Another kit: All-discrete power supply

[megajocke]

Do arrows indicating voltage have the arrowhead at the positive end or at the negative end? I've actually seen both used! It seems most people prefer the negative end to have the arrow (matching the electric field direction), and I guess maybe the other convention comes from thinking about vectors or something. I much prefer +- indications rather than arrowheads to avoid the ambiguity, and this was also the way they did it in all courses at my university (in Sweden).

But I also have one thing I was thinking about concerning circuit operation. With your current configuration, depending on at what point in the AC cycle the power switch is turned on, you might be without the half-wave rectified bias voltage for a full half cycle. The time constant for charging the main filter cap would be around 3 ms give or take (assuming a 30 V, 30 VA, 10% regulation transformer), so it seems the main supply has time to come up to almost full voltage in that half cycle, activating the regulator without activating neither the current limiter nor delay circuit.

It also looks like the current limiter and startup delay circuit will stop working on mains brownout.

[c4757p]

Good points.

I may have to change the bias supply to full-wave then. Grr.... I just ordered PCBs.......


I don't think brownout will be an issue. The "delay" circuit should activate whenever the bias supply sinks under 2.4V, and it barely requires any voltage (just Vbe IIRC, though I can't get the schematic to load on my phone). Should take a pretty severe brownout. Maybe I will increase the bias rail voltage spec to reduce the possibility.

It may still cause an issue at very severe brownouts.

Perhaps it will be better to derive the bias supply from a couple diode drops.

Looks like this will get a Rev B. Oh well, at least the main circuitry can be tested on the A boards just fine except for that.

[sync]

Is it possible to change the current sensing to the V+ side and get rid of the bias supply?

[c4757p]

Yes, but the current sense loop was a bit grouchy, so I would rather not restructure it and have to play around with stability. I don't want to have to do a Rev C after finding that B is unstable

OTOH it would be more elegant. Hmm... elegant vs. already works.

[edavid]

P.P.P.S. If anybody has any good info on temperature-compensating a Zener, I'm all ears. I used a 6.2V diode + 1N4148, which seemed like a good combination according to the datasheets. There appears to have been a series of compensated Zeners at one point, but I can't find any now. I'm not going to use a bandgap reference because that violates the "all-discrete" design pattern here. I'm not too worried about temperature compensation, as that's more of a long-term drift property, at least once the unit warms up, but I am making an effort to keep it from being dreadful. It's not a precision instrument, but I do like a power supply that starts up to the same voltage I set it to yesterday.

1N823 is readily available, although a bit expensive @ $3.

There's no reason why you can't build a discrete bandgap reference though...

[c4757p]

I don't know enough about semiconductor physics to really be qualified to explain how it works, and also, from what I understand about them it would be quite tricky to get good performance. The 6.2V Zener + standard diode seems to work well.


OK, I've changed the the power section. The bias supply is now made from diode drops. This version seems to lack all the problems found above, including brownout issues (the bias supply is active way before there are enough V_BE's across C1 to start giving output).

I haven't made any changes that can't be bodged onto the Rev A board, so I will thoroughly test the circuit like that before ordering Rev B. That gives me plenty of time to figure out what else I messed up...

Also: R18 and R19 go down to 1k5 and 390.

[c4757p]

Damn. Analog panel meters are freaking expensive. I think you're going to have to settle for a - *shudder* - digital meter. I'll try to set it up so an analog one can be used instead, but I'm not going to specify one.

[megajocke]

What are your goals for the current limiter?

Passing the load current through the diodes establishing the current limit reference would create some funny effects I think. The diodes heat up and the current limit lowers, for example.

How about instead changing the way the output darlington is biased? You could put in something requiring biasing current from the upper rail to activate the pulldown resistor on the base. I guess it could be done with the same number of transistors as the previous circuit.

[liquibyte]

Damn. Analog panel meters are freaking expensive. I think you're going to have to settle for a - *shudder* - digital meter. I'll try to set it up so an analog one can be used instead, but I'm not going to specify one.

Try looking over at Jameco.  Digikey and Mouser are ridiculous for those.  I don't visit Newark.

[Thor-Arne]

Damn. Analog panel meters are freaking expensive. I think you're going to have to settle for a - *shudder* - digital meter. I'll try to set it up so an analog one can be used instead, but I'm not going to specify one.

You can also try AliExpress if you don't mind the delivery time, like this store.

[c4757p]

I think I can do a custom 4-digit meter for $6. That's nearly on par with the cheapo ones you can buy (none of which seem quite right for the application ). A PIC16F677 ($1.44) to digitize and drive the display, an LTC-4627JR LED display ($1.96), a TL431 ($0.31), perhaps a transistor or two and a small handful of passives. $1 for the PCB, and that's $4.71 plus assorted costs. I'll probably bring back the hand-wound aux winding to allow this to float.

