Re: Truth About IRF MOSFETS
If I have ever seen a lateral MOSFET amp with any temp.
comp. it is a rare exception. Rather, AFAIK they are so
commonly used because they don't require one, as for
instance Slone points out in his book.
It could be they would benefint from temp. comp. sonically,
though, if the bias current is far from the zero. temp. coeff.
ppoint.
FACTOR said:
If I have ever seen a lateral MOSFET amp with any temp.
comp. it is a rare exception. Rather, AFAIK they are so
commonly used because they don't require one, as for
instance Slone points out in his book.
It could be they would benefint from temp. comp. sonically,
though, if the bias current is far from the zero. temp. coeff.
ppoint.
Certain IRF mosfets (like the IRF 230 - the TO3 metal cans type) can be used without temperature compensation. For the newer types like IRFP240, thermal compensation is necessary.
Lateral mosfets (Hitachi 2SK1058/2SJ162) do not need temperature compensation. This is one of the key factors that distinguishes 2SK1058 mosfets from the rest.
Lateral mosfets (Hitachi 2SK1058/2SJ162) do not need temperature compensation. This is one of the key factors that distinguishes 2SK1058 mosfets from the rest.
Hi.. I'm resurrecting zombie thread
Is is possible to directly replace (with minor modification) those Hitachi mosfets with TO3 IRF230/9230??
Thanks
okky
The IRF parts need higher bias voltage and thermal compensation. The Renesas parts have lower transconductance, but better (easier to work with) thermal characteristics.
Thanks jackinnj. I was referring to Michael Chua claim above that TO3 metal can IRF devices do not need thermal compensation. Is it true?
You can use the IRFP240/9240 without thermal compensation as Pass does -- in Class A. These are great sounding amplifiers (I have 4 of the F5's running at the moment.) The TO3's I have used are all bipolar.
Here's a link to the IRF datasheet -- it looks like this part of the line was not picked up when Vishay bought IRF's discrete transistor biz:
http://www.irf.com/product-info/datasheets/data/jantx2n6758.pdf
Here's a link to the IRF datasheet -- it looks like this part of the line was not picked up when Vishay bought IRF's discrete transistor biz:
http://www.irf.com/product-info/datasheets/data/jantx2n6758.pdf
Argh always this story about IRF/IRFP not needing thermal compensation. Maybe in some cases assuming certain temperature and heatsink, but in general YES and it's really only a minor problem. Just use a Vgs multiplier to supply bisa, using a similar family IRF MOSFET on the heatsink. It tracks so well that you actually want to avoid using source resistors for a single pair of MOSFETs - but then you need to use true complements.
So:
K1058/J162 - low gm laterals, have zero tempco around 100mA for most sensible operating voltages. They are also pretty good complements. Low gm gives them a bit more linearity and good crossover region assuming they are biassed to 100mA or more, but also high loss. They need about 10V of Vgs for full output current. Max Id is 7A (8A or 10A in some Exicon and Magnetec replacements) but this is quite desceptive - they are amazingly rugged. Vgs overvoltage is more likely to kill them than anything else. DO NOT trust the internal zeners, use external ones.
K1530/J201 - PI-trench types, more akin to hexfets. About 5x the transconductance of the laterals, but lower than HEXFETs of equal ratings. Their stable tempco point is close to half the maximum current so temperature compensation is a must. They have smaller cousins (K1529/J200), and have the advantage of a large MT100 case, which makes it possible to utilize well their relatively low 12A max Id. The treshold voltage is similar to laterals, about 1V. It is possible to use them instead of bipolar outputs with minimal changes to biasing, and that was at one time probably a selling point - quick conversion from existing bipolar designs into all-the-rage MOSFET.
IRF(P) and similar HEXFET parts: High gm, stable tempco is invariably close to maximum Id. High gmalso means high capacitances. Treshold voltages are quite high, about 3-4V, and the tolerance of the treshold voltage is also high from part to part, although parts from the same batch tend to show much more consistent characteristics in this regard.
