In BJT amps there's no question about it, but in MOSFETS? Are source-resistors really needed to prevent Iq from excursion? Even in parallell output, I believe in theory it shouldn't be needed. I've read about in the Wiki but I would like to see some non-theoretical evidence too it.
I'm gonna do some tests later this day about this, but I'd like to know if anyone has a comment about it
I'm gonna do some tests later this day about this, but I'd like to know if anyone has a comment about it
I believe it depends on the particular devices?, FET's are supposed to have a negative temperature coefficient, so automatically prevent thermal runaway and have no problems in parallel. However, many high power audio devices don't have a negative temperature coefficient, so require resistors just like bipolar.
Hi,
Lateral FETs have a -ve temp coef above about 500mA so are effectively self protecting when used in parallel.
Source resistors here help the circuit to stabilise the quiescent and dynamic currents.
Vertical FETS have a -ve temp coef above about 6A to 40A depending on model. At these kind of currents and at the voltages used in audio they are already destroyed before they get out of the +ve coefficient region.
Source resistors are mandatory for current sharing when driving the output.
At quiescent conditions, things are a bit different.
Very accurate matching can equalise individual Iq resulting in near identical junction and case temperatures when no output current is flowing.
Lateral FETs have a -ve temp coef above about 500mA so are effectively self protecting when used in parallel.
Source resistors here help the circuit to stabilise the quiescent and dynamic currents.
Vertical FETS have a -ve temp coef above about 6A to 40A depending on model. At these kind of currents and at the voltages used in audio they are already destroyed before they get out of the +ve coefficient region.
Source resistors are mandatory for current sharing when driving the output.
At quiescent conditions, things are a bit different.
Very accurate matching can equalise individual Iq resulting in near identical junction and case temperatures when no output current is flowing.
Hi,
Thanks for Your replies. Ok so the tempratur-coef can vary from devices. How about an old faithful IRF540/9540? I think Iv'e got these on my shelf so I shall try with these later. First single output and see if Iq rise when they are hot.
I had a look at the datasheet for IRF540 and the curve Temperature vs. Id is of interest but it seems like the current has to go very high to take these effects that we're talking about.
Thanks for Your replies. Ok so the tempratur-coef can vary from devices. How about an old faithful IRF540/9540? I think Iv'e got these on my shelf so I shall try with these later. First single output and see if Iq rise when they are hot.
I had a look at the datasheet for IRF540 and the curve Temperature vs. Id is of interest but it seems like the current has to go very high to take these effects that we're talking about.
Fritzell said:
Yes they have a neg TC but only at currents higher than normally used in audio bias, so the effect is pretty useless unless you use them in heavy biased class A.
Another reason for the Re is to smoothen the cross-over region to avoid hard xover distortion (it is local feedback), and in case of multiple parallel devices, to help equal load sharing.
Jan Didden
Hi all
actually there are competing mechanisms in mosfets.
But, I think you mean positive temperature coefficient when you said negative.
A positive temperature coefficient makes the resistance increase with temperature, a negative tempco decreases it.
In a Mosfet, the Vt has a negative tempco, while the channel resistance has a positive tempco. This means that for each fet there will be a temperature neutral point, which as people have said already, only works at relatively high current.
Because most fets are biased to a relatively low current (well, maybe 100 mA or so) to avoid crossover, the Vt is still usually in the negative tempco region, so needs a bias stabiliser.
Parallelling MOSFETS without local source resistors can be dangerous - especially at high frequencies and currents, the impedances need to be matched to make sure that one device does not hog the amps and destroy itself.
cheers
John
actually there are competing mechanisms in mosfets.
But, I think you mean positive temperature coefficient when you said negative.
A positive temperature coefficient makes the resistance increase with temperature, a negative tempco decreases it.
In a Mosfet, the Vt has a negative tempco, while the channel resistance has a positive tempco. This means that for each fet there will be a temperature neutral point, which as people have said already, only works at relatively high current.
Because most fets are biased to a relatively low current (well, maybe 100 mA or so) to avoid crossover, the Vt is still usually in the negative tempco region, so needs a bias stabiliser.
Parallelling MOSFETS without local source resistors can be dangerous - especially at high frequencies and currents, the impedances need to be matched to make sure that one device does not hog the amps and destroy itself.
cheers
John
Fritzell said:
Many designs are just good for french fries... Vbe multiplier is not well apted to mosfets biasing because mosfets havent not Vbe at all. But if nothing of better is available (as may be a dedicated biasing circuit as that provided by Linear Technology LT1166), a Vbe multiplier is surely more reliable of a raw potentiometer.
Hi
Piercarlo
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.