Hallo experts, please Mr. Pass,
What about the range of the yet not mentioned IRF-FETs?
There are some web offers with IRFP 460, 3710, IRFZ 44 etc. What would decide if they were champions or audio garbage?
Why do we see the 04x, 14x and 24x only? - please forget their price this time.
Are there some limits regarding Ciss, Rdson or other parameters? In my particular case, the A-X would be the basic topology, with single output FETs.
I think your help would be of general interest.
Looking forward to your quick kind answers,
Laci
What about the range of the yet not mentioned IRF-FETs?
There are some web offers with IRFP 460, 3710, IRFZ 44 etc. What would decide if they were champions or audio garbage?
Why do we see the 04x, 14x and 24x only? - please forget their price this time.
Are there some limits regarding Ciss, Rdson or other parameters? In my particular case, the A-X would be the basic topology, with single output FETs.
I think your help would be of general interest.
Looking forward to your quick kind answers,
Laci
differences
Hi Laci,
Normally the IR Mosfets are switches. When you "abuse" them
for linear purposes not all of them are well suited. From long
year experience the bigger the 2nd number, the less suited
is the device. EG 044 is much better (sonically) than 064.
There are many reasons for that e.g. gate-source capacity which
depends heavily on the number of parallel devices on the dye,
If you look at old NP designs (A75) he used 230. As they are
out of production you better use the devices with the lowest
number possible (mostly 140, 240, 9140, 9240).
There are even "rumours" that higher voltage devices such as
240 work better than 140 or even 044, but I personnally have
no proof of that.
Uli
Hi Laci,
Normally the IR Mosfets are switches. When you "abuse" them
for linear purposes not all of them are well suited. From long
year experience the bigger the 2nd number, the less suited
is the device. EG 044 is much better (sonically) than 064.
There are many reasons for that e.g. gate-source capacity which
depends heavily on the number of parallel devices on the dye,
If you look at old NP designs (A75) he used 230. As they are
out of production you better use the devices with the lowest
number possible (mostly 140, 240, 9140, 9240).
There are even "rumours" that higher voltage devices such as
240 work better than 140 or even 044, but I personnally have
no proof of that.
Uli
Hi, Uli!
That would mean, under my items IRFZ44 could be considered, while 460 were worse and 3710 the worst (?).
Regarding Cgs, its absolut value cannot be so critical considering the original Pass projects with a palmful of paralleled devices (4-6 x Cgs), more its nonlinearity.
BTW, in that case is the bandwidth the same because of the also paralleled series gate resistors or having the same driver will be altered?
Laci
That would mean, under my items IRFZ44 could be considered, while 460 were worse and 3710 the worst (?).
Regarding Cgs, its absolut value cannot be so critical considering the original Pass projects with a palmful of paralleled devices (4-6 x Cgs), more its nonlinearity.
BTW, in that case is the bandwidth the same because of the also paralleled series gate resistors or having the same driver will be altered?
Laci
Re: Hi, Uli!
As every fet forms its own lopass pole (Rg +Cgs) it depends
on Cgs too. Driving a huge capacity through Rgs needs a lot
of "overshoot" in the driver and a lot of current too. This in
turn leads to the need of more bias in the driver and more
nonlinearity between driver and output which has to be
corrected by global feedback and can cause distortion through
overload of the driver (SIM).
Uli
ljozsef said:
As every fet forms its own lopass pole (Rg +Cgs) it depends
on Cgs too. Driving a huge capacity through Rgs needs a lot
of "overshoot" in the driver and a lot of current too. This in
turn leads to the need of more bias in the driver and more
nonlinearity between driver and output which has to be
corrected by global feedback and can cause distortion through
overload of the driver (SIM).
Uli
.Rg + Cgs
The higher Cgs the lower Rg should be to achieve the same
pole frequency. As impedance level drops the amount of
transient current to drive that pole rises. Having such poles
in the NFB loop leads to SIM simply because those transients
saturate the driver stage without contributing anything to
open loop gain. Try it, you will hear the difference!
Uli
Edit: not to forget that a low Rg can lead to severe oszillation
as the source impedance of the driverstage drops too low!
The higher Cgs the lower Rg should be to achieve the same
pole frequency. As impedance level drops the amount of
transient current to drive that pole rises. Having such poles
in the NFB loop leads to SIM simply because those transients
saturate the driver stage without contributing anything to
open loop gain. Try it, you will hear the difference!
Uli
Edit: not to forget that a low Rg can lead to severe oszillation
as the source impedance of the driverstage drops too low!
Try it, you will hear the difference!
I would be happy to listen to differences, but also very pleased to be able to choose a proved set of MOSFETs...
Regarding the serial gate resistor, I did not mention dropping its own value, only the resulting parallel resistance in case of a "battery" of devices.
All told,
1. searching AUDIO at the irf.com were not the best solution (interesting results...),
more important
2. to prefere datasheets with low Ciss (IR-notation) and low voltages (?) and Rds (?). Or these letter not so cardinal?
Laci
I would be happy to listen to differences, but also very pleased to be able to choose a proved set of MOSFETs...
Regarding the serial gate resistor, I did not mention dropping its own value, only the resulting parallel resistance in case of a "battery" of devices.
All told,
1. searching AUDIO at the irf.com were not the best solution (interesting results...),
more important
2. to prefere datasheets with low Ciss (IR-notation) and low voltages (?) and Rds (?). Or these letter not so cardinal?
Laci
The first IR number usually indicates voltage, and the higher
number implies a higher voltage and lower transconductance.
This seems to be mostly process (doping), as opposed to chip
geometry.
The second number seems to relate to chip size, and the bigger
the number the bigger the chip. Within a voltage type, this
implies a character similar to parallel versions (just more area)
of the lesser sized chip. Thus a 240 looks like a couple of
parallel 230's, and a 250 looks like a couple of 240's. This is
just a very rough approximation, mind you.
As they chip size goes up, the capacitance increases with surface,
and so does the transconductance, Rds and wattage rating.
This might affect how many you choose to parallel, if any, but
otherwise I have been able to get similar performance from these
different chips by applying this logic.
number implies a higher voltage and lower transconductance.
This seems to be mostly process (doping), as opposed to chip
geometry.
The second number seems to relate to chip size, and the bigger
the number the bigger the chip. Within a voltage type, this
implies a character similar to parallel versions (just more area)
of the lesser sized chip. Thus a 240 looks like a couple of
parallel 230's, and a 250 looks like a couple of 240's. This is
just a very rough approximation, mind you.
As they chip size goes up, the capacitance increases with surface,
and so does the transconductance, Rds and wattage rating.
This might affect how many you choose to parallel, if any, but
otherwise I have been able to get similar performance from these
different chips by applying this logic.
Mr.Pass -
It seems I was successful in generating opinions of best value.
At the low rail voltage levels - common situation for this forum - this logic would mean a rational limiting of die surfaces, wouldn't it? Practically, one should choose possibly minor second rang figures and at increased current/power requirements better to multiply them than changing for higher numbers. Implicit, the catalog parameters sould be focused more on Cgs than Rds, n'est-ce pas (Pass)?
... And I can begin my holiday much more restful.
Thanks to all of you,
Laci
It seems I was successful in generating opinions of best value.
At the low rail voltage levels - common situation for this forum - this logic would mean a rational limiting of die surfaces, wouldn't it? Practically, one should choose possibly minor second rang figures and at increased current/power requirements better to multiply them than changing for higher numbers. Implicit, the catalog parameters sould be focused more on Cgs than Rds, n'est-ce pas (Pass)?
... And I can begin my holiday much more restful.
Thanks to all of you,
Laci
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.