Input and feedback capacitance nonlinearity is a problem - it becomes quite horrific when Vdg falls to about 5-8V. Don't even think about driving these with low current drivers like those often found in designs intended for lateral MOSFETs.
On the plus side, they are easy to find, cheap, easy to temperature stabilize using a Vgs multiplier made with the same technology MOSFET (usually a TO220 small current IRF part - has teh added benefit of really good thermal contact with the heatsink!), they are quite rugged and there is a huge variety available. They usually need to bi biassed 'hot' - 100-200mA per pair, although with some tricks 40-50mA may be enough. They are at their best in class A designs as linearity improves with increasing voltage and current (and so does the heat generated...)
On the minus side - IRF(P)xxx and 9xxx are NOT electrical but process complements. Because of this, finding good complements is challenging - and no, IRFP240 and 9240 are NOT complementary, one look at the datasheet should be enough. Just about the only thing that is the same, is the breakdown voltage. IRFP240 and IRFP9140 are, however, quite complementary except for the breakdown voltage. Ditto for other larger and smaller parts - like IRF640/9540, IRF710/9610 (although 610 an be used as a usable complement for the 9610 but this is really an exception). Often, one has to look at different manufacturers to find matching P and N parts - say IRFP240 and IXYS makes a 17P20 part which is a decent complement. Unfotunately, due to vast usage in power electronics, the N-type parts get continually improved, while the P-type stay the same and the number of available types seems to get smaller every day - this makes finding complements an even bigger problem.
TO220 parts are quite rugged but not nearly like their TO247 (IRFP) counterparts (or older type TO3 parts). IRFP240/9140 can push close to 100W reliably if on a huge heatsink, do not expect that from IRF640/9540 even though the specs look the same.
Always protect the gates - overvoltage will kill them. Some designs are so flawed that it may happen simply if the amp clips or drives a low enough impedance - even though with gate protection the amp may well fry it's fuses and keep it's MOSFETs alive.
So:
K1058/J162 - low gm laterals, have zero tempco around 100mA for most sensible operating voltages. They are also pretty good complements. Low gm gives them a bit more linearity and good crossover region assuming they are biassed to 100mA or more, but also high loss. They need about 10V of Vgs for full output current. Max Id is 7A (8A or 10A in some Exicon and Magnetec replacements) but this is quite desceptive - they are amazingly rugged. Vgs overvoltage is more likely to kill them than anything else. DO NOT trust the internal zeners, use external ones.
K1530/J201 - PI-trench types, more akin to hexfets. About 5x the transconductance of the laterals, but lower than HEXFETs of equal ratings. Their stable tempco point is close to half the maximum current so temperature compensation is a must. They have smaller cousins (K1529/J200), and have the advantage of a large MT100 case, which makes it possible to utilize well their relatively low 12A max Id. The treshold voltage is similar to laterals, about 1V. It is possible to use them instead of bipolar outputs with minimal changes to biasing, and that was at one time probably a selling point - quick conversion from existing bipolar designs into all-the-rage MOSFET.
IRF(P) and similar HEXFET parts: High gm, stable tempco is invariably close to maximum Id. High gmalso means high capacitances. Treshold voltages are quite high, about 3-4V, and the tolerance of the treshold voltage is also high from part to part, although parts from the same batch tend to show much more consistent characteristics in this regard.
Input and feedback capacitance nonlinearity is a problem - it becomes quite horrific when Vdg falls to about 5-8V. Don't even think about driving these with low current drivers like those often found in designs intended for lateral MOSFETs.
On the plus side, they are easy to find, cheap, easy to temperature stabilize using a Vgs multiplier made with the same technology MOSFET (usually a TO220 small current IRF part - has teh added benefit of really good thermal contact with the heatsink!), they are quite rugged and there is a huge variety available. They usually need to bi biassed 'hot' - 100-200mA per pair, although with some tricks 40-50mA may be enough. They are at their best in class A designs as linearity improves with increasing voltage and current (and so does the heat generated...)
