Spehro Pefhany said:
The 2N506x is a jellybean part, made by a bunch of manufacturers.
On Semi, Teccor, Central Semi, ST, Philips (there's 5). They have been
around (and reasonably priced) for maybe 30 years that I know of, and
are still going strong.
Eg.
http://www.onsemi.com/pub/Collateral/2N5060-D.PDF
Note: "preferred device"
The only gotcha to be a bit careful with is the Vdrm rating of
30V/usec typical, so a bit of snubbing might be required in some
applications.
voila.
You mean you don't just start production if the prototype works? ;-)
no, not at all. I worked on an existing mil UPS design in the states a few
years back. Many battery failures started occurring, so I looked at the
charger - a simple flyback converter with closed-loop voltage regulation,
and a current limit too. All seemed well. Then I measured its behaviour, and
found appalling line regulation - from 115Vac to 230Vac caused the output
voltage to change dramatically, around 50% IIRC. But at 115Vac in, Vout
happened to be right......obviously load regulation was appalling too.
Feedback was opto via TL431, I forget how the current limit worked exactly,
but through the same opto. how could it not regulate?!? This was designed by
the VP of eng, a twit with an ME from MIT, and no real skill. basically they
turned it on, it didnt explode, the voltage looked right, so they shipped
it!
close examination showed a host of layout problems, but the main culprit was
the ultra-slow rectifier diode (Trr spec anyone?) which successfully managed
to use up all the gain margin in the closed-loop controller.
speaking of which, closing a loop can do some interesting things. If you
think about closed-loop response to a disturbance, you can "inject" that
disturbance anywhere you like - for example a load change. Alternatively,
you can build a "self-disturbing" controller by screwing up the calculations
(or the circuit
- effectively injecting a disturbance in a different part
of the loop.
If loop gain is high enough, it will STILL WORK! Of course when doing
transient tests you'll get a worse response, as the control loop is busy
dealing with the self-inflicted problems - hence my comment about gain
margin.can be hard to diagnose, as the system works but with degraded
performance. I did this with a 200kW 3-phase inverter/rectifier controller -
I had a wide variety of screw-ups in my code (sine routines overflowing,
etc) yet it all worked, but not as well as expected. I went thru and
individually calibrated all my maths routines, and found some astonishing
errors - about 20 degrees of my sine calculation overflowed due to crap in a
look-up table, yet my PLL (sync. ref. frame PI controller really) worked
brilliantly, and the UPS wasnt too shabby - 10% load regulation. Given that
the sine routine was used to calculate the PWM switching times, i.e. the
overflow resulting in a burst of incorrect PWM for 20/360 = 5% of the time,
this is quite remarkable!