Another kit: All-discrete power supply

[c4757p]

Yep. It seems to me that they have some logic in place to determine whether quoting is necessary. It appears to be choking on V+ and V-, which presumably it doesn't like because of the symbols. Why the hell would you not just quote goddamn everything?? Do they pay royalties to the HDD manufacturers for any bits used to encode quotation marks? Freaking hell...

New PCB and schematic attached. I'll wait to order the PCBs until tomorrow morning in case I think of something else I missed.

...oh look, I did. Again:

You should always expect wipers to go open circuit, because they do all the time. Always design your circuit to do something safe if this happens.

Speaking of which, my voltage pot. Er.... thanks for bringing that up.

[Monkeh]

Watch out for those cheap 10-turn pots. They're seriously fragile.

Thanks for the heads-up, I was wondering why they were priced "too good to be true"...

Do you know what the common failure mode is?  e.g. should I expect the wiper to eventually go open-circuit?

The glue on the terminal bar fails and the hilariously thin wire within proceeds to break.

Those Bourns knock off ones. Bonens i think they have written on them. Garbage.

The pot in question is not a Bourns knockoff.

[calexanian]

Bonens....    Chinese names..... Still the best one i have found are "Lusty" batteries.

[BravoV]

Subscribed.

Chris, what is the targeted performance for this little gem of yours ? May be some preliminary/draft  specifications ?

[peter.mitchell]

... preliminary/draft  specifications?...

Don't catch fire.

I have tested the circuit, but using hand-matched transistor pairs without good thermal coupling, as I do not have the specified parts.Line regulation was 1mV on 30V over an input voltage range from 35 to 45V (0.03%), load regulation was about the same from 0 to 500mA. Once I order PCBs, I will fully characterize the circuit for line and load regulation, temperature coefficient and transient response with the correct parts installed,and will make a probably futile attempt at measuring the output noise with the equipment I have on hand...

I guess this is close enough for the moment.

[c4757p]

 

Yes.

(As a matter of fact, this is why I don't presently have more test data.... I stupidly set my prototype down on a couple wire trimmings, stupidly didn't notice until I smelled magic smoke - I was doing a load test and so the current limit on the input was high - and now I have a very tightly packed perfboard with many crispy transistors, including two tediously matched pairs. I couldn't be bothered to rebuild for a prototype that doesn't even have the same part numbers as the final circuit. I'll just wait for it.)

[BravoV]

I guess this is close enough for the moment.

Ok, fair enough, meanwhile, back to stalk&lurk mode. 

[sorin]

If anybody has problems opening the schematic, please let me know.

I cannot open the pdf file. All I get is this:

Delete Cookies and will work

[c4757p]

Something that has been tweeking my pots (as t'were), and I can't quite make my mind up, but: You have the load connect to the collector of the "output amplifier", so the loop gain of the voltage control is defined by the load - which is unpredictable - wouldn't it be better to use a voltage-follower output stage?.

Back to this.

The reason I couldn't get it to work is.... wait for this..........

After reconfiguring the error amp for the loss of one inverting stage, I never moved the compensation network. The negative-feedback compensation became positive feedback.

I was having one hell of a stupid day yesterday.

It's actually significantly better like this. Not just stable in practice, but in theory too. And the line ripple rejection has improved as well, not depending on the feedback loop anymore. Thank you for the suggestion. Just had to sleep on it

New SPICE, schem and PCB coming soon.

People... friends don't let friends design circuits in a hurry.

[megajocke]

Very nice project!

How about using this kind of recitifer circuit so you'd be able to use a single secondary transformer? The voltage might be a bit higher than you wanted though, but you could regulate it down (maybe using a simple zener + resistor shunt regulator?).

PS. Is your Kicad symbol library self-made? It looks really nice, almost like it's not Kicad!

PPS. What is J4 for? Looks like connecting it to the transformer winding would cause a short circuit.

[megajocke]

Maybe something like this?

[c4757p]

Very nice project!

How about using this kind of recitifer circuit so you'd be able to use a single secondary transformer? The voltage might be a bit higher than you wanted though, but you could regulate it down (maybe using a simple zener + resistor shunt regulator?).

That would work, but the voltage would be way too high. As in, 20V too high. I only need 2V there, up to 4 or 5. The intent was to just add an auxiliary winding by hand to the transformer - should only require a few turns of insulated wire between the existing, sealed bobbin and the core, for roughly 3Vrms.

Another possibility would be to drop those volts in a couple diodes, but I didn't want to waste headroom because the specs put me right around a couple common transformer voltages.

PS. Is your Kicad symbol library self-made? It looks really nice, almost like it's not Kicad!

Yup. Thank you!

PPS. What is J4 for? Looks like connecting it to the transformer winding would cause a short circuit.

Somewhat bodgy, though useful, versatility in transformer selection. You can connect a single 24-28V winding across J3 and J5 with a bridge rectifier, or place two 1N4004 diodes in the footprint instead and connect a 49-56V center-tapped winding across all three.

[AndersAnd]

If anybody has problems opening the schematic, please let me know.

