I ran one test at 8 Ω, ±28 V. On the "small" heat sink (130x130x1.6 mm aluminum sheet), the LM3886TF provided the full 38ish W output power. The heat sink reached 90.5 ºC at the peak power dissipation.
If the TF can survive on the small aluminum sheet, all the other variants can too. Also, if it can survive on the small heat sink, it will survive on the others as well. Hence, I didn't test any further at 8 Ω.
I did not, however, raise the supply voltage. I'll see if I can get some time today to test at ±35 V. I cannot do the ±38, as I'm using ±35 V electrolytics...
Stay tuned...
Yep. The peak power dissipation with a 4 Ω load at ±28 V rails happens at 40 W output power, just as the math predicted.
I'm not aware of any official spec for the TF package. There's a mention of 2 ºK/W in the BPA200 app note, but nothing in the data sheet. If it's not in the data sheet, it's not a controlled parameter...
~Tom
More data:
Test conditions: ±35 V supply, 8 Ω load, ambient temperature: 21 ºC.
On the "medium" heat sink, the LM3886TF provides a maximum of 60 W RMS into 8 Ω. The heat sink reaches 92 ºC at the peak power dissipation.
With the 32-tone "music" test signal driving the amp to just below clipping, the heat sink reaches 80 ºC.
~Tom
Test conditions: ±35 V supply, 8 Ω load, ambient temperature: 21 ºC.
On the "medium" heat sink, the LM3886TF provides a maximum of 60 W RMS into 8 Ω. The heat sink reaches 92 ºC at the peak power dissipation.
With the 32-tone "music" test signal driving the amp to just below clipping, the heat sink reaches 80 ºC.
~Tom
Thanks.
I stopped at Radio Shack; no TO-3 mounting kits. I'm going to browse some more, but I'm leaning towards the TF package. I would also have to scrounge up mounting screws and insulators to use with the T package.
I do realize that there is no guaranteed spec for the TF package. I like to be a control freak with my designs but I think I'll be OK, based on the data you provided. I'm also going to sit down and see if I can derive a theoretical thermal resistance from your data. If my suspicions are correct, the amplifier will experience a temp rise of around 100 degrees C driven steady full power - but much less with musical program. Still hot though.
I stopped at Radio Shack; no TO-3 mounting kits. I'm going to browse some more, but I'm leaning towards the TF package. I would also have to scrounge up mounting screws and insulators to use with the T package.
I do realize that there is no guaranteed spec for the TF package. I like to be a control freak with my designs but I think I'll be OK, based on the data you provided. I'm also going to sit down and see if I can derive a theoretical thermal resistance from your data. If my suspicions are correct, the amplifier will experience a temp rise of around 100 degrees C driven steady full power - but much less with musical program. Still hot though.
And the last bit of data:
Test conditions: ±28 V, 8 Ω load, ambient temperature: 21 ºC, "medium" (Pentium Pro) heat sink, LM3886TF.
Heat sink temp 67.6 ºC at the peak power dissipation. Max RMS power out: 39 W (limited by the supply rails).
At 1 W RMS output, the heat sink reaches 46.6 ºC.
At 0 W RMS output (idle), the heat sink temperature settles to 31.1 ºC.
~Tom
Test conditions: ±28 V, 8 Ω load, ambient temperature: 21 ºC, "medium" (Pentium Pro) heat sink, LM3886TF.
Heat sink temp 67.6 ºC at the peak power dissipation. Max RMS power out: 39 W (limited by the supply rails).
At 1 W RMS output, the heat sink reaches 46.6 ºC.
At 0 W RMS output (idle), the heat sink temperature settles to 31.1 ºC.
~Tom
The shoulder washers commonly used with TO-220 packages will fit the LM3886T. Use a #4-40 machine screw (M3x0.5).
That would be very cool. Please post your methodology and findings here. You should be able to use the data from Post #1, specifically, the cases where the output power was limited by the 165 ºC thermal limiter rather than the SPiKe protection.
~Tom
Tom, I don't see why you think that the AP comb signal reflects music's peak-to-average. The AP is a continuous signal. Sure, the difference between peak and average may be separated by the same difference in dB's as some nominal musical signal, but the peaks are very close together, rendering it essentially a continuous signal when compared to thermal time constants. I.E., the notion of 'peak to average' is different in the case of music vs the AP signal.
I have a bit of trouble decoding your question there, Brian as you actually give the answer in the middle...
Music is a continuous signal. Plug your oscilloscope into your amp to convince yourself. Sure, some music does have some quieter passages and some louder ones, but the time between those passages are much smaller than the thermal time constant of a heat sink, hence, it all averages out.
