29-05-2011, 01:12 PM
Hi Al,
That'll teach me for not looking up the transistor part numbers! I admit that I also didn't double-check with all the regulator ICs used, but certainly the LM340 recommends an external diode. Either way, it's certainly something I pay close attention to - oh, and I modify those Farnell PSUs to prevent them going wrong in the future, as the driver transistor is slightly awkward to source these days... When I'm back at work, I'll email you a design that I did recently for a simple bench PSU - it uses a mix of op-amps and discrete stuff and it might be interesting for future reference.
As to the temp-co of LEDs, yes it does vary with type, and most types are worse than a standard PN junction (-2.1mV/C), and certainly worse than a 4V7 Zener (5V6 is just as good, perhaps better). But here, I would argue that the temp-co of the reference at this point doesn't matter a jot; this current source is just to isolate the error amplifier from the ripple component on the input. In other words, anything would do (I generally prefer the two-transistor version). But, the advantage of the lower voltage reference is to reduce the drop-out voltage, which, I confidently predict, will improve ripple at higher output voltages - perhaps you'll be able to remove C2, which will probably improve the transient response...
If you'd like to worry about the temp-co of anything, the two diodes (and BE junction of Q5) would be my target. But, although the output might drift slightly, it's probably not worth worrying about IMO. But on the other hand, if you used digital monitoring, then any small drift is somehow "magnified" because of our "human" response to numbers on a digital read-out.
Again: a really nice job.
Mark
That'll teach me for not looking up the transistor part numbers! I admit that I also didn't double-check with all the regulator ICs used, but certainly the LM340 recommends an external diode. Either way, it's certainly something I pay close attention to - oh, and I modify those Farnell PSUs to prevent them going wrong in the future, as the driver transistor is slightly awkward to source these days... When I'm back at work, I'll email you a design that I did recently for a simple bench PSU - it uses a mix of op-amps and discrete stuff and it might be interesting for future reference.
As to the temp-co of LEDs, yes it does vary with type, and most types are worse than a standard PN junction (-2.1mV/C), and certainly worse than a 4V7 Zener (5V6 is just as good, perhaps better). But here, I would argue that the temp-co of the reference at this point doesn't matter a jot; this current source is just to isolate the error amplifier from the ripple component on the input. In other words, anything would do (I generally prefer the two-transistor version). But, the advantage of the lower voltage reference is to reduce the drop-out voltage, which, I confidently predict, will improve ripple at higher output voltages - perhaps you'll be able to remove C2, which will probably improve the transient response...
If you'd like to worry about the temp-co of anything, the two diodes (and BE junction of Q5) would be my target. But, although the output might drift slightly, it's probably not worth worrying about IMO. But on the other hand, if you used digital monitoring, then any small drift is somehow "magnified" because of our "human" response to numbers on a digital read-out.
Again: a really nice job.
Mark
