My numbers refer to the lateral surfaces. Tom quoted the epoxy thickness between the copper plate and the epoxy edge as 1 mm.
Regards
I fully support pushing the envelope on things. But please do the math first. The LM3886 will work for years in any application if you stay within the limits listed in the data sheet.
If you deliberately exceed the limits only to turn around and claim that the IC is bad, then I support Bills opinion.
If you don't know how to do the math, I suggest hitting the books. I'm sure you can request a copy of Sedra/Smith and Kreyszig through your local library.
~Tom
Ah, just went back to your attachment.
You used 3 to 1. It appears to be closer to 5:1.
You mention calculating the epoxy thermal conductivity by using the difference between isolated and non isolated. That method does not work. Remember, they add gain and offset to both versions recommended values, so you do not have adequate information to the the calcs correctly.
jn
Well, to paraphrase Mr. Bennet: "I do not have the pleasure of understanding you"!
Could you be more specific, please? What is 3:1 and what is 5:1?
The calculation method based on using the difference of the thermal resistances with and without the epoxy layer is scientifically sound. I take the respective thermal resistance values as the only existing ones, and use them in good faith. If you claim that they are rubbish, then how should we design our equipment?
Regards
The thickness of the epoxy under the copper is 1mm. The thickness above the copper is 5mm. That is a 5 to 1 ratio.
You used .5 and 1.5, or a ratio of 3.
The numbers cited by the manufacturers are not rubbish. They are values the manufacturer states you can expect when you use the device as recommended. They are not the actual values, but have been changed to provide a margin of safety. For example, if the device was measured by the manufacturer as being a value 1, they will tell you that it is 1.5 give or take, so that even if you mount is badly, it will not blow up.
You cannot use those values in computations to derive the epoxy thermal conductivity. To do so will produce errors.
jn
What is in those books that are needed to design a power amp using the LM3886? They seem overkill. The math I've seen on the datasheet only requires high school algebra.
Of course, the fact that the math isn't that complicated makes it that much more of a shame when it isn't used.
sorry tom but i dont understand your point.learned more from burned silicon intentionaly than datasheets, thank you on books suggestions.
Sorry, but not quite right.
My figures are 0.5mm underneath the copper plate, and 2.5mm above. Gives the ratio of 1:5.
I really wouldn't mind seeing an alternative calculation, possibly with insider knowledge included. Might learn something new. Seriously.
Regards
Correct. However, to push the limits, it is often helpful to fully understand the system. To gain that understanding, I suggest reading Sedra/Smith. In particular the section on power dissipation in a class AB output stage will be useful. You can also check out the Thermal Design Section of my Taming the LM3886 page if you feel like it. The Power Supply Design Section may be useful as well.
~Tom
Um, you said exactly what I did..
your pdf seems to have different numbers, giving 1:3.
I wrote a tutorial for Linear Audio. It shows how the thermal system works, shows the 45 degree model, and provides a clear and concise look at how the IR hexfet line was modelled, was spec'd, and what offset/gain is required to arrive at their specified numbers. The numbers they give are roughly twice the theoretical calculation with .25 degrees c per watt added across the entire product line. edit: I'm not sure about figure 12 however, as it may have been lost in the editing process.
As to the chip being discussed, a single point calculation (one die size) would require the total transistor area and the percentage of the transistor topology that was actually involved in the calculation.
jn
Last edited:
lm-s and rf dont go together well,especialy without math knowledge. better to use transistors made for switching application
One advantage of the LM3886 is that it is quite rugged. When you try to get the most from it in a small total design, there are lots of things one needs to do. You just don't go one way or the other in most cases, but optimising usage of varying methods in different ranges.
Um, you said exactly what I did..
your pdf seems to have different numbers, giving 1:3.
jn
The ratio of 1:3 is valid for the lower/upper surface of the tab, and it is correct.
Regards
Amazing.
Let's try to put this geometry quibble aside once for all.
The epoxy thickness below the copper plate is 0.5 mm (hopefully no objections to that).
Your 5:1 ratio gives 2.5 mm for epoxy thickness above the 1.5 mm thick copper plate.
Summing the above three thicknesses gives overall tab thickness of 4.5 mm.
My 3:1 ratio gives overall tab thickness of 3.5 mm
Tab thickness measured with a digital slide caliper is 3.35 mm.
So no cherry garden in sight from where I stand.
BTW, thanks for the tip re. your article in Linear Audio, I wasn't aware of it. LA Vol. 9 arrived last week, but was left unopened on the post stack.
Looks very interesting, and I'll have look at it.
Regards
I am only looking at the cross section tom provided, with the ruler next to it.
I am not talking about the copper thickness, just the thickness of epoxy above it in relation to the epoxy below it. It looks like 1:5.
As to the tutorial, I suspect there may be a few errors left, time was not on my side. I have not yet opened my copy either.
jn
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.