~Tom
Music is a continuous signal. Plug your oscilloscope into your amp to convince yourself. Sure, some music does have some quieter passages and some louder ones, but the time between those passages are much smaller than the thermal time constant of a heat sink, hence, it all averages out.
~Tom
OK, I'll try again. One can calculate a peak-to-average ratio of anything -- even a sine wave. In terms of the thermal performance of the system we're talking about, it would be pretty meaningless for a sine wave. Whereas music tends to have peaks that are more 'infrequent' compared to the average than does a test tone. My point is that the AP test tone will look (thermally) more like the sine wave than real music. I would expect that with the AP test tone, the junction and die temperature will come to some stasis and not change much. With music I would expect a lot more variation in die/junction temperature. I've done a lot of thermal impedance measurements at work, but all steady-state. It would be instructive to pop the top on an LM3886 and watch the junctions on a high-speed IR camera.
Bottom line: while most music playback will not overly stress your LM3886 thermal system, at loud listening levels peaks may trigger protection differently than the AP test signal.
Bottom line: while most music playback will not overly stress your LM3886 thermal system, at loud listening levels peaks may trigger protection differently than the AP test signal.
Surely that varies between the types of music. Some recent pop CDs have almost no dynamic range. A classical CD I bought yesterday has around 30dB difference between quiet and load passages. You have to put a stake in the ground. Measuring against a known reference that anyone else with AP gear can replicate would seem to be a good stake, otherwise you are into 'music power' type data, or something that is optimised only for your speakers and the music you play at your preferred SPL.
The other problem of course is that peak dissipation occurs at around half power...
The other problem of course is that peak dissipation occurs at around half power...
Ah, see what you mean. FWIW on my build I have gone way OTT on the heatsink just in case, but I am old skool, so If I can run for 30 mins sine at peak dissipation without the sink getting above 50 degrees I will be happy.
Doesn't stop me wondering just how far you could push it if you went mad with the thermal design
Doesn't stop me wondering just how far you could push it if you went mad with the thermal design
I disagree. If you look at the AP test signal on the scope, it looks like music. It sounds like a dissonant chord on a pipe organ. The 32-tone test signal has a crest factor of 13 dB, which is rather similar to music.
The AP 32-tone signal is also repeatable. I.e. anyone capable of playing back a WAV file can reproduce the results. It also conveniently avoids the discussion about what constitutes "music".
A sine wave, on the other hand, has a crest factor of 3 dB, which is at least 3 dB worse than the most heavily compressed music. If you design your heat sink for a sine wave, you'll end up over-designing by 3 dB minimum.
~Tom
Doesn't stop me wondering just how far you could push it if you went mad with the thermal design
Well... The LM3886 is limited to ±42 V as I recall, so even if you submerge the IC in chilled mineral oil, the highest output power you'll be able to get is about 100 W into 8 Ω.
There are less messy ways to build 100 W amplifiers...
~Tom
Well... The LM3886 is limited to ±42 V as I recall, so even if you submerge the IC in chilled mineral oil, the highest output power you'll be able to get is about 100 W into 8 Ω.
There are less messy ways to build 100 W amplifiers...
~Tom
No sense of fun
That depends on the thermal mass of the heat sink. On the "small" heat sink, it took a couple of minutes. The "medium" (Pentium Pro) heat sink stabilized fully in 10-15 minutes. The "large" 0.4 ºC/W heat sink took about an hour to stabilize thermally.
~Tom
Since we are talking about device to sink thermal stackups, a relatively new "grease" is the Loctite Thermstrate TC liquid crystal material. I have used it on some Pentium4 to heatsink installations and it seems to work well. This "grease" has some interesting properties and might be ideal for some amplifier build ups. Has some one else experience with this material?
I've looked at that before. Sadly for amplifier builders, I don't think it provides any electrical insulation, though. The thermal pads commonly used with semiconductors serve two purposes: Provide a good thermal interface and provide electrical isolation.
It may be useful with the isolated packages, though.
~Tom
Ive used this stuff during my over clocking days. Its still the gold standard even today, it comes out of Austria, but Egghead carries it, and its none conductive. It might provide better thermal transfer.
http://www.noctua.at/main.php?show=productview&products_id=13&lng=en
Noctua NT-H1 Thermal Compound - Newegg.com
http://www.noctua.at/main.php?show=productview&products_id=13&lng=en
Noctua NT-H1 Thermal Compound - Newegg.com
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.