MOSFET ESP's P3A?

Status
Not open for further replies.
Jump to Latest
#5
Light bulbs change their resistance dramatically when they heat
up. I learnt this many years ago when I measured the resistance
of an ordinary 60W 220V bulb. The value I got was way off from
what it "should" and would result in the bulb dissipating several
kW rather than the 60W it was supposed to. After scratching my
head for a while I realized the reason had to be that the
resistance increase when the filament heats up. Hence, light
bulbs are particularly unsuitable as resistors unless you want
this effect. Light bulbs are, for instance, often used in the
feedback loop of sine oscillators to stabilize the amplitude where
they have a suitable time constant for the task at hand.
 
#6
Christer said:
Light bulbs change their resistance dramatically when they heat
up.
Click to expand...

wouldn't that be a desirable property here? effectively, you are talking about a vari-ster (?) that will protect the output device, especially bjts?

Afterall, isn't that the reason light bulbs even exist? Otherwise, each and everyone of those would have gone up in flame, 🙂
 
#7
millwood said:


wouldn't that be a desirable property here? effectively, you are talking about a vari-ster (?) that will protect the output device, especially bjts?
Click to expand...

let me expand this a little bit. Say you have a "heating element" that was 1ohm at <1amp, and 100ohm at >100amp. What will happen when you short the output terminals? the heating element will start to glow, limiting the current that otherwise would have killed the output device.

so an IRF with such a power resistor would be similar to a hitachi L-MOSFET, in terms of how they react to overcurrent.
 
#8
those biasing diodes

I was thinking about the diodes this morning. They should generate just enough current for the drivers to conduct. The author explicitly warned against using signal diodes as their forward voltage can be too high (he is right on that if you look at the data sheet for 1n4001-4007 vs. 1n4148 for example).

then why don't the author just use two amplified diodes from the same transistors as the driver? That way, you match the vbe characters almost exactly, short of some process variations.
 
#9
millwood said:


let me expand this a little bit. Say you have a "heating element" that was 1ohm at <1amp, and 100ohm at >100amp. What will happen when you short the output terminals? the heating element will start to glow, limiting the current that otherwise would have killed the output device.

so an IRF with such a power resistor would be similar to a hitachi L-MOSFET, in terms of how they react to overcurrent.
Click to expand...


I suppose you are referring to the 0.22Ohm resistors at the
output. While I do see your point, it might be difficult to find
a bulb, or thermistor or whatever, that has the right non-linear
behaviour to do the limiting job and still not prevent the amp
from working. Even then, you introduce even more non-linearity
than there already is in the output stage, so I don't think it is
a good idea.
 
#10
Christer said:
Even then, you introduce even more non-linearity
than there already is in the output stage, so I don't think it is
a good idea.
Click to expand...

that non-linearity is no different from the non-linearity inherent in a hexfet or l-mosfet (in terms of how they work).

but your first point is well taken.
 
#11
millwood said:


that non-linearity is no different from the non-linearity inherent in a hexfet or l-mosfet (in terms of how they work).
Click to expand...

I don't really know, but I suspect a light bulb would be much
less linear.

BTW, you said you wanted to use MOSFETs to get rid of the
drivers. Since the outputs are now Szikloy pairs, do you intend
to use source followers instead?
 
#12
Christer said:
BTW, you said you wanted to use MOSFETs to get rid of the
drivers. Since the outputs are now Szikloy pairs, do you intend
to use source followers instead?
Click to expand...

not really. I could use n and p-channel irfs (for example, I have tons of 540 and 9540). No, there are not really complementry but I have used them before with great success.

However, after giving it some more throught, I realized that the simplicity of the biasing circuitry is indeed due to the the CF driver stage. so if I were to get rid of the bjt drivers, the biasing circuitry would have to be redone - something I do not want to do.

so for now, I will settle with two amplified diodes biasing the cf ops.
 
#13
millwood said:
not really. I could use n and p-channel irfs (for example, I have tons of 540 and 9540). No, there are not really complementry but I have used them before with great success.
Click to expand...

Well, I did mean complementary source followers, of course.
If you skip the drivers and still want a complementary OPS
it would have to be either source followers or common source
(at least I cannot come up with any other possibility, well
quasi-complementary, but then we need drivers again).


However, after giving it some more throught, I realized that the simplicity of the biasing circuitry is indeed due to the the CF driver stage. so if I were to get rid of the bjt drivers, the biasing circuitry would have to be redone - something I do not want to do.

so for now, I will settle with two amplified diodes biasing the cf ops.
Click to expand...

Yes, that was the reason I asked. Since you can't have Szikloy
pairs (or CF or whatever else they're called) without drivers,
there will be some redesign necessary (perhaps not so much,
though).
 
#18
there are two ways to do,

1) as suggested by Nelson. It has the advantage of fewer component counts but need a Vbe multiplier and adjustment.

2) keep the p3a topology and replace the output transistors with mosfets. You can still use the diodes for biasing so no adjustments. But will need two transistors and two power resistors.

personally, I would go with the 2nd strategy.
 
Status
Not open for further replies.
Top Bottom