Light BIAS

[CBS240]

Hi

The HEC hexfet output circuit is one of the best output arrangements I’ve used, in terms of ruggedness, bandwidth, and distortion. This being said, it has a significant problem besides being a pain to stabilize sometimes. The bias can be difficult to maintain at a constant level because of the temperature coefficient of the outputs and of the error devices which both affect the bias. Being able to use high ft, small signal transistors for the error amplifier and not having to route that circuit around the outputs and all the way to the heatsink gives a significant improvement in bandwidth and thus distortion. However, this brings up the problem of having a different temperature, ambient or just above, for the error amp transistors than the temperature of the outputs. Obviously in this case the error devices will not track the temperature of the outputs, thus some sort of a Vbe multiplier action is needed. While it is true the typical Vbe bias spreader on the VAS works with this circuit, the amount of compensation and its effect is mitigated by the positive feedback node. This makes it difficult for the Vbe multiplier transistor to sense the temperature of the outputs, and be able to correct bias. The conductance change required of the multiplier transistor may not be exactly a ‘linear’ function wrt temperature and the outputs and EC transistor temp co. This is probably one reason the circuit is not as popular as it should be.

BUT, I haven’t given up quite yet.

My simple ………….

IMO, the answer is in the biasing of the error amplifiers, right where Bob has his pot for bias control. If you set up a feedback loop to sense the current in each output source resistor and use that info to control a photo-coupler to do the job of the bias pot, it just might be possible to control the bias without interfering with the audio or error signals. The idea is to make the conductance of the photo-diode relate to the common mode current. The trick, and what I’m still playing with now, is to get the correct amount of gain to properly track the bias current. Since this is a circuit concept that can’t be easily simulated, I had to make another mock up of the output stage on the same PCB as the control circuit. I screwed up and placed the photo-diode too far from the EC part of the circuit exposing a “sensitive” node to RFI by stretching a wire across the PCB. It makes it more of a pain to stabilize and I have to compensate more and cut the BW of the EC stage short, though it is not terribly important now as that is not the part of the circuit I’m focused on.

For my attempt, I’m using a LT300 opto-coupler from Vishay, a pretty common part. A quad op-amp, and a few transistors, just nickel and dime parts…..BC850, BC860, FJV1845, FJV992, a couple J-fets ect. Since the outputs cost less than a buck a piece, the real expense is in the 0.1 Ohm 1% power metal film resistors from Caddock that are almost $4.00 each.

IMO, the results thus far warrant spending some more effort on the circuit and idea. Since I’m swimming in uncharted waters, and I have not seen something similar in the forum, any thoughts would certainly be salutary. If there is any interest here, I think there could be good discussion and I would like to continue it. But if no-one has any interest, I will continue anyway, just won’t bother the forum with it. Surely some of you have contemplated ways to bias output transistors as AB followers that is based on common mode conductance instead of temperature…………

Here is a sketch of the circuit thus far. The +/- 40V and +/- 12V comes from the driving stage PS. P1 controls the symmetry of the current sense and P2 sets the amount of bias in the outputs. With the output bias at 200mA, current in the LT300 LED is 14.9mA, leaving each photo diode to conduct ~150uA. The circuit action does track the temperature increase and adjusts the bias, but I think it needs little more gain, or more change in diode conductance vs temperature. Also I could re-adjust the fixed bias so as less conductance of the diode is needed. It tracks fairly well except for at extreme temperatures.

[Bonsai]

Why don't you use a microcontroller to do this?

1. You can make it self calibrating (no potentiometers . . . wow)
2. Accurate
3. Opto provides in-built isolation - no noise problem
4. Easily handles non-linear power/temp/gate voltage variation
5. Easy to make it work with tight control over a very wide temperature range - essentially the full range of the output device temperature capability
6. Cheap
7. Fun to write some C code.

[CBS240]

Hi Bonsai

Yes, that would be quite a trip. Gosh, it's been a long time since I've written anything in C. I wonder if the intended operation would be complicated enough to warrant such a device. I built this simple circuit to see if and how the concept of using opto-isolation to control the bias with HEC might work. I was thinking of using a PWM w/filter to control the diode conductance. It might be simpler than say a PIC, though probably not as accurate. I would have to ponder on the best way to represent the current flow in the outputs as ‘data’ since all you have is the voltage across 0.1 Ohms with mV sensitivity. But with hexfets, the bias doesn't have to be spot on if it requires much more complexity to achieve those results. However, wouldn't it be cool to develop a general microcontroller based bias circuit that could be easily adapted to work with FET's or BJT's for the obsessed DIY'ers? Then you could control the outputs so as neither one ever turns completely off but rather has a programmed minimum conductance. I may look a little deeper into this after all. Looks like I need to get familiar with Assembly Language……