...

You can also try AliExpress if you don't mind the delivery time, like this store.

Even better! Clearly I suck at AliExpress. I missed that completely. Good prices! I don't see any of the type I was looking for (the old-fashioned 10/50/100uA meters meant for use with external circuitry). Am I wrong in thinking that if I open up one of these, I'll be able to remove a series resistor to convert it? I'll probably have to open them anyway, to put in a custom scale.

Speaking of AliExpress, how wary should I be of small semiconductors? MMBT3904, 3906, MMBD4148? They've recently jumped in price on Mouser etc. from "sensible" to "hell no"...

Passing the load current through the diodes establishing the current limit reference would create some funny effects I think. The diodes heat up and the current limit lowers, for example.

Yep. I should have seen that already... Simple solution though - use three 4001 for the drop, and then two 4148 plus a resistor as the reference.

I don't want to restructure the current loop, as I said, because it works well already and it was tricky to make it play nice.

[Monkeh]

I think I can do a custom 4-digit meter for $6. That's nearly on par with the cheapo ones you can buy (none of which seem quite right for the application ). A PIC16F677 ($1.44) to digitize and drive the display, an LTC-4627JR LED display ($1.96), a TL431 ($0.31), perhaps a transistor or two and a small handful of passives. $1 for the PCB, and that's $4.71 plus assorted costs. I'll probably bring back the hand-wound aux winding to allow this to float.

You can probably buy one for that. Speak to Franky.

[c4757p]

I think I'm going back to an analog meter. Fits the style of the design better, and Thor-Arne found some for me for a decent price. I'll still have to pop them open and add a custom scale, though, unless I can somehow find one labeled both 0-35V and 0-600mA...

[edavid]

Speaking of AliExpress, how wary should I be of small semiconductors? MMBT3904, 3906, MMBD4148? They've recently jumped in price on Mouser etc. from "sensible" to "hell no"...

If you are ordering small quantities and you are OK with random Chinese parts, Tayda is pretty good.

If you are ordering a reasonable quantity, you can do just as well with a US distributor.  Arrow is usually cheapest, but their website is frustrating.   Avnet's is a little easier to deal with.  Do you know about findchips.com and octopart.com?

[Thor-Arne]

I think I'm going back to an analog meter. Fits the style of the design better, and Thor-Arne found some for me for a decent price. I'll still have to pop them open and add a custom scale, though, unless I can somehow find one labeled both 0-35V and 0-600mA...

I popped open a 91C4 5V i have here. There is indeed a series resistor, don't know about the other models though.
Soldering isn't that great, but who complaints at that price. 

Edit: url

I also tested it on 5.0012V, varies a little with the angle, but not much.

[c4757p]

Who thought it was a good idea to put the null screw on the inside?

I'll take my chances with the one I said, that one has it in the right place. Thank you for the picture, though. I'm sure they're all pretty similar.

[Thor-Arne]

Who thought it was a good idea to put the null screw on the inside?

I'll take my chances with the one I said, that one has it in the right place. Thank you for the picture, though. I'm sure they're all pretty similar.

It is probably the same inside.

You can ask it they're able to get the exact ones you are looking for, they're pretty easy to deal with. Just use simple languange.

[c4757p]

I'm not going to bother, as I need a custom scale (and don't need enough units to have them do it for me). Might as well snip out a resistor while I've got it open to change the scale plate.

I just ordered a couple, and I'm going to order some parts tonight.

Just so I can do some planning: can I get a PM from anybody who is likely interested in buying one? You aren't committing, just helping me out a bit. To make this clear: this is a "most parts" kit. You must provide the heavy parts (enclosure, heat sink and transformer) as I simply cannot afford to buy these items in any quantity and ship them. I've tried to keep the design relatively versatile in terms of what can be used, so if you have access to any junk parts they may be of use.
 - Transformer: 28V, 30V, 56VCT, or 60VCT, 20VA min, 30VA preferred; recommended: Hammond 187D28.
 - Heat sink: Any TO-220-compatible heat sink that can take 18W.

Two notes on power:
1) I am beginning to seriously consider either a tap-switcher or thyristor preregulator, so that heat sink requirement may go down.
2) If I don't do #1, I am going to prepare a version (this just requires changing component values, nothing else) with a 20V maximum instead of 30V. That cuts the power requirements (both transformer and heat sink) quite a bit.

The anticipated cost of either version will be $25, including PCB, electronic parts, front and back panel parts, and a printed manual.

