Member
Joined 2009
Paid Member
I simulated different values over a range of temperatures to find the value that looked more or less the right one, and picked a 'standard' resistor value that was closest. I wasn't looking for perfect, just much better than standard Vbe. In the circuit I measured the bias across the output device emitter resistors between turn-on and later after playing music - it was pretty darn stable.
The value you need might depend on the implementation details of your amplifier, but if you use the same Vas current as in my design I think it'll be close.
Why do you think this? Trial and error in your particular circuit or is there some rationale? To compensate a typical amplifier with a Miller capacitor is better than a shunt load. I think there is a broadly applicable principle that explains why and that would apply to a CFP as well because it's really just a little amp itself.
I am interested in CFP stability because I do plan to use one to improve the thermal stability of the amp - but in a different way. A CFP that drives a EF for the OPS will practically remove the driver thermal drift so the Vbe multiplier can be simpler and more precise. Not sure the oscillation risk is worth the benefit and would like more information.
Best wishes
David
Hi Dave. Partly trial and error, partly rationale. Miller compensation is unresponsive to loading so what happens when you have a reactive load? Bzzt. This can be pretty bad since the base impedance of output stages can be overreactive ( = 2nd order impedance curves). Furthermore Miller compensation loads the driver even when there is no output load, which is a crying shame. Since it is a rail-referenced high-Q reactance against a ground-referenced signal it can get into all sorts of local parasitic mischief. Miller compensation is more appropriate for current-drive amps.
Please folks, stop the presses. I've made a mistake. A CFP with high source impedance is a totally different thing. For Vbe multipliers, Bonsai's suggestion should work well; but it's still a bit overreactive and resonates with its own parasitic capacitance (no problem with lytic bypass). A 15nF B-E cap on the slave can quench overreactivity but slows it down to a grind.
A CFP with a low source impedance however is a totally different thing, and usually the capacitive Gm bypass is the best method here.
CFP stability with high and low source impedance are two totally different things. Stability with reactive source impedance can be different as well. As with load impedances. This is something that feedback theory and open-loop analysis, despite their incredible popularity, have had a hard time teaching us, as it seems you have discovered. But this is not the right thread. Why don't you try the B-E cap and if you're not satisfied, show me the circuit? A CFP for EF drivers can be done, but a CFP power output stage is much more difficult.
I'm curious to know whether anyone at all still wants to use a CFP Vbe multiplier.
Please folks, stop the presses. I've made a mistake. A CFP with high source impedance is a totally different thing. For Vbe multipliers, Bonsai's suggestion should work well; but it's still a bit overreactive and resonates with its own parasitic capacitance (no problem with lytic bypass). A 15nF B-E cap on the slave can quench overreactivity but slows it down to a grind.
A CFP with a low source impedance however is a totally different thing, and usually the capacitive Gm bypass is the best method here.
CFP stability with high and low source impedance are two totally different things. Stability with reactive source impedance can be different as well. As with load impedances. This is something that feedback theory and open-loop analysis, despite their incredible popularity, have had a hard time teaching us, as it seems you have discovered. But this is not the right thread. Why don't you try the B-E cap and if you're not satisfied, show me the circuit? A CFP for EF drivers can be done, but a CFP power output stage is much more difficult.
I'm curious to know whether anyone at all still wants to use a CFP Vbe multiplier.
Member
Joined 2009
Paid Member
You likely won't be too far off if you try the resistor value I used and see what happens.
For LTspice, if I remember correctly, you put a temperature command into the value of the transistor. For example, instead of "2N5550" you put "2N5550 temp=70" for a junction temperature of 70 degrees celcius.
You can get spice to automate it a bit and step through a range of temperatures by using a parameter for the temperatures. For example, you would use a value for the transistor of "2N5550 temp={t}" and then use ".step param t list 40 60 80" which lists the temperatures you want the simulator to use, in this example it is 40C, 60C and 80C. When you run the simulation it will run successive simulations, one for each temperature and then display all the curves on the graph at the end.
Physics demonstrates that broadly applicable rules are very powerful. Feedback theory attracts me because it is applicable despite specific details of the circuit.
The benefit of feedback may manifest as a decrease or increase in, say, impedance, dependent on what sort of source is considered - but there will be a predictable improvement in distortion etc.
So your statement that stability is totally different looks to me like an incomplete analysis. Where would you like to discuss this if I have more specific ideas?
I see little reason to use a CFP OPS in a Class B-ish amp anyway.
Best wishes
David
Feedback theory may be all you need to make a circuit stable, if you are a master at math. But I don't know anyone who understands it well enough to confidently stabilize simple circuits such as the CFP. Don't you think that's strange? There are comparably expedient ways to understand stability.
My method of understanding stability is different but is still broadly applicable, and quite possibly less abstruse to the hobbyist. I would write everything here, but there is so much to tell I really want to write an article about it so things aren't fragmented and confused. So if you want to discuss this I guess you should contact me privately.
My method of understanding stability is different but is still broadly applicable, and quite possibly less abstruse to the hobbyist. I would write everything here, but there is so much to tell I really want to write an article about it so things aren't fragmented and confused. So if you want to discuss this I guess you should contact me privately.
with the two complementary transistor version one could mount one sensor to monitor the main outputs and the second could be mounted on the driver heatsink, if both output and driver were of the EF style.
For a CFP, the two sensors could be on the pre-driver sink and on the driver heatsink.
Last edited:
Like I said in a previous post, if you use a CFP spreader, it will need to be decoupled adequately across the clamp terminals, and it will need some compensation. I dont know of many designers that struggle with this. Most of the problems around spreader design seems to be with the tempco and getting good system thermal stability - the problem being manifestly more difficult with a triple. Cordell discusses quite a few variants in his book. I would treat loop compensation/stabilization of a CFP spreader as pretty trivial - you dont need high bandwidth (hell, 1Hz is enough), and the shunt current for the most part is limited to a pretty narrow range.
I think the main issue is the choice of the transistor which leads some to use more complex multipliers which is unnecessary like cfp. Cfp is prone to oscilation, Darlington is not and has about the same gain and same packages as single tranasistor. In fact using just a single super beta transistor is more than enough for 98 percent of designs and has a positive benefit above the others. The higher the hfe the more temperature sensitive the transistor, as is a lower voltage type to a lesser extent.
I don't understand for what OPS type could be used the Hagerman(CFP only?), it can provide 2Vbe max and for double EF you need 4Vbe and for Triple you need 6Vbe.
I have had no problems with it. Zero. Zip. Nada.
Clearly I must be doing something wrong with it then.
I used a diode string in that position in my TT amp prototype and I was not satisfied with it, not easy to trim the tempco.
I am not referring to a CFP multiplier.
I am referring to the dual transistor (complementary in series) version of the multiplier.
Both of these transistors have equal sensitivity to temperature.
The CFP I do refer to, is the output stage where it can be EF, or CFP.