19-02-2018, 12:12 PM
Personally, it's far easier to measure Zout than to try to calculate it 
I prefer the "ring of two" current source. It maintains the current using negative feedback, so generally performs better than these sorts of "hold a known voltage across a known emitter resistance" sources. Cascoding these types of sources won't make them inherently better - they just allow you to use different parts to get the voltage handling. Another approach is to use a Darlington instead.
Of course, the hFE has relatively little to do with the output current directly, but if you cock up the design by making R1 and R2 too high, then the base current might make the base voltage change, and that in turn reduces the emitter voltage, which in turn reduces the output current. The "stiffer" the base voltage is, the more constant the current will be. You still have the problem that the emitter current is not the same as the collector current (because Ie = Ic + Ib), but for the application, I can't see that being a real-world issue. It would be good to know why you're so worried about the tail current in this application, BTW...
Luckily not an issue here, but this circuit has poor "compliance" - meaning that the voltage can only get to within ~6V of the negative rail at best. When you add the cascode transistor, you reduce the compliance further, depending on the values of R1 and R2. Generally, you choose them such that the base of the MJE340 is only a couple of volts above the base of the BC337. So that answers one of your questions about the design of this - decide what the zener current should be (perhaps 10mA), add a bit for the base currents (perhaps a milliamp, but it's really not critical), then choose the R1 + R2 combination to provide this. Then, as you'd only like a couple of voltage across R2, that comes out at 180 ohms, give or take.
Back to the "ring of two" source - that can get to within a volt of the rail, typically. Choosing the two resistors is easy - one carries the current, so that is 0.6V / required current. The other needs to provide a milliamp or two, and the voltage across it is the rail minus 1.2V. Easy
Regarding temperature stability, I'd be astounded if that matters - and if it does, there are easier ways to fix it in a solid-state amp, at least. However, I've always understood that 5.6V zener diodes have a positive tempco of 2.1mV/C, and a "standard" PN junction has a tempco of -2.1mV per degree C. So-called "temperature-compensated" zener diodes consist of a 5.6V zener in series with a standard signal diodes, and such a combination can be implemented for a fraction of the cost using a 1N4148. Obviously, all numbers are "engineering" numbers, meaning they are approximate, but good enough.
In a current source, the zener doesn't generate much heat of its own and is likely to stabilise fairly quickly at power-on (ignoring the influence of the surrounding temperature rises). But the transistor is likely to run much warmer, so it's the change in Vbe that will contribute most to any change in current. The "ring of two" current source is much less affected by this, because the voltage across the sense resistance is compared to the Vbe of the control transistor, not the one handling the current.
Yet another problem is Early effect. With the high voltages present, this might raise its head. The "ring of two" is much more resistant to that, too.
If the current really matters, put an op-amp in there. This will give excellent results - limited only by the quality of the voltage reference at the input - but it is over-engineering, which is why it's rarely if ever done commercially. Again, why does the current source stability and impedance matter in this application?
Final thought on that - presumably the LTP is at the input to the amp? What is the common-mode voltage seen by this LTP?
I prefer the "ring of two" current source. It maintains the current using negative feedback, so generally performs better than these sorts of "hold a known voltage across a known emitter resistance" sources. Cascoding these types of sources won't make them inherently better - they just allow you to use different parts to get the voltage handling. Another approach is to use a Darlington instead.
Of course, the hFE has relatively little to do with the output current directly, but if you cock up the design by making R1 and R2 too high, then the base current might make the base voltage change, and that in turn reduces the emitter voltage, which in turn reduces the output current. The "stiffer" the base voltage is, the more constant the current will be. You still have the problem that the emitter current is not the same as the collector current (because Ie = Ic + Ib), but for the application, I can't see that being a real-world issue. It would be good to know why you're so worried about the tail current in this application, BTW...
Luckily not an issue here, but this circuit has poor "compliance" - meaning that the voltage can only get to within ~6V of the negative rail at best. When you add the cascode transistor, you reduce the compliance further, depending on the values of R1 and R2. Generally, you choose them such that the base of the MJE340 is only a couple of volts above the base of the BC337. So that answers one of your questions about the design of this - decide what the zener current should be (perhaps 10mA), add a bit for the base currents (perhaps a milliamp, but it's really not critical), then choose the R1 + R2 combination to provide this. Then, as you'd only like a couple of voltage across R2, that comes out at 180 ohms, give or take.
Back to the "ring of two" source - that can get to within a volt of the rail, typically. Choosing the two resistors is easy - one carries the current, so that is 0.6V / required current. The other needs to provide a milliamp or two, and the voltage across it is the rail minus 1.2V. Easy
Regarding temperature stability, I'd be astounded if that matters - and if it does, there are easier ways to fix it in a solid-state amp, at least. However, I've always understood that 5.6V zener diodes have a positive tempco of 2.1mV/C, and a "standard" PN junction has a tempco of -2.1mV per degree C. So-called "temperature-compensated" zener diodes consist of a 5.6V zener in series with a standard signal diodes, and such a combination can be implemented for a fraction of the cost using a 1N4148. Obviously, all numbers are "engineering" numbers, meaning they are approximate, but good enough.
In a current source, the zener doesn't generate much heat of its own and is likely to stabilise fairly quickly at power-on (ignoring the influence of the surrounding temperature rises). But the transistor is likely to run much warmer, so it's the change in Vbe that will contribute most to any change in current. The "ring of two" current source is much less affected by this, because the voltage across the sense resistance is compared to the Vbe of the control transistor, not the one handling the current.
Yet another problem is Early effect. With the high voltages present, this might raise its head. The "ring of two" is much more resistant to that, too.
If the current really matters, put an op-amp in there. This will give excellent results - limited only by the quality of the voltage reference at the input - but it is over-engineering, which is why it's rarely if ever done commercially. Again, why does the current source stability and impedance matter in this application?
Final thought on that - presumably the LTP is at the input to the amp? What is the common-mode voltage seen by this LTP?
