Some of this is related to my amateurish discussion with JensRasmussen of aspects of his circuit design, some not.
1) What is the mechanism of the instability that gate resistors are supposed to ameliorate? If you have a compact enough layout and drive the gate from a high impedance (as high as the get resistor would have been) does a gate resistor do any good?
2) How would you achieve temperature-independent biasing, like current proportional to absolute temperature for better control of BJT transconductance without lots of matched devices for making the bias circuit?
3) What sorts of time constants / and ESPECIALLY what capacitor types (low leakage electrolytics??) are good to use for the DC rejection cap that makes the amp have unity gain at DC? Or is it better to use a servo op-amp...
I hope these (basic) questions aren't considered annoying.
1) What is the mechanism of the instability that gate resistors are supposed to ameliorate? If you have a compact enough layout and drive the gate from a high impedance (as high as the get resistor would have been) does a gate resistor do any good?
2) How would you achieve temperature-independent biasing, like current proportional to absolute temperature for better control of BJT transconductance without lots of matched devices for making the bias circuit?
3) What sorts of time constants / and ESPECIALLY what capacitor types (low leakage electrolytics??) are good to use for the DC rejection cap that makes the amp have unity gain at DC? Or is it better to use a servo op-amp...
I hope these (basic) questions aren't considered annoying.
This thread on biasing might have some useful info.
http://diyaudio.com/forums/showthread.php?s=&threadid=3881
GP.
http://diyaudio.com/forums/showthread.php?s=&threadid=3881
GP.
Thanks, Circlotron!
What I was really trying to get at, with the question on biasing, is really a kind of silly question, becasue I was thinking about input stages, second gain stages, etc, not so much output stages,
and of course it is in the output stage where the temperature change is so significant.
But if you are trying to achieve controlled gain, then even ignoring rapidly varying thermal effects, the ambient temperature has a significant effect on gain, and it seems there are two ways to deal with it:
Either use lots of degeneration resistance, which has other benefits too but kills the gain,
Or bias the circuits with a current that increases with temperature.
The simple and elegant Vbe/R circuits that people tend to use to set bias currents give the opposite result ...
Come to think of it, as long as you can adjust the resulting current with a trimpot ( I hate trimpots though ), you don't necessarily need precise matching to build a delta-Vbe current source, just trim out the error...
With any luck I'll just convince myself that the idea is silly because the operating temperature of the small signal stages is just not going to change enough to make a difference.
What I was really trying to get at, with the question on biasing, is really a kind of silly question, becasue I was thinking about input stages, second gain stages, etc, not so much output stages,
and of course it is in the output stage where the temperature change is so significant.
But if you are trying to achieve controlled gain, then even ignoring rapidly varying thermal effects, the ambient temperature has a significant effect on gain, and it seems there are two ways to deal with it:
Either use lots of degeneration resistance, which has other benefits too but kills the gain,
Or bias the circuits with a current that increases with temperature.
The simple and elegant Vbe/R circuits that people tend to use to set bias currents give the opposite result ...
Come to think of it, as long as you can adjust the resulting current with a trimpot ( I hate trimpots though ), you don't necessarily need precise matching to build a delta-Vbe current source, just trim out the error...
With any luck I'll just convince myself that the idea is silly because the operating temperature of the small signal stages is just not going to change enough to make a difference.
Random responses:
MOSFETs are scary fast under the right conditions, and
when the circuit they sit in is not ideal, they can turn
that non-idealness into a fine multi-MHz oscillator. The
Gate resistor slows them down enough to play well with
others.
in BJT's the thermal drift we deal with is usually Vbe, not
current gain. This does not require matching as such, since
the Vbe is quite consistent from part to part, and we never
parallel BJT's without an emitter resistor due to the potential
for thermal runaway and current hogging.
)
Somewhere below 1 Hz is pretty acceptable. If you could
find a perfect capacitor, there would be little reason to
pursue servos and such. Aaarr, but there's the rub: the
capacitor be the devil himself, says I...
MOSFETs are scary fast under the right conditions, and
when the circuit they sit in is not ideal, they can turn
that non-idealness into a fine multi-MHz oscillator. The
Gate resistor slows them down enough to play well with
others.
in BJT's the thermal drift we deal with is usually Vbe, not
current gain. This does not require matching as such, since
the Vbe is quite consistent from part to part, and we never
parallel BJT's without an emitter resistor due to the potential
for thermal runaway and current hogging.
)Somewhere below 1 Hz is pretty acceptable. If you could
find a perfect capacitor, there would be little reason to
pursue servos and such. Aaarr, but there's the rub: the
capacitor be the devil himself, says I...
voltage gain stage
Suppose you wanted to have well defined gain in a BJT stage that had little or no degeneration. I'm not talking about power output stage.
You would want the current to increase proportional to absolute temp.
For some reason, this doesn't usually seem to matter. Is it because we assume that the low power stages of amplifiers are going to operate at or near room temperature, so don't worry about temperature compensation, or is it becasue they are usually degenerated with emitter resistors to the point that their gain is defined as a resistor ratio.
I'm just probing for general design philosophy, hoping anyone interested will pipe in.
And about those devilish capacitors. They are a problem. The latest digikey catalog does list some suggested replacements for panasonic Z-series low leakage electrolytics.
Servos are looking better all the time though, I suppose.
Suppose you wanted to have well defined gain in a BJT stage that had little or no degeneration. I'm not talking about power output stage.
You would want the current to increase proportional to absolute temp.