Edit: #1is out. That's a bit much. Again, out of scope. But I will add a thermal cutoff.

[Bored@Work]

I would delay the decision about the meter once you have done your homework about the error calculations.

What's the point of ruining a (potential) good precision with an imprecise meter?

Or

What's the point of adding a precise meter to a power supply that is (hopefully not) staggering all over the place like a drunken sailor?

[megajocke]

I don't want to restructure the current loop, as I said, because it works well already and it was tricky to make it play nice.

I was thinking along the lines of putting a switch in series with the resistor that turns on the output transistor (collector resistor of the common emitter amplifier after the voltage regulator differential pair), and to have this switch require current from a current source on the upper rail to turn on.

[c4757p]

Maybe I'm being dense, but I don't see how that will work for current limiting. It sounds more like an alternate version of the startup delay circuit, which I no longer need due to the new power input arrangement (bias rail always starts first).

Also, I do not want a current limit that overrides the voltage error amp, causing it to saturate, because I don't want the voltage to overshoot when the load is removed. By pulling on the feedback instead, the voltage amp is still in control and running at full speed.

I would delay the decision about the meter once you have done your homework about the error calculations.

What's the point of ruining a (potential) good precision with an imprecise meter?
Or
What's the point of adding a precise meter to a power supply that is (hopefully not) staggering all over the place like a drunken sailor?

I'm not going to use a very precise meter because they are expensive and it's not really necessary. For example, my Power Designs TP343A has specs that are even better than these, and it has the same old-style analog meters and is just fine. Most of the time you don't need to set the voltage to the millivolt, and if you do, you can just test with a voltmeter. If it's that critical you should be doing that anyway, unless you're using a really nice, newer, software-controlled PSU with a million self-tests.

Like I said before (though I might not have made it clear): I want line and load regulation to be excellent, because those help isolate the effects of the PSU from the effects of the DUT. One of my cheaper PSUs, for instance, has rather poor line regulation, and if I'm powering a circuit and measuring with a 4.5- or 5.5-digit meter, I can see the readings fluctuate as nearby loads on the mains cycle.

I'm not so concerned with long-term drift and things like that. I want the voltage to stay constant during the hour while you are using it; I don't really need it to stay constant tomorrow. But as I added, that would certainly be a nice bonus.

Of course, if the PSU is staggering like a drunken sailor it needs to be fixed, not matched to a crappier meter

By the way, I fixed everything you mentioned in the text, except the error calculation (didn't have time to do that). I'm going to work on that now, and then I'll post an update. Thank you, again, for taking the time to go through it for me.

[megajocke]

Maybe I'm being dense, but I don't see how that will work for current limiting. It sounds more like an alternate version of the startup delay circuit, which I no longer need due to the new power input arrangement (bias rail always starts first).

Yes, that's what I meant - as an alternative if you wanted to keep the separate bias winding. But if you feel you can live with a somewhat higher droupout voltage caused by the diode string it seems like a good solution, and won't require modifications to the transformer.

If you are interested in some more unusual operating conditions you might want to look at what happens if the output current is externally forced higher than the current limit (throught the output diode) or if the voltage is forced higher than the voltage limit. This could happen if supplies/batteries are put in series or parallell respectively.

It might also be interesting to look at what happens if the output is connected to a voltage source, e.g. a battery, while the power is off. Such operating conditions make many power supplies unhappy, so having survival as a requirement might be overkill.

[c4757p]

Thanks for the suggestion. I thought of these things, but forgot a necessary part to make them work properly: a reverse protection diode across the output transistor.

If the output current is forced high through the diode, the sense resistor won't see it (as usual - those diodes are often connected right across the jacks). I may add a slow-blow fuse inline with the output to prevent blowing the diode up. Those low-voltage inline fuse holders used in automotive audio apps are very inexpensive.

Forcing the output high saturates the amplifier, but doesn't do any damage if I add the reverse protection diode.

Connecting to a voltage source while powered down also is harmless with the addition of a diode; the external voltage source will actually power the control circuitry, which will just saturate as usual like when the voltage is forced high. Anything up to 50V will not damage the circuitry (as it is designed for that much in case of excessive mains voltage feeding the transformer - which reminds me, I need a MOV on the input - and R1 must change to 22k, and a 4148 must be added in series with Q12/Q13 to keep the 50V from feeding back up to VEE), and anything over about 48V will trip the TVS diode on the output.

[BravoV]

Like I said before (though I might not have made it clear): I want line and load regulation to be excellent, because those help isolate the effects of the PSU from the effects of the DUT.

This, is exactly what I wanted to hear, and confirms that I'm not subscribing the wrong thread. 

Way to go Chris. 

[c4757p]

 

Did I say a magic word?