juniorhifikit wrote:Just wanted to report back on this project of mine...
Wayne's 1570 stereo bus amp is performing wonderfully. The oscillation problem I was having was due to a ground loop from some not-so-neat flying leads to correct for a pcb error in the prototype.
As for the DAC problem, it's come back to haunt me. The tracking CV bias supply has stopped the DAC's output amps from frying (when having to sink too much current before they're properly powered up), but I'm still having DAC chips die on me. I've been able to monitor the I2C serial connection between the microcontroler and the DAC, so I can see that the DAC is no longer responding (dead). Today I'm going to monitor the communication with a known-good DAC. I have three hunches:
1) supply voltage spike - but after 2 regulators in series (main +-18V regulator, then down to +5V with another regulator in series), with filter caps on each, I doubt this is it, but it's still a possibility. I'll probably need a better scope to know.
2) -12V bias frying the DAC's output amps - but we already addressed this with the bias supply tracking the DAC's +5V supply. The DAC's are no longer showing the same symptoms of fried output amps (0 volts or full rail output - nothing in between) so I don't think this is the problem either.
3) The microcontroler is powered by one 5V regulator (fed from the +7.5V digital supply) and tied to digital ground; while the DAC is powered from a separate 5V regulator (fed from the +18V analog supply) and tied to analog ground. The two chips are tied together by the I2C bus, which I can't remember if it's biased with the +5V digital supply or the +5V analog supply. Perhaps there's some difference in potential that's causing current to flow where it shouldn't... On the schematic, the I2C bus is biased from the +5V analog, but in the hand built prototype, I can't remember. It's currently being beta tested at a studio at the moment, so I can't check for a few days.
4) Pin 4 of the DAC, which is the LDAC pin, was left floating (as per the instructions in the data sheet), but in a design note they stated it should be grounded. When this pin goes from high to low, it's possible to change the DAC's bus ID, and also power-down individual outputs. That's not something I want to do, so I've since grounded the LDAC pin, replaced all the dead DAC's, and put it back into service. Crossing my fingers and waiting is not a very solid problem solving technique!
Any thoughts would be greatly appreciated, as usual!
Sorry for the huge image - smaller was unreadable. Might have to open it in another window
That huge image didn't show the whole thing anyhow... maybe you need to make it bigger.
I still think the tracking bias rail is over kill.
We've probably gone around this tree before but you should be able to diode clamp the resistor string going to the control voltage node so a) it can't go very negative, and b) with resistor in series that current can't be enough to hurt anything.
I seems from a casual inspection that perhaps breaking the R from the DAC into two series R's and a diode clamp there to ground should work and still allow - control voltage more than -1 diode drop... probably several ways to skin this cat...
If you need to you could put a diode in series with the r from the DAC so clamp would hold DEC side to no lower than ground, but that shouldn't be necessary.
The divider from the dac could also be redesigned, say with a first divide from the output of the dac to ground. then from that node another resistor feeding the divider resistor to -V this should put less stress on the DAC.
The two diodes approach could insure that DAC never sees any negative voltage.
?? Is there any transient or otherwise difference between analog ground and logic ground voltage (VSS)?
Those parts look pretty robust... should be harder to kill than that.
JR
PS: I am not immune from blowing stuff up... I recently fried a 44pin PIC and two LED latch drivers, on a proto when I lost a posiitve bias string on A/D inputs and apparently hit the PIC input with too much negative juju for too long... FWIW I changed the design to cap couple all A/D inputs so even a serious DC fail in the future will only be transient.
Odd to see the failure vector reach right through the PIC A/D inputs and out to a couple LED latches through the serial com lines. but fault current has to go somewhere, and positive supply only chips don't like negative voltages at all....
Cancel the "cancel culture", do not support mob hatred.