03-08-2014, 12:01 PM
Honestly, I wouldn't wish to argue against that current source.
It is primarily an active load for Q3. This transistor needs to have a high gain as it is the error amplifier. The better this is, the better the DC regulation will be, but also, the lower the output ripple will be.
A bipolar transistor - which is voltage-controlled like a valve, FET or MOS-FET - has a value of gm (transconductance) that depends on the collector current. That's about 1mA here, which gives a gm of 35-40, depending on temperature... If more gm is required, the current source can be turned up, subject to the usual limitations of heat, etc... A fixed resistor would result in a gm that falls as the output voltage is turned up, but increased output voltage means higher closed-loop gain, so the last thing you want here is falling open-loop gain.
The voltage gain of the amplifier is gm times RL, but as RL is a current source, it will have a high resistance. Now, whatever the amplifying device - BJT or MOS-FET - the gain will be gm.Rl. So, arguing that with a MOS-FET doesn't require the current source load doesn't really make sense to me - they normally have pretty low values of gm in my experience.
Also, as well as increasing open-loop gain, which will reduce ripple at the output via NFB, the active load will prevent "ripple injection" by isolating the collector of Q3 from the incoming unregulated supply. If you wanted to use a resistor here, to avoid this problem you'd have to split it in two and decouple the mid-point with a big capacitor. Which again reduces the voltage gain... In my humble opinion, the current source is a more elegant solution here.
I would say that this design is pretty optimal for this topology. In terms of drift, the Zener isn't the best reference, but the metering won't show the drift, and any circuit being powered by it won't mind! For the intended purpose, it's a job well done.
The only thing I'd add, other than to echo the gate-protection mentioned earlier, is to comment on the voltage rating of the resistors. I spend a lot of time replacing failed resistors that are >100k with >100V across them (approx!). Standard carbon-film devices aren't the best in this location, so it might be worth making up series combinations that will have less voltage stress. Or consider metal-glaze devices. But I will say that this is general advice - I'm not saying that you will run into this problem with the unit as constructed, it's just something that would be at the front of my mind if I was building such a thing.
Oh, and I'd add a resistor in series with the base of Q4, just to protect the BE junction from "spikes". A sudden overload will turn on Q4, of course, which in turn will turn off the pass device Q2. But that process takes a few µs to take place... 1k or thereabouts will do.
Once fitted, you then have the option of including foldback current limiting. A resistor from the base of Q4 to ground is all you need.
Finally, I always like to include a thermal cut-out in a power supply. A simple bi-metallic switch is all I tend to use - they are cheap insurance.
Good job
Mark
It is primarily an active load for Q3. This transistor needs to have a high gain as it is the error amplifier. The better this is, the better the DC regulation will be, but also, the lower the output ripple will be.
A bipolar transistor - which is voltage-controlled like a valve, FET or MOS-FET - has a value of gm (transconductance) that depends on the collector current. That's about 1mA here, which gives a gm of 35-40, depending on temperature... If more gm is required, the current source can be turned up, subject to the usual limitations of heat, etc... A fixed resistor would result in a gm that falls as the output voltage is turned up, but increased output voltage means higher closed-loop gain, so the last thing you want here is falling open-loop gain.
The voltage gain of the amplifier is gm times RL, but as RL is a current source, it will have a high resistance. Now, whatever the amplifying device - BJT or MOS-FET - the gain will be gm.Rl. So, arguing that with a MOS-FET doesn't require the current source load doesn't really make sense to me - they normally have pretty low values of gm in my experience.
Also, as well as increasing open-loop gain, which will reduce ripple at the output via NFB, the active load will prevent "ripple injection" by isolating the collector of Q3 from the incoming unregulated supply. If you wanted to use a resistor here, to avoid this problem you'd have to split it in two and decouple the mid-point with a big capacitor. Which again reduces the voltage gain... In my humble opinion, the current source is a more elegant solution here.
I would say that this design is pretty optimal for this topology. In terms of drift, the Zener isn't the best reference, but the metering won't show the drift, and any circuit being powered by it won't mind! For the intended purpose, it's a job well done.
The only thing I'd add, other than to echo the gate-protection mentioned earlier, is to comment on the voltage rating of the resistors. I spend a lot of time replacing failed resistors that are >100k with >100V across them (approx!). Standard carbon-film devices aren't the best in this location, so it might be worth making up series combinations that will have less voltage stress. Or consider metal-glaze devices. But I will say that this is general advice - I'm not saying that you will run into this problem with the unit as constructed, it's just something that would be at the front of my mind if I was building such a thing.
Oh, and I'd add a resistor in series with the base of Q4, just to protect the BE junction from "spikes". A sudden overload will turn on Q4, of course, which in turn will turn off the pass device Q2. But that process takes a few µs to take place... 1k or thereabouts will do.
Once fitted, you then have the option of including foldback current limiting. A resistor from the base of Q4 to ground is all you need.
Finally, I always like to include a thermal cut-out in a power supply. A simple bi-metallic switch is all I tend to use - they are cheap insurance.
Good job
Mark
But I suspect that the importance of the voltage rating of any particular type of resistor construction goes far beyond that. So, in summary, for the various types of resistor construction that are commonly available, what happens - short, medium and long term - if that voltage rating is exceeded? Or, to put it briefly, what is the importance of voltage rating in resistors, and why?
I suppose such things will happen from time to time. 