On the minus side - IRF(P)xxx and 9xxx are NOT electrical but process complements. Because of this, finding good complements is challenging - and no, IRFP240 and 9240 are NOT complementary, one look at the datasheet should be enough. Just about the only thing that is the same, is the breakdown voltage. IRFP240 and IRFP9140 are, however, quite complementary except for the breakdown voltage. Ditto for other larger and smaller parts - like IRF640/9540, IRF710/9610 (although 610 an be used as a usable complement for the 9610 but this is really an exception). Often, one has to look at different manufacturers to find matching P and N parts - say IRFP240 and IXYS makes a 17P20 part which is a decent complement. Unfotunately, due to vast usage in power electronics, the N-type parts get continually improved, while the P-type stay the same and the number of available types seems to get smaller every day - this makes finding complements an even bigger problem.
TO220 parts are quite rugged but not nearly like their TO247 (IRFP) counterparts (or older type TO3 parts). IRFP240/9140 can push close to 100W reliably if on a huge heatsink, do not expect that from IRF640/9540 even though the specs look the same.
Always protect the gates - overvoltage will kill them. Some designs are so flawed that it may happen simply if the amp clips or drives a low enough impedance - even though with gate protection the amp may well fry it's fuses and keep it's MOSFETs alive.
Last edited:
You made a very good summary ,Ilimzn.
These ones, and their TO3 variants 2SJ48..50/2SK133..135 ,
from Hitachi, are the only ones that make sense in an "audiophile" class amp...
The other mentioned from toshiba and IR can only be used in
class A amps in respect of their Id/Vgs transfer function.
There is no speciffic sweet spot except that it's usually around 100mA - as there are tolerances. I've seen variance from about 70 to 200mA across many dozens of devices (had to make several sets of matched pairs)...
Wahab, Toshiba and IR can both give very good results if you know how they work. It does sometimes require some specialcircuits - like error correction or feedforward, or similar techniques (eg. CFP-like drive, but this tends to be difficult to tame re oscilaltions). Toshibas are more forgiving. Also, adding degeneration can help but it increases loss quite a bit, as rather high value source resistors are required to make any difference. When using simple circuits, I've found that Toshibas definitely have a sweet spot regarding bias current, usually between 150 and 200mA per pair. IR and the like seem to work the better the higher the bias, but with Toshibas, if you go too high, they tend to start sounding dull. Of course, it's not a simple matter of comparison as this depends largely on the rest of the amp, what I'm talking about are simple linn-derived topologies, with refinements such as cascoding and current mirrors - kind of what you would expect if someone took a proven classic BJT output design and replaced the output pair(s) with MOSFETs.
Wahab, Toshiba and IR can both give very good results if you know how they work. It does sometimes require some specialcircuits - like error correction or feedforward, or similar techniques (eg. CFP-like drive, but this tends to be difficult to tame re oscilaltions). Toshibas are more forgiving. Also, adding degeneration can help but it increases loss quite a bit, as rather high value source resistors are required to make any difference. When using simple circuits, I've found that Toshibas definitely have a sweet spot regarding bias current, usually between 150 and 200mA per pair. IR and the like seem to work the better the higher the bias, but with Toshibas, if you go too high, they tend to start sounding dull. Of course, it's not a simple matter of comparison as this depends largely on the rest of the amp, what I'm talking about are simple linn-derived topologies, with refinements such as cascoding and current mirrors - kind of what you would expect if someone took a proven classic BJT output design and replaced the output pair(s) with MOSFETs.
I tend to use 240/9240's as they are cheap and easy to get.
I know you can get higher wattage IRFP's but I think 150watts is enough to dissipate in that package.
Most obvious is the transconductance of the P-type is half that of the N-ch. When I chose cheap hexfets for my EC amp, this is one very important factor, more than the difference in capacitances.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.