building a first audio amplifier; isolation, heat transfer issues

[[email protected]]

So I found this page:
http://www3.national.com/mpf/LM/LM3875.html

.... and I'm confused by what they mean by "isolated TO220". It seems
to be encased in plastic, which seems a very strange thing to do,
since heat transfer is critical.

If I get the ceramic one, is there a need to isolate the back of the
chip from ground?

 

[Bob]

So I found this page:
http://www3.national.com/mpf/LM/LM3875.html

... and I'm confused by what they mean by "isolated TO220". It seems
to be encased in plastic, which seems a very strange thing to do,
since heat transfer is critical.

If I get the ceramic one, is there a need to isolate the back of the
chip from ground?

So if you look at the data sheet it explains the difference between the
isolated and non-isolated version. So if you get the isolated version then
the potential of the heatsink doesn't matter. So you need to learn how to
calculate what your die temperature will be for your particular application
and then you'll know what to do.

Just learn what thetaJC and thetaJA for ICs means, and learn what thetaSA
for heatsinks is.

Bob

 

[[email protected]]

So if you look at the data sheet it explains the difference between the
isolated and non-isolated version. So if you get the isolated version then
the potential of the heatsink doesn't matter. So you need to learn how to
calculate what your die temperature will be for your particular application
and then you'll know what to do.

Just learn what thetaJC and thetaJA for ICs means, and learn what thetaSA
for heatsinks is.

Bob

Thanks for the hints. Reading closely, yes I see the tab is V-. I
thought it was ground.

Interesting, the data sheet for the LM3886 shows a ground, but the
LM3875 (discontinued?) does not have a ground pin.

Thanks again,

Michael

 

[Eeyore]

So I found this page:
http://www3.national.com/mpf/LM/LM3875.html

... and I'm confused by what they mean by "isolated TO220". It seems
to be encased in plastic, which seems a very strange thing to do,
since heat transfer is critical.

Isolation is often critical too.

Isolated TO-220s are very common. They're very useful too. The isolating
hardware is often a real nuisance. The data sheet states 1C/W thermal resistance
from junction to case which is absolutely fine for an application like this.

If I get the ceramic one, is there a need to isolate the back of the
chip from ground?

What are you talking about ? There isn't a 'ceramic one'.

Graham

 

[Eeyore]

Thanks for the hints. Reading closely, yes I see the tab is V-. I
thought it was ground.

Interesting, the data sheet for the LM3886 shows a ground, but the
LM3875 (discontinued?) does not have a ground pin.

Op-amps don't have ground pins either.

Graham

 

[[email protected]]

Isolation is often critical too.

Isolated TO-220s are very common. They're very useful too. The isolating
hardware is often a real nuisance. The data sheet states 1C/W thermal resistance
from junction to case which is absolutely fine for an application like this.


What are you talking about ? There isn't a 'ceramic one'.

Graham

ceramic one = non-isolated one... right?

http://www.national.com/mpf/LM/LM3875.html
LM3875T, vs. LM3875TF.

Or, since National is implying the 3875 will be discontinued, LM3886T
vs. LM3886TF.
http://www.national.com/mpf/LM/LM3886.html

Thanks guys

Michael

 

[Robert Adsett]

Interesting that they appear to rate the thermal resistance identically
for isolated and non-isolated versions. I'm not sure I believe the
package material's thermal resistance matches that of the bare metal.

ceramic one = non-isolated one... right?

No. Generally it means bare metal (ceramic is usually an insulator
after all). In this case it does mean bare metal, note that the
datasheet refers to the tab on the non-insulated version being live.

Robert

 

[Eeyore]

ceramic one = non-isolated one... right?

No ! Where does NS mention ceramic ?

Non-isolated simply means the plastic doesn't cover the tab.

Graham

 

[Eeyore]

Interesting that they appear to rate the thermal resistance identically
for isolated and non-isolated versions. I'm not sure I believe the
package material's thermal resistance matches that of the bare metal.

Probably too little difference to be worth printing. The tab is quite large, so any
difference is likely to be in the region of 0.1C/W or less in my estimation.

Graham

 

[Robert Adsett]

Probably too little difference to be worth printing. The tab is quite large, so any
difference is likely to be in the region of 0.1C/W or less in my estimation.

An iso pak from IR appears to add about 2 C/W (see
http://www.irf.com/product-info/datasheets/data/irliz44n.pdf and
http://www.irf.com/product-info/datasheets/data/irlz44n.pdf) The
national package appear to be about twice the area so I would have
expected a contribution of around 1 C/W from the isolation. While they
might be better tham that, I'm having trouble with them being 10x
better.

I've seen TO-220 down to 1 C/W so 0.5 C/W for the National case if the
die is large would not be shocking but I'd have expected a significant
difference between the isolated and non-isolated versions.

Robert