Quote:
A quick question, will the LM6172 work well in this board and be stable and all? |
Good question. I'd intended to try it for the heck of it, and composing this reply has helped me to think the problem through more carefully. Thanks.
We put in a place for a pair of bypass capacitors for difficult chips like the 6172. We even went to some trouble to make the path from the bypass caps to the opamp's power pins very short. (About 4mm!)
But in my brief experience with this chip, that isn't enough to make the LM6172 happy. I didn't try soldering the chip into place instead of socketing it, so perhaps that will help. In addition to the bypass cap, what I found necessary was adding a small (10pF) film cap in parallel with the feedback resistor. (Not the one that goes to ground, the other one.) This rolls off the high-frequency response of the chip, which helps keep it stable.
We had a place for such a cap in the META42 for a while, but we took it out at the last minute for a few reasons:
1) It was taking valuable board space, but it was only there for the LM6172, which is a cranky, cantankerous chip. Better alternatives exist, especially for buffered amplifiers like the META42. In an amp like the Corda, where the opamps are directly driving the load, the 100 mA current drive capability is a nice feature, but it's not helpful for us.
2) ppl said to, and heaven knows ppl's always right
I think there's still hope that the LM6172 will work anyway. If you configure the amp for multiloop, the inner loop sets the opamp for a very high gain, like 100. If you tune that gain right, you can intentionally limit the bandwidth of the opamp, which effectively rolls off the HF response of the chip. This may have the same beneficial effect as adding the feedback capacitor without the bad phase shift effects added by a capacitor.
(For those interested: To tune the chip's bandwidth, you find the chip's "gain/bandwidth product" in the datasheet and use that to find a gain that gives you enough bandwidth for audio plus a decent safety margin. 100 kHz, say. The LM6172's GBP is 100 MHz. GBP=A*B where A is amplification in dB and B is bandwidth in Hz. To limit the bandwidth of this chip to 100 kHz, you need a gain of 1000, which is probably too high to achieve. I'll have to play with this to decide if I can get enough bandwidth limiting to stabilize the chip.)
Thanks again for the prompt to try this. The results should be interesting.
EDIT: I just remembered that if a feedback capacitor turns out to be necessary, all is not lost. You can take advantage of the multiloopability of the board to add a capacitor where one of the multiloop resistors would normally go. You configure the amp for global feedback and then make use of the now-unused inner feedback loop to place the capacitor from the op-amp's output to the -IN. (R6) If you leave the output buffers out (which is reasonable with the LM6172), you've basically made a CMoy w/LM6172 and a studly power supply.