I cannot open the pdf file. All I get is this:

That looks like a problem on your end, or with the forum software. The .pdf extension has been stripped. Try saving it and then adding the extension manually.

This happens to all (or most?) PDF-files I have downloaded from this forum. Just gets named "downlaod (1)" or whatever number unused. Very annoying to have to rename the PDF files.

Only happens to PDF-files uploaded to this forum. No problem with from other sites.

(Using Chrome browser on Win7 atm., but can't remember if it happens in FF and IE too).

[c4757p]

Linux FTW! Doesn't give a damn what the extension is, if it looks like a PDF and quacks like a PDF it is a PDF.

New schematic and PCB attached; I'll add the SPICE stuff after I add a loop stability analysis to it.

[AndersAnd]

Sigh, I fear this topic will turn this topic into yet another Linux vs. Win debate. 

[GeoffS]

Sigh, I fear this topic will turn this topic into yet another Linux vs. Win debate. 

No it won't.  I won't let it and I'm a Linux bigot from way back

[calexanian]

Maybe something like this?

You need that resistor in series with that zener or the zener will release its smoke.

[c4757p]

Sigh, I fear this topic will turn this topic into yet another Linux vs. Win debate. 

I will be very cross with anyone who does that.

I have no strong feelings about Linux vs Windows. I use Linux because I haven't used Windows seriously in at least six or seven years and I'm no longer comfortable with it. I've no interest in learning another system just to say I have. But I don't care what you use. Both have their advantages and disadvantages.

Not that it wouldn't be just a result of my thread derailment karma to have this turn into an OS war.

Oh damn. I'm not uploading a new schematic until the SPICE is done, but add a 12k resistor in parallel with C7 (for open-circuit wiper failsafe)

[megajocke]

You need that resistor in series with that zener or the zener will release its smoke.

There _is_ a resistor in there, just not in the place one might expect.

But sure, for such a low voltage, not that many extra turns are required on the transformer.

[c4757p]

After spending some time checking and re-checking the PCB, I sent it off. I do somewhat regret not including dummy footprints for layout-dependent frequency compensation, but I don't foresee that being an issue. A few minor, cosmetic changes were made.

I'm not going to prepare a full SPICE sim with stability analysis. That's going a good bit outside the scope of the kit; it does take some intuition to interpret the results and I'm not going to teach a whole lesson on stability and control loops here. I don't feel qualified to write that explanation.

And besides, it's an old-fashioned all-transistor power supply that I designed in the first place without SPICE. Kind of doesn't feel right.

Here's the beginning of the manual, with two sections of the Theory of Operation completed. There's a good bit of Theory of Design in there too. I'll just edit this post to update it as I go. Feel free to help me proofread

[Galaxyrise]

Feel free to help me proofread

From reading this, I think I'm probably the target audience.  Thanks for putting something like this together!

Figure 1.1: I found the green line confusing, how it can connect to the black side of the top rectifier without turning it green.  You do explain later that you consider the circuit having two grounds, but that wasn't apparent to me from the label "local ground".  Even knowing what they mean now, I think I'd be just as happy without the green lines in that diagram.

Figure 1.2: I found the top spans much clearer than the bottom spans.  Perhaps use the same pattern, so have a 32.4V and a >= 4.6V instead of two overlapping spans?  Also, Q1/Q2 are missing a part number but everything else is labelled.

I'm not sure I agree with "there is no one ground"; I can call anything the ground that I want!  I'm sure you told SPICE that one net was the one ground I get what you mean, though, there's a variable voltage gap between the two parts of the circuit. 

I was surprised by the C6 charging math.  I recognize the basic capacitor charging formula, but expected VMax would be 2.4, (and thus a charge time of 5RC.) I played around in LTSpice a bit, and I think I get what's happening.  But your formula is still just a close approximation, right? The LTSpice curves diverge around 85% charged.

I had trouble following the tempco of the zener.  First, you list the tempco as 800ppm (which, as I read the datasheet, is optimistic.  27.4 - 33.4mV/C translates to 830 - 1012ppm.)  But then that's followed with "1.6% in either direction over the specified range" and I thought you meant +-800ppm, but that didn't work out (and zeners don't have a +- tempco anyway.) Perhaps, "-1.6% at 5C and +1.6% at 45C" would have been easier for me to digest. 

I liked the explanation of the constant current source.  Maybe add where the value for R1 comes from?

nits
Page "1": "front-panel Voltage potentiometer".  I don't think voltage should be capitalized there.
Page "1": the next paragraph starts with an awkward sentence.  Perhaps: "This potentiometer feeds a differential voltage into the error amplifier, which compares it to the differential voltage sensed from the output."

[codeboy2k]

You need that resistor in series with that zener or the zener will release its smoke.

There _is_ a resistor in there, just not in the place one might expect.

Yep. +1 to c4757p for posting a V+ referenced schematic on an Australian forum

I hear that's how they keep the electrons from falling out over there

[Bored@Work]

Here's the beginning of the manual, with two sections of the Theory of Operation completed. There's a good bit of Theory of Design in there too. I'll just edit this post to update it as I go. Feel free to help me proofread

Figure 1.2: DC voltage has direction, so all the blue voltage arrows should be uni-directional (one arrow head), not bi-directional. Direction is from higher to lower potential, so the arrow head should be at the more negative end.

