Curious thermal effects in SMT part

Thread Starter

Veracohr

Joined Jan 3, 2011
772
I've been investigating drift in an analog synth sawtooth oscillator circuit I designed based on the LM13700 OTA. It showed significant drift over at least 10 minutes, and looked like it would continue drifting down in frequency a little more if I had recorded the output for longer. After eliminating the bias current source I used as the cause, I was left with only the LM13700. I thought perhaps it was drifting due to self-heating, because the output current is temperature dependent but the ambient temperature did not change.

However, I discovered that the same circuit on a breadboard using the DIP version of the OTA settled to a stable frequency in about one minute. The test board I made with SMT parts drifted down for at least 10 minutes (the original circuit had a more irregular, but still downward, drift). Out of curiosity I cut the power traces that go underneath the OTA and ran wires from the power inputs to the pins of the OTA and downstream parts. This cut the drift time down to about 5 minutes, even though there is only a few mA flowing through these traces. The two trace widths in this picture are 10mil and 20mil:

U3.png


I don't know how a small current going through even small traces generates enough heat to affect the part, but apparently it does. Before I cut the traces I also tried to measure the case temperature and found that it went up by roughly 7 degrees C over a 10 minute period. The original design using this circuit had 8 identical oscillators on the board, so now it's not surprising to me that it drifted so much.

The datasheet for the LM13700 shows that the DIP version has lower thermal resistance in all but one aspect:

thermal.png

Now I suppose my next step is to try redesigning the board using DIP parts and not routing the power traces underneath the parts. That will unfortunately mean a larger board and ultimately a larger end unit but it may be the only option. I was beginning to despair of whether I could make this idea work at all, but there's hope if the DIP parts will perform better.
 

MisterBill2

Joined Jan 23, 2018
18,061
If wire jumpers are an option that might solve the problem of the traces under the part. Not quite as pretty but possibly an adequate alternative.
 

Thread Starter

Veracohr

Joined Jan 3, 2011
772
Even though I removed the supply current from underneath the part it still didn't fix the drift as much as I wanted. Plus I realized after I posted that I checked the function not long after touching a soldering iron to the part, so it could be that the reduced settling time was only because it was already at a slightly elevated temperature at the beginning.
 

DickCappels

Joined Aug 21, 2008
10,140
One thought is that the chip is being heated unevevly. Putting a heat spreader such as a copper covered area under the chip might mitigate this effect.
 
Top