cmos, mosfets, other mosfets, and biasing

[duck_arse]

I was looking at a contest winner, with the to3 transistors and vintage components, and I was looking at dis' creep somewhere else, and I was thinkink. effdub was converting a cmos tube-sound to discrete mosfets. I have some 2sj49/2sk134 power mosfets left over from the eighties. why not use them in a booster?

(imageshack has done it's magic and shrunk my image, but it's all still there)


so I converted his circuit, part shown top left of the diagram, to a 4007 complementry pair, part shown bottom left to see if it worked. it seemed to, produced buckets of nicely rounded squarewaves (although I didn't listen to it ....). then I replaced the 4007 wit the power fets, as shown on the right. after rearranging the fets so they faced the right way, and without source resistors, I switched on. there was some distorted signal, but the power supply dropped to ~5V, so I switched off.

I thought about what was chewing the supply, and stuck (first 100R) the 27R in. fabulous, 9V stays good, and some nicely rounded output signal. only it wouldn't go much above 2 Volts or so, and something smelt warm. the voltage readings suggest the fets are sucking 119 mA, which won't last a battery too long. lower source R = more current, etc. (12R = 3V6 and 266mA. goodbye, battery!)

so, my question: should I continue because there is an easy fix to reduce current and produce useable gain and output level, or is it going to need driver transistors to bias properly and I should chuck the whole thing?

any help would be helpful.

[Jazznoise]

Why no source resistor? There's nothing there to really limit the supply flow besides Rds!

[R.G.]

Let's think for a minute.

Enhancement mode MOSFETs (that's what you have there) don't conduct at all until you put some voltage on their gate relative to the source, then they begin conducting at a rate of Yfs amperes per volt of additional gate-source voltage.

The hookup you have is using the MOSFETs to actively set the common gate voltages at about half the power supply, depending on the relative gate threshold voltages of the P and N type devices, but feedback is trying to put them in the middle, along with the output voltage.

If the gates are approximately in the middle of the power supply, then both devices are getting a Vgs of about 1/2 the power supply, minus any voltage applied across the source resistors by current flowing through the source resistors. The 2SJ49/2SK134 pair start turning on at about 1V of Vgs, and thereafter allow about 1 ampere per additional voltage of Vgs to flow.

So the circuit is going to try to put half the power supply between the gates and the sources. That's more than a volt, so current starts flowing through both devices at the same time. If the sources are tied to the power supplies, then about 3.5V appears across the gate-sources, and the devices try to produce about 2.5A of current flow through them, on average.

Using source resistors applies negative feedback by reducing the gate-source voltages as the current through the sources makes a voltage across the resistor which is subtracted from the half-the-power-supply the circuit is trying to make happen. Bigger resistors make this happen at lower currents by subtracting more voltage-per-ampere.

So there are two ways to do this. You can increase the source resistors til it pulls less current, or you can lower the power supply til it pulls less current because the total voltage available for the gate-source voltages is less.

Driver transistors will not help the basic problem except by incidentally reducing the available gate-source voltage.

[PRR]

> 2sj49/2sk134 power mosfets left over from the eighties. why not use them in a booster?

Well, they have a ton of input capacitance, relative to small-audio devices and circuits. Sucks highs. Won't hurt nothing, so try it.

However....

> without source resistors, ...the power supply dropped to ~5V

R.G. says it one way. Let me try another way.

As said, the MOSFETs suck "no" current up to 2.5V gate-source, then they suck a LOT of current. OK, you may have to go over 5V gate-source to hit a 10A current, but for "small" systems anything over 2.5V g-s is a LOT of current.

So the two parts act a lot like a 5V Zener.

What happens when you put 9V into a 5V Zener? There's a fight. The battery pulls-up to 9V, the Zener pulls-down to 5V. Who wins? The bigger guy. If you have a 9V 10 Amp bench supply to a half-watt Zener, the Zener smokes. If you have a DuraCell feeding a pair of 100 Watt MOSFETs, the DuraCell loses.

How do you normally use a 5V Zener in a 9V system? Put a resistor in series. A 4K resistor will give 1mA in the Zener and nothing fries. A 400 ohm gives 10mA and the battery lasts all night but not all week. 40 ohms.... well 27+27= 54 ohms gives (9V-5V)/54= 74mA and battery life is short, perhaps an hour.

So how-come the CMOS works? The CMOS transistors are much-much smaller (about 1,000 times smaller). You could even say they have parasitic 500 ohm resistors. Full-on they can hardly pass 18mA.

Use some source resistors. Offhand, anything under 500 ohms is too low for pedal work.

You can instead ground the bottom Source and run the top Source through a resistor to +9V. That is asymmetrical; you probably want to bypass that top source to ground with maybe 100uFd.

[amptramp]

I have done a few bizarre things during my career and one was to use a standard CD4000 series CMOS inverter as a linear amplifier while submerged in liquid nitrogen.  At a mid voltage input, the normally tiny CMOS current went up to 8 mA with just one of the six inverters being used. Every once in a while, we would get huge spikes in the signal that we suspected were due to differential expansion of silicon and ceramic at low temperature (we used the ceramic packages that used to be more common).  8 mA is a large current for a COMS gate but consistent with what you would see in linear service.

[duck_arse]

thanks rg, prr. I'll have another go.

damn those smaller transistors! I thought larger source resistors would reduce the output swing, so didn't go much further with that.

the input capacitance I remembered, from one side of the eti477 having 470pF caps added to the gates of the n channels. (input cap. 900pF/600pF)

[R.G.]

You might want to use smaller, TO-92 MOSFETS. They're about US$0.40 each, and will behave the same way, but at smaller currents like the  CMOS logic stuff.

As an aside the 2SJ/2SK pairs are out of production and highly thought of for some hifi stuff. They may well be worth selling.

[duck_arse]

worth something? they cost me a fortune when I bought 10 pairs way back. I built a mate an eti 477 as a guitar amp. his bandmates were impressed by the amount of smoke that came out, but niether he or I was impressed by the low output and oscillations at high volume. I only found out the other day this amp was referred to as "old smokey".

I put in 100R source resistors, and I get a gain of around 13x, with overloaded valve type output. sounds like a booster, now all I need is some tone control, a non-clanking input buffer and an external supply.

[R.G.]

worth something? they cost me a fortune when I bought 10 pairs way back. I built a mate an eti 477 as a guitar amp. his bandmates were impressed by the amount of smoke that came out, but niether he or I was impressed by the low output and oscillations at high volume. I only found out the other day this amp was referred to as "old smokey".

Yeah, the ETI amp design was... well, novel is a good word, I guess. Sorry about that. There are better designs for using those later MOSFETs.

I put in 100R source resistors, and I get a gain of around 13x, with overloaded valve type output. sounds like a booster, now all I need is some tone control, a non-clanking input buffer and an external supply.

Great. If it works, good.

If you have to rebuild it, do consider the smaller TO-92 parts. And, it still may be possible for you to resell the big ones.