fins. Those are really cheap over here.

Another reason for using the 3886 would be that i dont know where to buy the 4780 here in sweden....

I probably will go with the servo version of the BPA design too.
I will use the DRV134 to split the signal aswell. it keeps the design simpler imo.

BR

Nicks

 

[motherone]

Does the lm4780 have twice the heat conducting area as 2 lm3886?

No, I don't believe it does. National rates the 4780 at 60 watts per channel versus the 68 watts from a 3886.

 

[slackman]

Does the lm4780 have twice the heat conducting area as 2 lm3886?

No, definetely not!
I mounted a heatspreader made of copper (6mm thickness) between chip and heatsink to achieve better heatflow (copper is much more better than alu). With this trick i increase the active heatcoupling area of the chip. The package of the lm4780 is too smal to handle all the heat in full power; the contact area (and the area nearby) can't take all the heat.

So without fan the bridged version is 'forbidden' for this chip, parallel should work. In this case better take two lm3886 (what technical is the same) and give them some space between ...

A tipp for heatflow cream: this schould only fill the mikroscopic small holes in the surface of the material - so DON'T place a thick film of it between components! Heatflow will be much better if you take care of this...

mic

 

[motherone]

So without fan the bridged version is 'forbidden' for this chip, parallel should work. In this case better take two lm3886 (what technical is the same) and give them some space between ...

I think Bridging would work okay if your speakers are 8ohm or higher.. If they dip below 8ohms, it would probably tax the poor chip on heavy transients. I agree there.

I would say that in a bridged/parallel arrangement (where each 4780 has both it's amps paralleled, then a pair of 4780's are bridged), the heat dissipation shouldn't be as much of an issue.

A tipp for heatflow cream: this schould only fill the mikroscopic small holes in the surface of the material - so DON'T place a thick film of it between components! Heatflow will be much better if you take care of this...

mic

A great way to do this (from my computer overclocking days) is to use the edge of a credit card:

1) Apply enough heatsink compound to cover the chip.
2) Apply the chip to the surface you wish to mate it with
3) Remove the chip carefully
4) Use the edge of a credit card to wipe all the excess compound off of BOTH surfaces (chip and heatsink)
5) Mount chip to heatsink surface

This way you ensure there's no air-gaps at the heatsink or chip side, since you filled those in with the material.

Mike

 

[Nicks]

So folks,

What do you think of my idea? Should i do something else or stick to my concept?

What about parallelling 4 on each side and then bridge the 2 sides?

As i said before, i started this thread to get some inspiration and good ideas for a massive monster GC!



Nicks

 

[motherone]

Go for it.

Jeff Rowland made/makes two amps based on bridged/paralleled 3886. One uses 3x per phase the bridged (6x 3886 per channel).. I believe that's in the "integrated" stereo amp. The other is a monoblock that goes with 6x 3886 per phase, then bridged (12x 3886 per channel)

I'm going to go with 4x paralleled and then bridged and see how it does. I think it should handle any kind of speaker you can throw at it.

 

[Nicks]

From what i have read on this forum and other GC sites i understand that each 3886 requires about 80-90 VA transformer power each.

So with the 2x500VA i have, six 3886 per channel would be a good match. so i think i will make 2 monoblocks with six 3886 each. And the other 2x500VA will be a stereo amp with similar setup as the monoblocks.

That should be enough power

Thanks for the inspiration folks!!!

I needed it.

BR

Nicks

 

[tlmadsen]

Hi

Have a look at my picture in:

http://www.diyaudio.com/forums/showthread.php?threadid=19157&perpage=10&pagenumber=2

post#15.

you might get inspired for your monster amp.

If possible, "spare" the LM3886 as far apart af possible or maybe use fan-cooling to keep the temp down.

Have fun

Thomas

 

[autoexec]

say if you bridged 2 banks of 4 in parrallel, with an 8 ohm speaker would that mean that each chip will see 16 ohm and therefor each chip can output about 35w, times by the 8 chips gives 280w in total, right?
and with a 4 ohm load you would get 8 ohm load on each chip, so 50w a chip gives a total of 400w
is this all correct?

 

[tlmadsen]

Hi Autoexec

Yes in theory you are (about) right. However you will have different kind of "losses" in your system, that makes is difficult to reach the theretical max.
With a 500VA72x30 V AC transform you should reach something like 200W/8ohm and 350W/4 ohm. If you have a bigger/stiffer tranformer you might get it 5-10 % higher.

Have fun

Thomas