LM3886 T vs TF (simple question, not an argument)

[BrianDonegan]

Okay, I am not trying to argue a point here.

This is not a new topic, and I am not asking which package is better or worse. Just looking for some numbers.

So, it makes sense to me that the T (un-insulated package) would have better thermal transfer then the TF (insulated) package. I was wondering how great the difference was, but can not find any information on the National site about it.

All the datasheet has to say about it is this (other than a single thermal resistance value):

Note 12: The LM3886T package TA11B is a non-isolated package, setting the tab of the device and the heat sink at V− potential when the LM3886 is directly mounted to the heat sink using only thermal compound. If a mica washer is used in addition to thermal compound, èCS (case to sink) is increased, but the heat sink will be isolated from V−.

The lack of information got me thinking that maybe there is not a great difference in thermal resistance. And in fact, dispite what I thought would be obvious, the TF may be better than T with insulator (not a claim, just pondering).

So my question is, can someone point me to some data showing the actual Thermal Resistance numbers for the two packages? I just want to be able to make proper calculations, that is all.

[kevinkr]

I've worked with both, and there is a very large difference in thermal resistance. Oddly enough I did not find any information on the TF package in the current LM3886 data sheet. (2003) The TF part has a thermal resistance of something like 2C/W or worse. If you are planning to build an amplifier around this chip and require more than 15Wrms per channel and a reasonable sized heatsink I'd probably use the T part with insulators.

edit

Download AN-1192 from the National website, it has all the thermal information you need to make your decision. I confirmed that the T package has a thermal resistance of 1C/W and the TF 2C/W. As I indicated above I would not use the TF if I needed anything close to the rated power and a reasonable sized heatsink.

[BrianDonegan]

Thanks Kevin. I did see that already. The only numbers they provide are in Table 2, which is specific to their heatsink. While they specify their heatsink dimensions and chip spacing, they do not specify the mounting methods, such as insulators used (if any), thermal componds, etc.

I'm not really trying to decide on which chip to use, as I have used both many many times. I have never run into any thermal issues with the TF package, usually running at 35V rails or so into 4-ohms. I am actually running a balanced BPA arraingement with the TF packages and the heat dissipation is not a factor for me (28V rails).

What I am looking for are numbers to provide heatsink calculations for others. I may need to determine these for myself if I cannot find anything documented. How did you arrive at the 1C/W and 2C/W numbers? How was the T package insulated from the heatsink?

[Grandma´s_SUB]

offtopic; Can someone do me a favour?That would the the measurement of the resistance on LM3875 between + and - leg.I have some probs.

Ive made my GC again, and how the hell could be 50 Ohm between + and - leg on the chip?This is like inserting 50 Ohm resistor between + and - rail voltage.

thx

[kevinkr]

The 1C/W 2C/W numbers are directly from the AN-1162 application note. (And have nothing to do at all with how they are heatsunk, those numbers are the thermal resitance from die to tab in the case of the T and from die to the exterior plastic insulation covering the tab in the case of the TF) Absolutely everything needed to calculate the thermal resistance of the heat sink is included in that ap note. The part spec provides the germaine information for the T version, and substituting the 2C/W for the TF will allow you to derive the required heatsink thermal resistance to ambient.

The equations to calculate the dissipation for a given supply rail, load impedance, and target output power are also in the specification. The specification also mentions the T package thermal resistance in the absolute maximums section as being 1C/W die to case.

Reading both AN-1162 and the LM3886 specification will answer all of your questions. I designed three commercial amps using this chip about 4 yrs ago and everything I needed to know was in those docs.

One thing that is not clearly spelled out is that thermal resistance is not necessarily a constant, at dc and very low frequencies the effective thermal resistance may be higher, I generally add a fudge factor of 20% for this if the amplifier will see extensive use below 100Hz.

A TF package will be used with thermal grease only. There can be as much as 1C/W - 2C/W thermal resistance between the tab and the heatsink depending on thickness of grease, (less is better) clamping force, co-planarity (flatness), and microscopic voids in the surface of the heat sink. Clamping across the body rather than bolting at the tab generally results in an improvement in heat transfer, and this can be as much as 50% better than the tab alone.

Black anodyzed heatsinks are about 25% better at transferring heat to the environment than a clear or unanodyzed heatsink everything else remaining the same. The difference is due mostly to black body radiation.

