Bridged Parallel Amp Questions

[h_andree]

Hello all,

I'm trying to put together the specs for a BPA-200 amp.
I have read the application notes and used the
overture spreadsheet.

I have the following plans:

Transformer 625 VA +/- 25 V (Voltage rails ~ +/- 37 V)
1 per stereo chassis.
The filter caps will be 2x68000uF per side.
I would use 4 chips per side.
I want to bolt the chips to an aluminum channel
(5" high 0.2" thick and 15" long). Each channel will have
2 heatsinks of 5.25" long 7" wide and 3" high.
The total has a dissipation of 0.43 oC/W per channel
(Exluding the aluminum channel). The case will be aluminum
with some ventilation holes.

I would like the amp to be able to use at 4 ohm loads.
The application notes state that this should be fine but the
overture spreadsheet notes a 0.8 oC/W heat sink is required
per chip. That is more then I would have.

My questions: Is the tranformer adequate or do I need a
higher rating (The application notes talk about 365 VA per
channel)? I'm assuming that the total continuous load will be
lower then the maximum load.

Should I use 6 chips per channel and reduce the load per chip
(Higher cost) and improve heat dissipation?

Or should I reduce the voltage and go for less power?

Should I go for the lm3886 chip (higher power) or the lm3875 chip?

Would a 20 Ga power hookup cable be enough per chip?

I would like some input on these design decisions and I
would like to hear peoples personal experiences with these
kind of designs.


Thanks

Harry Andree

[tlmadsen]

Hi H_andree

Have a look at

200 watts Gäjnklon (Gainclone) - BPA200

Your 625 VA transformer would be fine, but the 25 V AC will only give you about 34 V DC, not that it makes a big difference.

The 2 x 68000 uF is more than enough.

You need some kind of "soft/slow" turn-on circuit with that size tranformer and level of capacitans

With the 34 V DC voltage level I think that your cooling and number (4 pices) of LM3886 are fine. Just make sure that you take care to get the best possible cooling of the chips.

Have fun


Thomas L. Madsen

[h_andree]

About the voltage.

I agree that with the theoretical value of 1.41 x the secondaries
you should get 35 Volts. However when I measure the
mains it's not 110 V but 122V. That's why I got 27.1 V
out of a 18V secondary.

This 10% higher voltage will give me 37V in my current house.
I do not know if it's important but that's why I want to
stay away from +/-30V like in the application notes.
I'd rather be safe then sorry.

Harry

[peranders]

Originally posted by h_andree
This 10% higher voltage will give me 37V in my current house.
I do not know if it's important but that's why I want to
stay away from +/-30V like in the application notes.
I'd rather be safe then sorry.

As long as you have less than 42 V DC it's fine.

[macboy]

Should I use 6 chips per channel and reduce the load per chip
(Higher cost) and improve heat dissipation?

Or should I reduce the voltage and go for less power?

When going for high power, make sure to use the non-isolated package (T), not the isolated one (TF). The thermal characteristics of the isolated package are very poor.

Using 6 chips wouldn't hurt (except the budget) and would help to ensure that the amp will never limit the current.

Should I go for the lm3886 chip (higher power) or the lm3875 chip?

The 3886 is much better for high power. It has a higher maximum power dissipation of 135 W vs. 100 W, and it has a current limit of 11 A vs. only 4 A for the 3875. For a 4 ohm load, the choice is a no-brainer: 3886.

Would a 20 Ga power hookup cable be enough per chip?

If it is very short, then probably. But you will start to experience voltage losses in longer runs. Try for 16 ga or higher.

[h_andree]

Thanks for the suggestions so far.

So I will use the lm3886 T.

I've read somewhere that
people have difficulty insolating the chips. I wonder
if there are any pitfalls in this process?
Should I use nylon screws? Somebody recommended
TO-3 mica insulators. I'm using a single heat sink
so I cannot keep the back of the chips hot.
I will use the heat transfer compound.

I do not want to go into the discussion of fast
switching diodes vs bridges since that has been
done in different treads. I'm planning to
uses separate bridges for each channel.
What should be the current rating for bridge diodes?
The amp can draw 300 W max at 60 V. This is 5 amps.
But the diodes are not constantly on. Also 2 diodes should
be on at the same time. Thus a 5 amp rating would probably
enough. I would double that to at least 10 amps. Is this a
fair assesment?

So the power hookup wire should be thick. Have people
experimented with different geometry here. I was planning
to have a braided design. But maybe a parallel wiring is
better. The capacitance would eliminate noise?
I read in the Jeff Rowland litterature something about this.
They have the conductors very close together though.
Anybody any thoughts on this subject?

Thanks again

Harry

[peranders]

If you are going to use the amp in light duty (at home) I suspect that the TF will perfectly and you won't have to worry about isolation. If you are thinking of heavy duty then you must choose the "T" model.

Before you do too much I'll suggest that you read listed documents carefully and also try to understand them:
http://www.national.com/an/AN/AN-1192.pdf
http://www.national.com/appinfo/audi...gn_Guide13.xls
http://www.national.com/appinfo/audi...sign_Guide.pdf

and check once more Mr. Madsen's pcb.

You can't connect the IC's in parallel pin by pin and I'll doubt that you can get away with this without a pcb.

[h_andree]

I will use the amplifier for home use. I expect the
amplifier hardly to run out of power. I'm using 12W
of power now and run out of power only one
few occasions.

On the other hand I would like the amp to be reliable and
prefer not to experience shut down.

Harry

[peranders]

Then it's very clear, use the TF model! You will have a very powerful amp despite the fact that you can't take out continuous power at full throttle. Note also that the IC's will run rather cool so you must play for a while to get them hot (I believe, haven't tested yet).

The rectifier bridge should be 200 V / 10 A (if you have one bridge per channel) on heatsink or at least bolted in the chassis. This will also fit, I'll gather.

[h_andree]

Maybe I missed it but what do people typically do
for input buffer power supply. For the regular
gainclone the lower voltage is ok so you can
use a voltage regulator. The bridged parallel
version has rails from +/- 35-40 V. A buffer opamp
typically will draw < 100 ma

I see a couple of solutions for the power supply
of the opamp buffer:

1) Using a resistor and a Zener diode
to down regulate the voltage to 24V.
Then add regular voltage regulators.

2) Design a power supply from scratch
with Nationals tool. Seems complex.

3) I'm wondering if this is possible: use only
half of the rail with a separate diode bridge.
(Which will give me 35-40V single
ended and I divide it into +/- 17-20 V) and use a
voltage regulator. Is this possible?

4) Is it wise to just add an extra 15 W transformer
and diodes? I'm not sure I have enough space in my case
for this option.

I wonder which one would be easy and which
one would give the best results.

Any thoughts or help?

Thanks

Harry