Explain a current source?
In layman's terms?
Why, sure, glad to.
Let's come at this from the backside. Let's talk about voltage regulators. So, what is a voltage regulator? A voltage regulator is a circuit that takes a raw, incoming DC voltage with all its warts and imperfections and smooths it out into a constant, predictable voltage. No bumps, no wiggles, just pure, clean, steady DC voltage. If you look at it on an oscilloscope, there's nothing there; it's just a straight line, nice and boring, just the way we want it to be.
Let's say that we've got a load that draws current in spurts, like the output stage of a class B amp. If there's no regulator in the circuit, just a transformer and rectifier feeding some caps, the voltage on the caps will drop when the circuit demands a large amount of current (unless the power supply is very, very large indeed). This isn't good. It changes the behavior of the amp. So we stick a regulator in.
The regulator works by chopping off all the voltage above a certain preset level (yes, this is wasteful, but necessary). If you set a regulator for 20V, then feed it 30V, it will lop off the 10V that it doesn't need, allowing 20V to come on through. If the circuit demands more current, the regulator delivers more current, but at that same preset 20V. This is the crucial point for our present purposes--a voltage regulator holds the voltage steady, but allows the current to vary according to the needs of the moment.
Okay, let's turn back around and look at current sources. The thing that makes a current source mysterious is that it has multiple names: current source, current sink, and...wait for it...current regulator. Fasten onto that last name and things will become clearer. Whereas a voltage regulator maintains a steady voltage and allows the current to vary, a current regulator (aka current source) maintains a steady <i>current</i> and allows the voltage to vary.
Think of a circuit as a variable resistor. From the current source's point of view, there's somebody out there twisting the knob back and forth, back and forth, with no rhyme nor reason. The load changes unpredictably. That's okay. Current sources are patient little fellows. They're also obedient. If you tell your current source to deliver 10mA, it will do everything in its power to deliver 10mA no matter what the load does. If the load goes to an effective 10 ohms, the current source will set itself to .1V (10 ohms * .010A =.1V, simple Ohm's Law stuff). If the load suddenly goes to 1k, then the current source will set itself for 10V. At all times and in every way possible the current source will do its dead level best to deliver 10mA, simply because that's what you told it to do.
Okay, so where's that leave us in terms of an Aleph output stage?
The bottom transistor is the actual amplifier. It gets its signal from the front end differental (the two IRF9610s back-to-back to the left of the schematic) and amplifies it, in this case adding both voltage and current. The current source is the MOSFET and NPN (plus a few resistors and caps) up above. Normally, and there are numerous commercial and DIY examples of this, you would expect to see a conventional current source up there. The resistor on the upper MOSFET's Source sets the current and the MOSFET delivers it. The NPN is there to oversee the operation and kinda nudge the MOSFET back into line if it does a less than perfect job; this being the real world, the MOSFET benefits from a little help in accomplishing its task. (In fact, you can build an even simpler current source without the NPN, but performance would suffer a bit.)
If we were to leave the circuit here, this is what would happen: signal would enter the lower MOSFET and cause an output signal to appear at the Drain. The current source, seeing that the voltage is varying, and seein' as how varying voltage is second nature to a current source, goes along for the ride. As the voltage varies, the speaker (whose other side is connected to ground--nominal 0V) will see current travel back and forth, depending on whether the signal happens to be swinging above 0V or below. Music will come out of the speaker.
Back to the Aleph part of the deal, which is the crux of the matter. What Nelson did was add a few parts to the current source, so that it's even more attuned to what the lower MOSFET is doing.
As the signal leaves the circuit, the very last thing it does is pass through an array of resistors (R22-25 in the Aleph 5). The fact that these resistors are in series with the load means that they serve to sense the current going out the back door. This signal travels back up into the innards of the current source via R21 and tells the current source that it's okay to vary a little bit, as long as it does so in accordance with what the lower MOSFET is doing.
Seems a little weird to be telling a current source <i>not</i> to be delivering constant current, but as long as it does so in synchrony with the lower MOSFET, it's cool. If you want to see a voltage analog to this, take a look at Part 3 of the current Zen series over at
www.passdiy.com, wherein Nelson puts together a voltage regulator that...well...it varies the voltage. Again, in synchrony with the output of the amp.
If you want more of the grisly details, check out the patent. It's not all that difficult to follow.
Incidentally, Nelson likes to point out that although the Aleph current source varies on an instantaneous basis, when taken as an average over time, the current is indeed constant.
So what happens when you remove R21 is that you end up with a "normal" current source, i.e. one that doesn't dance along with its partner. With R21, you give the current source a little bit of wiggle room to dance along with the music in the current domain.
Incidentally, the IRF9610 sitting by its lonesome up above the differential is also a current source. It's one of the conventional kind that doesn't dance, current-wise. The Zener diode and the resistor to ground set a reference voltage. The resistor above its source is what sets the current, when referenced against that 9V. In principle, you could use a similar arrangement for the output current source, using the same 9V Zener and 10k resistor at the Gate of the upper MOSFET, but substituting a 10 ohm resistor below the source (better make it a 5 or 10W resistor), but you'd have to give up the "Aleph" part of the circuit to do it. You're better off with the NPN.
Howzzat?
Grey