No, I’m not looking to use a DC servo with a gainclone. Just considering using them AS a DC servo amp where a very low output impedance must be maintained. Not looking for massive output current but higher than +/-15 volt rails IS required As well as fractional ohm Zout (ruling out NE5532). +/-30 to 45 would be nice, more correction range the better.
Is the LM1875 or 3875 my best bet, or should I keep researching? It will be used in a loop with a separate differential integrator, and probably set to a DC gain of around 20.
Is the LM1875 or 3875 my best bet, or should I keep researching? It will be used in a loop with a separate differential integrator, and probably set to a DC gain of around 20.
Sounds intriguing! How low must be the impedance at 0Hz? What impedance is permissible as servo gain droops with frequency? If you graphed requirements vs. frequency, what would the they look like?
Interesting...
Interesting...
Haven’t figured out all of the requirements yet, and of course the overall Bode plot is not worked out yet. Its a bias servo for a VERY LARGE tube amplifier being developed over say the next 5 years. I want an IC solution to drive the normally grounded end of a fixed-bias pot string. Not the usual 470k either. Several orders of magnitude lower impedance to REALLY get the g1 circuit impedance down. That grounded end needs to look like a couple of ohms, max, across the entire audio band - with NO dead zone. Needs to source and sink, both DC and the AC riding on it. Dead zone won’t hurt the DC solution, but will inject distortion unless that IC can directly drive say a 100-1000 uF capacitor bypassing that end of the string. I’m guessing it WONT, so it’s Zout must be well controlled through the crossover region on its own. If I could live with a dead zone I’d just use a pair of discretes, but biasing into AB requires all that thermal compensation nonsense increases the package count. One IC ought to do the trick if it’s the right one. I may use an asymmetrical supply (say +15/-50) to get more negative swing as vg1 is always negative. I know the +/-13 from an op amp will run out of swing and putting a follower on it gives even less. I just need to figure out what to go buy on the next mouser order, now that chip amps are back on the menu.
Most of the DC gain will be provided by small SMD quad op amp - two followers for the cathode current capture (and sending to the overcurrent protect) difference circuit, and integrator. I’ll figure the time constant as I go along. Needs to be pretty low, as the amplifier‘s poles will be a few Hz. The purpose is Not to set a quiescent current (as some drift will be tolerable) but to ensure BALANCE over time and under heavy load since it’s output transformer is toroidal. Multiple pairs in parallel will be used, each with independent balancing circuits. This requires some development time.
Most of the DC gain will be provided by small SMD quad op amp - two followers for the cathode current capture (and sending to the overcurrent protect) difference circuit, and integrator. I’ll figure the time constant as I go along. Needs to be pretty low, as the amplifier‘s poles will be a few Hz. The purpose is Not to set a quiescent current (as some drift will be tolerable) but to ensure BALANCE over time and under heavy load since it’s output transformer is toroidal. Multiple pairs in parallel will be used, each with independent balancing circuits. This requires some development time.
It sounds like the entire project will consume a lot of time, money, and resources. In that case, the additional expense of building a small class-A power amplifier out of high voltage discretes, to perform the DC servo function, becomes a negligible sliver of the total budget. Run it in class A and you don't need "all that thermal compensation nonsense," just set gobs of bias current and attach adequate heatsinks. Figure 20.10 of Douglas Self's Small Signal Audio Design 2nd edition "Class-A headphone amplifier with EF output stage" might be a convenient starting point. 13 discrete devices and an Audio Precision measured 0.0008% THD at 10 kHz -- when driving 30 ohm headphones. You of course will be driving heavier loads and thus need more coals on the fire.
You could look at something like an OPA549 which can use +/-30 and deliver 8A. An OPA462 will go even higher to +/- 90Vbut with lower current of 45mA. Ok, according to the datasheets.
But I fail to understand some of your requirements. It is DC servo but it has AC at its input and at the same time has to drive a 1000µF capacitive load. Zout of an opamp is also determined by its loop gain.
But I fail to understand some of your requirements. It is DC servo but it has AC at its input and at the same time has to drive a 1000µF capacitive load. Zout of an opamp is also determined by its loop gain.
I’d rather NOT have to drive a big capacitive load. That adds another low frequency pole to the overall loop. It is a solution for a driver amp with a dead band (let the cap filter out the crossover distortion) that I’d rather not have to resort to. If I can get a low ENOUGH AC Zout the bias networks will remain clean. Putting a 10 watt class A amplifier in each position isn’t really a palatable option either, since the tubes will be generating a lot of heat (not good for a sand amp dissipating a lot of power) and I want to keep the drivers physically close to the tubes and bias network. Running a bunch of lines all over hell and gone has in the past given me stability troubles for the main amplifier. Keeping it compact is a must, hence the idea of a single package 5 or 11 pin TO-220. The smaller parts can be SMD, also physically close to the tube sockets.
