Amp driver section: What's better, many parallel lower current mosfets, or, as few as posible using high current mosfets?
Say, for a typical Single ended/or/class AB 100 watt rms amp.
Say, for a typical Single ended/or/class AB 100 watt rms amp.
Hi Brian,
Many parallel drivers transistors allow any of them works in a more linear region and disperses less power. The distortion is smallest than few transistors.
Many parallel drivers transistors allow any of them works in a more linear region and disperses less power. The distortion is smallest than few transistors.
With more FETs, though, the greater the total capacitance will be as seen by the drivers (in general, at least). Thus, you may need to increase the drive capability to be able to source/sink this much higher capacitance, i.e. use a Darlington stage rather than a simple single emitter follower.
From a theoretical point of view, more than one device is always bad, because there is no way thal all devices are identical in all respects.
From a practical point of view, it is often very useful to run devices in parallell. When you use local degeneration, you also reduce the effect of each device not being perfectly identical to the next. Now, running several devices in parallell will typically also tend to reduce noise.
So, there are two schools of thought here. Your mileage may vary depending on application and taste.
Petter
From a practical point of view, it is often very useful to run devices in parallell. When you use local degeneration, you also reduce the effect of each device not being perfectly identical to the next. Now, running several devices in parallell will typically also tend to reduce noise.
So, there are two schools of thought here. Your mileage may vary depending on application and taste.
Petter
Petter said:
This is exactly what I wish to analyze before designing the drive stage of my amp. What are the pros & cons with regard to each philosophy?
I would expect that as a designer you would evaluate
the performance of both approaches by building both.
the performance of both approaches by building both.
Isn´t like this;
Many paralelled devices may look good (sound good🙂 but for a given idle current that would mean that every singel device operates in a less than optimum range of its transfer curve?
Few devices biased higher (for the same idle current) would mean they operate closer to the most linear region.
On the other hand, the "many devices" situation will deliver less current/device dynamically, meaning less dist. And the "few devices" situation while starting out in a more linear region, while give more distortion due to higher current demand from each device?
Guess it´s a question of balance and that good results are possible from both "schools" if implemented correct.
With two similar devices of the same voltage rating but one having double power rating, which case gives least input-capacitance; one high power chip or two "1/2 the power" chip?
Guess this is something that needs to be considered when choosing the final solution.
Now... if I could stop reading and speculating and start to build a design from scratch myself some day so I learn for real😀
/Peter
Many paralelled devices may look good (sound good🙂 but for a given idle current that would mean that every singel device operates in a less than optimum range of its transfer curve?
Few devices biased higher (for the same idle current) would mean they operate closer to the most linear region.
On the other hand, the "many devices" situation will deliver less current/device dynamically, meaning less dist. And the "few devices" situation while starting out in a more linear region, while give more distortion due to higher current demand from each device?
Guess it´s a question of balance and that good results are possible from both "schools" if implemented correct.
With two similar devices of the same voltage rating but one having double power rating, which case gives least input-capacitance; one high power chip or two "1/2 the power" chip?
Guess this is something that needs to be considered when choosing the final solution.
Now... if I could stop reading and speculating and start to build a design from scratch myself some day so I learn for real😀
/Peter
I think that more devices with output degeneration would be better solution, just look at Krell amps. There are paraeled devices in the input stage and also in output. Nelson Pass use many output devices in parael (x1000 - 80 devices, I think) so they know what they do and if it works well in output it must work in other stages too. I think that only disadvantage is in increased input capacitance, so you have to calculate how many devices you can drive and of course there is more current needed.
Just don't forget to make good output degeneration!!!
Just don't forget to make good output degeneration!!!
DarkOne,
I think your reasoning is flawed. Mr. Pass uses many output devices beceause he has to, but he does not use paralled input devices. Matching all those devices could be a nightmare and I would only use paralled input and drivers if I could not find a single devive to do the job, besides I don't think Krell's sound that good.
Jam
I think your reasoning is flawed. Mr. Pass uses many output devices beceause he has to, but he does not use paralled input devices. Matching all those devices could be a nightmare and I would only use paralled input and drivers if I could not find a single devive to do the job, besides I don't think Krell's sound that good.
