Loading a gain stage with a constant current source

[Koreth]

I was reading the preview chapter of Merlin's book again, and was going over the section towards the end when it talks about baising. Normally with a preamp gain stage cathode biasing is used because it it simple and straightforward. But when using a resistor as is normally done, the cathode current feedback reduces the gain of the tube. Thus cathode bypass caps enter the picture, to partially or fully bypass the cathode current, and boost the gain of the stage back closer to what it could be if there were no cathode current feedback. Okay, so far so good, but now we have the new problem that adding the capacitor creates a filter which messes with the frequency response of the stage, making gain frequency dependent. Whee.

In the same section, Merlin mentions using one or more diodes with the correct forward voltage drop to bias the tube. Since diodes are not ohmic devices like resistors, once the forward voltage is reached, the current can change all it once, and the voltage across the diode is going to change negligibly, if at all. Cool! Now we've come up with a constant voltage source to bias our gain stage with, eliminating cathode current feedback and allowing us to get both the full gain of the tube, but also across the full frequency response of the tube. Yay!

This got me thinking about tricks that could be done on the anode of the tube. Loading the tube with a constant voltage source would be counter-productive, as the whole point of the stage is to get voltage gain -- a larger voltage swing out the anode than went in at the grid. But what about using a constant-current source to load the tube with? If I'm understanding the wikipedia article on current sources correctly, a constant-current source flows the same current, regardless of whatever voltage is across it. Considerations of unnecessary complexity aside what effect would this have on a gain stage? Best I can figure, the load line for that stage would be a flat line with no slope, but I'm not sure what the implications of that would be. Best I can figure, it probably doesn't have enough benefit (if any) to be worth it, or someone somewhere would have done it already.

Could someone who knows more about this than I do chime in and help me figure out what would happen in this scenario and why?

[mooreamps]

[QUOTE=Koreth;130362]

Cool! Now we've come up with a constant voltage source to bias our gain stage with, eliminating cathode current feedback and allowing us to get both the full gain of the tube, but also across the full frequency response of the tube. Yay!

I wouldn't be say "yay" too quickly..... He doesn't have those circuits set up correctly....


-g

[R.G.]

I've had some experience with CCS loaded circuits. Like constant voltage circuits, there is a difference between ideal sources and real sources, and there's also some "be careful what you want" mixed in.

A real constant current source is easy to do kind of, but harder to do well. A good one usually needs at least two active devices, sometimes more, and they tend to be noisy, as the active devices in them contribute noise that appears directly on the output of whatever you're using this as a load for.

In addition, CCS's are high impedance by nature. In fact, the higher the impedance, the more they approximate a theoretical CCS. That means that for AC signals, they're an open circuit. The real load seen by a CCS loaded stage is the following stage's input impedance and the input of the CCS loaded stage through its output-to-input capacitance.

The input impedance of the following stage is composed of the series resistance and shunt capacitance of any wiring to it, the shunt loading of any bias network, the series impedance of any DC blocking caps, and then the input impedance of the next stage itself. That may be as high as a MOSFET gate (i.e. very pure glass, 20 volts thick) or as low as the base of a bipolar. Even if the following stage has a high nominal input impedance, its reverse feedback capacitance can dramatically lower the effective input impedance if it has any voltage gain.

In practice, the reason to do a CCS loaded stage instead of a resistor loaded stage is either power supply voltage limitations or the serch for high gain. CCS loading gives you high gain, but you lose it easily to any loading. The following stage you can isolate with a high impedance buffer, like a source follower (bipolar emitter followers may have too low an input impedance for really big gains) but you have to do something like cascoding to take the voltage variation off the loaded node to isolate input capacitance. This solves the feedback capacitance problem by loading your gain device with the low impedance input of an impedance coverter to let the high impedance side drive that CCS.

By the time you get all done with this, you either have to not look too closely at what you really get with a simple CCS, or you have to use upwards of 5-6 active devices to get the benefit of CCS loaded stage.

Mother Nature never makes it easy unless you're trying to do what she was gonna go anyway.

[Koreth]

Cool! Now we've come up with a constant voltage source to bias our gain stage with, eliminating cathode current feedback and allowing us to get both the full gain of the tube, but also across the full frequency response of the tube. Yay!

I wouldn't be say "yay" too quickly..... He doesn't have those circuits set up correctly....


-g

I don't follow. I see a tube with a plate resistor, grid leak resistor, and a diode holding the cathode a few volts above the grid when idle, regardless of the current through it, as that's how a diode would act in that position.

What is incorrectly set up about the example diode bias in the circuits given?

[Merlinb]

Could someone who knows more about this than I do chime in and help me figure out what would happen in this scenario and why?

It is quite common to use a constant-current source (CCS) on preamp stages in hifi amps.
You are correct, the load line would become horizontal and the gain would become equal to the mu of the tube. (In practice the following load appears in parallel with the CCS, so you may not achieve a gain of mu exactly).
What's more, you would still get a gain of mu even if you used an unbypassed cathode resistor! A guy named Alex who used to hang around AX84 used to use CCS loads for the input stage of his high gain amps, apparently with excellent results.

However, strange things can sometimes happen when you overdrive a CCS loaded stage, as the current through the valve drops to zero and the CCS is forced to switch off and on again, which can lead to spurious transients. This doesn't matter somuch in the input stage, since it is rarely driven to cutoff.

The cheap way to make a CCS load is to use a bootstrapped anode resistor (see page 131 of my book), and this doesn't suffer from switch-off transients. I have had great results with this method, especially when leaving the cathode resistor unbypassed, which gets you a gain close to 0.9*mu in practice.


I would also be interested to hear why Gary thinks I have those LED biased stages set up wrong (I recently discovered Gary's amps use a lot of this kind of biasing), and Gary seems to enjoy picking holes in my work!