The lowest thermal resistance configuration would be either the thinnest mica washer you can find - .010" thick is not too thin! and thermal grease sparingly applied to both sides. Too much is worse than not enough.. IMO! Less messy and quite practical are some of the silpads - go for the lowest resistance ones you can find. The thermal resistance measurements on these are per unit area and are very misleading - in practice it is not unusual to see thermal resistances 3X higher than spec'd with these particularly if bar clamping is not used. I had a lot of problems with this specific issue.

I always set a specific thermal resistance budget for a high ambient of at least 40C and a maximum die temperature under worst case conditions (somewhere between 1/4 - 1/3 pwr typically for consumer audio apps) and find the most reasonable split between HS thermal resistance, package and insulator resistances. Usually commercially that requires about the best you can do coupling the device to the heatsink. In DIY a slightly larger heatsink is usually a safer way to go.

Edit: typo

[kevinkr]

I also recommend the purchase of a couple of very inexpensive digital thermometers with k type thermocouples. It is very illustrative of the problems faced to measure the tab temperature, and the temperature of the heatsink as close as possible to the device. Knowing just these two things along with the ambient temperature and the power dissipated in the amplifier will allow you to determine all of the thermal resistances in the path to ambient.

[kevinkr]

Hi GrandmaZ,
You really ought to start a new thread for your question, it's so far off topic.

However I think you answered your question. Most probably the chip has an internal short, particularly likely if it is the same in both directions. I am not familiar with the 3875 so I can't say for sure, but I'd recommend you start a separate thread.

[AndrewT]

Hi,

There can be as much as 1C/W - 2C/W thermal resistance between the tab and the heatsink depending on thickness of grease, (less is better) clamping force, co-planarity (flatness), and microscopic voids in the surface of the heat sink. Clamping across the body rather than bolting at the tab generally results in an improvement in heat transfer, and this can be as much as 50% better than the tab alone.

I do not have data for the wide To220 used in the chipamp but "as much as 1C/w - 2C/W" is above the 0.3 to 1.0C/W i have seen quoted for a normal width To220. I would expect the double width used in the chipamp to be nearer 0.1 to 0.5 found in To3 and To247/264 which are nearer in area.

Black anodyzed heatsinks are about 25% better at transferring heat to the environment than a clear or unanodyzed heatsink everything else remaining the same. The difference is due mostly to black body radiation.

I have seen +8% extra thermal resistance for the non black sink.

The lowest thermal resistance configuration would be either the thinnest mica washer you can find - .010" thick is not too thin! and thermal grease sparingly applied to both sides. Too much is worse than not enough.. IMO

Mica I have measured are more usually 0.002" to 0.005". I have heard of but not seen 0.001" for mica insulators. 2 to 5thou mica should achieve 0.4 to 0.6C/W and would bring T upto about 1.4 to 1.6C/W cf 2C/W of the TF.

So things may not be quite as bad as first sight (calculation).

[kevinkr]

There is often a difference between quoted values under ideal conditions in the lab and what is most often achieved in practice. I have measured the said "bad" results I cited in more than one application. It does not take much to get results that are far worse than what the manufacturers quote unfortunately.

The device thermal resistances are directly from the manufacturer's documentation and relate to the things other than the external surface area of that tab. (The size, thickness and location of the die on the tab as well as the thermal conductivity of the adhesive used to secure it to the tab.)

The best I have ever achieved in commercial production with silpads was about 0.25C/W. (IIRC - its been a couple of years and I don't have the lab notes.)

I use mica insulators in a lot of stuff, and the latest purchase from digikey were claimed to be 0.003" - although I have seen much thicker. I should grab my calipers and see if they are really that thin.
0.5C/W should be achievable with some care with mica imo.

One of the big issues with this modified TO-220 is to get enough clamping force to assure a low thermal resistance. I ended up using bar clamps..

The point I was mainly trying to make is to make some measurements they're comparatively easy to do.

[BrianDonegan]

Reading both AN-1162 ... will answer all of your questions

You are correct. I have put down the crack. Thanks!!

These are for kits, so the other factor is making the kit easier/more reliable, which is, to me, the biggest benefot of the TF (fewer smoke generators).

I am again leaning toward offering a choice of T or TF, with the T costing slightly more (base price is higher - due to copper content I suspect, thermal pad, shoulder washer, etc). Or, just stick with the TF and "KISS."

Thanks for the info and the dope slap