Jam
It bears repeating:
A high current Mosfet is pretty much the same as parallel
low current Mosfets. You can go either way, and there
are advantages and disadvantages.
For a given process, the capacitance and current and so on
are a function of die area. In this regard there will not be
enough difference to worry about. For example, and IRF250
is about the dies size of 2 X IRF240.
With the big device, there is less/no concern about matching.
With the small devices, there usually is more efficient heat
transfer to the sink for a given chip surface area.
When you are dealing with really big industrial mosfets like
some IR makes, you will find usually 4 parallel chips under the
hood, with no matching assurances.
As long as you are matching and using Source resistance, there
is no difficulty paralleling smaller devices to get big current/
wattage ratings.
A high current Mosfet is pretty much the same as parallel
low current Mosfets. You can go either way, and there
are advantages and disadvantages.
For a given process, the capacitance and current and so on
are a function of die area. In this regard there will not be
enough difference to worry about. For example, and IRF250
is about the dies size of 2 X IRF240.
With the big device, there is less/no concern about matching.
With the small devices, there usually is more efficient heat
transfer to the sink for a given chip surface area.
When you are dealing with really big industrial mosfets like
some IR makes, you will find usually 4 parallel chips under the
hood, with no matching assurances.
As long as you are matching and using Source resistance, there
is no difficulty paralleling smaller devices to get big current/
wattage ratings.
A related question is whether lower transconductance MOSFETs
sound better or worse than higher transconductance MOSFETs
for the same array size and idle current.
For example, does an output stage with four pairs of
IRFP140/9140 devices, with a total idle current of 1A, sound any
better or worse than a similar array of IRFP240/9240 devices?
My limited experience with class AB MOSFET output stages so far
suggests that there is a sonic benefit to using the lower
transconductance devices, even when their higher voltage rating
is not needed by the application.
Of course, this is a subjective judgement based on listening to a
particular system and environment. Has anyone else done this
comparison, and if so, what were your findings?
sound better or worse than higher transconductance MOSFETs
for the same array size and idle current.
For example, does an output stage with four pairs of
IRFP140/9140 devices, with a total idle current of 1A, sound any
better or worse than a similar array of IRFP240/9240 devices?
My limited experience with class AB MOSFET output stages so far
suggests that there is a sonic benefit to using the lower
transconductance devices, even when their higher voltage rating
is not needed by the application.
Of course, this is a subjective judgement based on listening to a
particular system and environment. Has anyone else done this
comparison, and if so, what were your findings?
Nelson Pass said:
Nelson,
I love your posts... ...Most of the time I read them and think... ...Well, duhhh, that makes sense!
Brian,
It's probably wise to investigate both methods, if you have some good ears (And it probably won't hurt to try to back it up with some good test equipment!). Often times what works well for one design won't work well for another. If you're designing an amplifier for 10 Watts, you would look at things a little differently than if you were designing one for 100 Watts.
But, for a hundred watts, you're probably going to have to parallel a few output devices. Are you using fets or BJT's?
Also, it's often times more affordable to use several smaller components in parallel than one larger single device. Not sure from a audio performance standpoint though. Like nelson said, try them both. Maybe you won't tell the difference, maybe you will! If you do, let us know! I'm sure we'd all love to hear your results!
-Dan
Joe Berry said:
Are you talking about an amplifier with -Feedback, or open loop?
Interesting point though, and I've wondered it myself. Maybe I'll try some different devices in my amp when I pull that project back out again. Only a few $$ for the experiment... ...no big deal.
-Dan
Dan, my experience with this so far has just been with an output
stage that is NOT included in a feedback loop. So I imagine I was
mostly hearing the effects of a higher vs. lower amplifier output
impedance. This is one reason why I suspect my results may be
system- and room-specific.
stage that is NOT included in a feedback loop. So I imagine I was
mostly hearing the effects of a higher vs. lower amplifier output
impedance. This is one reason why I suspect my results may be
system- and room-specific.
- Status
- Not open for further replies.