For some reason, this doesn't usually seem to matter. Is it because we assume that the low power stages of amplifiers are going to operate at or near room temperature, so don't worry about temperature compensation, or is it becasue they are usually degenerated with emitter resistors to the point that their gain is defined as a resistor ratio.
I'm just probing for general design philosophy, hoping anyone interested will pipe in.
And about those devilish capacitors. They are a problem. The latest digikey catalog does list some suggested replacements for panasonic Z-series low leakage electrolytics.
Servos are looking better all the time though, I suppose.
Mirlo,
Problem with gate of jfets is not instability, but oscillation at high frequency (50 MHz or so) This is very dependent of the layout, wiring, and occurs with high transconductance devices such 2SK170/370.
The solution is to insert a resistor in the gate connection (470 ohms) as close as possible of the device, even on the device lead itself. It's not a good idea to drive a jfet by a high impedance : no effect on HF oscillation (because of stray capacitances) and possible increase of distortion.
Regards, P.Lacombe.
Problem with gate of jfets is not instability, but oscillation at high frequency (50 MHz or so) This is very dependent of the layout, wiring, and occurs with high transconductance devices such 2SK170/370.
The solution is to insert a resistor in the gate connection (470 ohms) as close as possible of the device, even on the device lead itself. It's not a good idea to drive a jfet by a high impedance : no effect on HF oscillation (because of stray capacitances) and possible increase of distortion.
Regards, P.Lacombe.
PSRR
Well, since you normally care about PSRR at some frequency other than DC, to _simulate_ it you have to do two AC simulations, maybe make two copies of the circuit inside the simulator, one with AC stimulus on the power supply and one with AC stimulus on the input. Divide the output from the first circuit by the output from the second.
I wouldn't want to try to actually _measure_ it at frequency. To measure it at frequency you probably need to have a power amp that can handle capacitive loads, and transformer-couple it in series with the power supply of your amplifier under test to stimulate it through the power supply input. Ick.
Well, since you normally care about PSRR at some frequency other than DC, to _simulate_ it you have to do two AC simulations, maybe make two copies of the circuit inside the simulator, one with AC stimulus on the power supply and one with AC stimulus on the input. Divide the output from the first circuit by the output from the second.
I wouldn't want to try to actually _measure_ it at frequency. To measure it at frequency you probably need to have a power amp that can handle capacitive loads, and transformer-couple it in series with the power supply of your amplifier under test to stimulate it through the power supply input. Ick.
mirlo said:
The reason why MOSFET's are instable is the large capacitances which in combination with very small inductances (the connections to the transistor) forms a very good oscillator. I have seen the equations for that in some application note somewhere (Hitachi?) but I can derive them because I'm too tired.
When you insert a ferrite bead or a resistor in series with the gate you create losses, enough to stop the oscillation.
Rule:
Small (50-100 ohm) or no resistance = risk for oscillation but a fast transistor
Medium value (100-470 ohm) = no oscillation (if you are lucky) and a little bit slower transistor
Too high (470-2 kohm) value = Way too slow transistor!
mirlo said:
Check my monster phono amp for the design of the DC-servos. I use "plastic" caps in the DC-servo and acheive 0,4 Hz. I use try to use max 1 µF plastic for DC-servos.
Check also my input bias servo (at the input of the preamp), this servo is extremely slow with MC-pickup, minutes! With MM pick under 1 minute (don't remember the exact value)
If you use plastic caps there is no problem with leakage and if you don't have so high resistor values you won't get trouble with bias currents. Don't use more the 100 kohm to 1 Mohm (depending on opamp).
To answer your question: If you don't use DC-servo (and use a ordinary DC-blocking cap) you don't need to think about leakage. An electrolythic cap has around 1 Mohm parallel resistance as worst case and this compared to 100-2200 ohms is nothing. The gain will be slightly higher than 1.
What is the best?
DC blocking cap
Pros: Cheap, small space
Cons: Don't last forever, limited lifetime when it's warm.
DC-servo
Pros: Easy to get desired speed, long lifetime
Cons: Complicated (more than one cap), takes up space, can be more expensive than a single cap.
Distortion? Not really a big difference.
Audiophileness: Some like DC-servos, some don't
Mirlo,
I don't know the exact mechanism which causes distortion when driving impedance is high, but I have experienced this, and this phenomenon has been reported by Jean Hiraga in l'Audiophile, 15 years ago.
Very high transconductance jfets have a large gate, which can exhibit significant non-linear leakage current. Such current, for instance, 1 nA, in 1 Mohm impedance, can cause 1 mV voltage drop. If the input signal is 100 mV, this can cause 1% distortion... depending on the exact nature of the driving impedance.
And, as you have written, non linear capacitances can cause further distortion, by non linear Miller effect, but especially at high frequencies, and depending on the effective drain load.
Another possible cause is reverse transconductance, but I have no informations about this...
Regards, P.Lacombe.
I don't know the exact mechanism which causes distortion when driving impedance is high, but I have experienced this, and this phenomenon has been reported by Jean Hiraga in l'Audiophile, 15 years ago.
Very high transconductance jfets have a large gate, which can exhibit significant non-linear leakage current. Such current, for instance, 1 nA, in 1 Mohm impedance, can cause 1 mV voltage drop. If the input signal is 100 mV, this can cause 1% distortion... depending on the exact nature of the driving impedance.
And, as you have written, non linear capacitances can cause further distortion, by non linear Miller effect, but especially at high frequencies, and depending on the effective drain load.
Another possible cause is reverse transconductance, but I have no informations about this...
Regards, P.Lacombe.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.