Also, I would give all of the voltages quantity names like V1 = ..., V2 = ... This would make it much easier to refer to them later in the text. E.g. nstead of

which loses about 650 mV: from -33 V to -32.35 V.

which loses about Vx= 650 mV: from Vy = -33 V to Vz = -32.35 V.

or

which loses about Vx= 650 mV: from -Vy = -33 V to -Vz = -32.35 V.

Depending on the direction of the arrows.

Quantities and units:

The notation you use for the quantities and units is inconsistent. Your tax dollar helped to produce this document  http://physics.nist.gov/cuu/pdf/sp811.pdf  and the short version http://www.nist.gov/pml/wmd/metric/upload/SP-811-2008-Checklist-Reviewing-Manuscripts.pdf They describe how it is done (although I don't agree with the idea of using the period as the decimal separator, but well, it is the US, you know ...)

Examples:

* The equation at the start of page 3. The math is done without the units, this is an big no-no. And suddenly the result is in ms.

* You also have some magic equations like 40m/330 inline in the text. Well, 40 meters divided by 330 is 12.12 cm

* Taking the last two paragraph as an example:

This 2 mA is drawn through DZ1, a BZX84-series, 2_%-tolerance, 33 V Zener diode, and this is precisely the
current specified for the 2_% tolerance. The dynamic impedance (apparent output resistance at a specific bias
point) of DZ1 is specified by the datasheet to be at worst 80 ?, so the above 128 µA current error will cause a
voltage error up to the product of the two values: 10.2 mV, or 0.03_% on top of the existing 2_% tolerance. Further
error will be caused by DZ1’s temperature coefficient, given as 800_ ppm/deg.C: 1.6_% in either direction over the
specified range. The total voltage error is 2_% + 1.6_% + 0.03_% = 3.63_%.

The output voltage is buffered by Q3, a simple emitter-follower, which loses about 650 mV: from -33 V to
-32.35 V. It will have a 2_mV/deg.C error just like Q1, contributing an additional 0.1_% error, for a grand total of
3.73_%. There will be a slight, additional error in this 650 mV caused by self-heating and by load current, but this
error is nearly constant and will be calibrated away by the trimmers.

Also you should double check your error calculation. I am too lazy to do it now, but error propagation calculation doesn't work by just adding percentages.

[dannyf]

made entirely from discrete components

It is nice to design a discrete power supply for the sake of designing a discrete power supply. It is nicer if it can beat a lowly 3-terminal regulator in performance.

[c4757p]

Oh wow, thanks to both Galaxyrise and B@W, looks like you really spend some time reading through it.

Galaxyrise:
- Yup, no green lines would be better. I agree, they are misleading.
- I'll add an explanation rather than changing the voltage spans. They're like that for a reason: I'm showing the differential voltages that matter: Vbias->VCC (which biases current sources), VEE->V+ (which supplies the regulator), the total V- ->V+ span (which shows the possible output voltage span, as I will address in a section I haven't yet written), and the V+->Vbias headroom (which proves that the current sources won't ever saturate). All others are incidental.
- I did indeed pick a net as ground in SPICE, but it was inconvenient. Many voltages made more sense relative to a different rail. I used V+ as ground because it ruined fewer nice numbers than any other. That's pretty common in discrete-transistor circuits: "ground" is whatever an individual transistor sees at its emitter.
- C6 charging math: 3.5V is Vmax because until the Zener diode begins to conduct, it's not really there. The capacitor charges as if 2.4V doesn't exist and it's just charging towards the rail. The Zener will suddenly clamp it at 2.4V. I'll add a note to this effect.
- No comment on the tempco yet - from reading both you and B@W, it looks like I've got to spend a bit of time rechecking that. There probably is something wrong with it. I'll dig into it a bit later - just got up for morning coffee and have to do other things first.....
- You're right, I should explain R1. Added.

nits:
- Front-panel Voltage potentiometer: Capitalization is correct, and I will add quotation marks as well. It's not a "voltage potentiometer", all potentiometers are (potentiometer!!!!). It's a "Voltage" potentiometer, as in, that is its label.
- Yep, that sentence is awful.

Bored@Work:
- You are absolutely correct about DC voltage needing a direction label. I was careless to just assume the "positive up" convention. I will change the arrows to all point to negative, but I will also add + and - labels, as arrow direction convention can confuse beginners.
- I wasn't sure whether I wanted to label quantities like that. It does make it easier to refer to them, and it is of course standard practice in academic papers and such, but in my experience the uninitiated can find that it makes the text seem denser and harder to follow. I'm still not sure. I'll have to consider this one.
- I really like that NIST checklist. Definitely going to add that to my collection. And use it here, of course
- Ew... magic equations and math without units. Did I really do that? I'm so sorry for anyone who had to see that. Fixed!
- And as mentioned above, I think you are right about the error calculation. I'll go over that again a bit later today.