LED cathode biasing in push-pull input stage

[exeric]

There's one other really interesting thing that the spice simulation shows. I was worried that I wouldn't get all the good effects of a CCS as a plate load if I put it in the cathode circuit, namely a flat load curve and a more efficient load. It seems its plenty efficient in the cathode circuit. I'm getting 85 volts Peak to Peak with the CCS in the cathode and only 47 volts without the CCS and using cathode resistors unbypassed to create the equivalent bias. Everything else in the simulated circuit is exactly equivalent.

I'm beginning to think that putting a CCS in the cathode circuit has exactly the same effect as putting it in the plate circuit as strange as that seems. Of course this simulation was with a "perfect" CCS that has infinite dynamic impedance. I have a hunch that a real world imperfect CCS behaves slightly different when placed in a cathode or the plate circuit.

[Tweeker]

Whether a plate CCS is perfect or not is not so significant once you consider the effects of the grid resistor of the next stage.

[SY]

The effect in cathode and plate are quite different. Most of the standard texts (RDH, Terman, Landee, Jones...) dissect the diff amp in great detail, and it is worth spending an hour or two working through the equations and equivalent circuits. Morgan Jones, in particular, goes into great detail about the use of CCS.

"Real world" CCS are not perfect, but they can easily be made perfect enough that they can be considered as perfect. A cascode bipolar job in the cathode can easily be made to have a source Z in the 10M+ range, compared with a typical cathode impedance of a few k-ohms.

[exeric]

I thought I'd update you on some new information I've found using LTspice simulations. As usual things are never quite as simple as they seem at first, both regarding my initial assumptions and (I think) other people's assumptions.

Most of my original mistakes were made from accidentally optimizing different circuit topologies. I've really tried hard to make equivalent DC operating points in my latest comparisons at the very least. I used a model for each CCS that is perfect, i.e., infinite dynamic impedance. I created a typical Push Pull cathode bias input stage schematic in LT spice. The tubes are 5687s that have a subcircuit defined amplification of 18.5. The input on each tube is 2.8 volts peak to peak, which is the equivalent to the standard 2V RMS/2 or 1 V RMS input per tube.

Rather than describe each circuit in turn I'll just say that I tested for LED biasing/ resistor biasing and combined that with No CCS/cathode CCS/plate CCS. So I had a total of 6 circuit topologies to test. I didn't include cap bypass of resistors because I found that the amplifications and waveforms were identical to LED biasing. This to NOT to say that led biasing probably doesn't sound superior to caps. I suspect they do. Here's the results.

No CCS/resistor biasing = 32V P-P , 11.5 AF (amplification factor)
No CCS/LED biasing =39V P-P, 14 AF
CCS on Cathode/resistor biasing =34V P-P, 12 AF
CCS on cathode/LED biasing = 40V P-P, 14.25 AF
CCS on plate/resistor biasing =49V P-P, 17.5 AF
CCS on plate/LED biasing = 50V P-P, 17.85 AF

There is a clear advantage to putting the CCS on the plates. Several things can be noted but probably the most important one isn't shown in the numbers. I did a waveform image of the power expended instantaneously through the 5687s and its grossly distorted between the upper and lower half of the sine wave. The only case in which this distortion did not occur was where a CCS is in the plate circuit (one CCS is required for each tube plate circuit if the amp is capacitor coupled.) This indicated to me if there is no CCS in the plate circuit, and the more the following stage draws current, then the more distorted the ampfication will be and the less symmetry between the upper and lower halves of the sine wave.

One more thing. LEDs in the case of circuits using CCS do not address the problem of equal voltage division both sides of the push-pull circuit. It is questionable whether they are equivalently accurate to resistor division of voltage. I just put them in to show how amplification changes with them. This is another case where a CCS in the plate circuit is superior in a push-pull circuit to the other types. You lose almost nothing in amplification by using resistors over LEDs, just 1 volt P-P.

[exeric]

I should add that when I said that the waveform is grossly distorted you gotta know I meant that it was for the waveform of each individual tube. I know they combine and cancel in push pull. But you never have "identical" tubes so it's still be to get rid of it rather than depending on cancelling it.

I've got some screen shots I'd like to attach of LTspice simulations but when I do previews they never show up. So since I'm new to this apparently I'm doing something wrong...