Another kit: All-discrete power supply

[c4757p]

A current limit LED would be very easy. Not a bad idea

[c4757p]

I had to move up to the 100x100mm PCB size bracket, and for the sake of spreading things out a bit more to make a neater PCB, I went all the way to 100x100. Here's the revision B PCB.

[Thor-Arne]

Sorry about the late reply on this, busy with kids and life in general. 

I haven't ordered much semiconductors from AliExpress, but quite a few project boxes and such.
For inexpensive parts like the ones above I don't think the risk is that big, especially if the seller has good feedback score.

If you order parts from Ali, I recommend that you compare prices for with/without free shipping. For some of the tings I've ordered the free shipping ones actually became more expensive.

I've only ordered some diodes and leds from AliExpress,  nothing wrong with those at least.

Anyway,

• MMBT3904 x100 USD2.42
• MMBT3906 x100 USD1.81
• MMBT4148 x200 USD15.00

Seems like the MMBT4148 was expensive. perhaps this can be used instead?
Also, searcing for "3904 sot-23" and "3906 sot-23" both yield better prices than for the MMBT* types.

Tip, if you need bigger quantities than the seller lists, always ask for a quote.

[mrkev]

Hi c4757p,


Can you explane how your Voltage Reference is working...
Q4a base and collector are shorted togetter?
Schematic error or is Blackdog is a ...

OK, the 100 a 150 Ohm base resistor is always necesary, but you already now that :-)

Well, Blackdog is a ... then
Just kidding... You probably missed that Q4a is effectively just a diode and it can be, as the active load Q16a and Q16b is two P-N passes higher than V+, and that differential pair (Q5A and Q5B) can (therefore) go quite high...
Also, because Q4 is a diode, any base resistor is pointless..

[megajocke]

I don't know if I'd trust a 4148 to survive charging the smoothing cap in the D11 position. And with that diode in place is there really a need for the diode in series with the voltage error amplifier? Using a midpoint of the diode string won't help much in any case though. It's a circle of four diodes so it would conduct in both directions anyways.

[c4757p]

You're right, I_FSM is only 4A. I thought it was higher. I'll change that to another 4001.

Shame on me for thinking I was more familiar with a common part.

And with that diode in place is there really a need for the diode in series with the voltage error amplifier? Using a midpoint of the diode string won't help much in any case though. It's a circle of four diodes so it would conduct in both directions anyways.

You know, I don't remember why I did that anymore. I do remember that it stopped a large flow of current in the simulation under some unlikely output conditions... but I can't remember what those conditions were, and I can't make it happen again.

[c4757p]

The text has moved along a bit further - some corrections, some updates to the circuits, and the ToO is written through the voltage error amplifier. I'll post again when I have the full ToO complete.

I've also updated the PCB. Not much has changed, but I labeled things a bit better. I also moved the output and sense jacks to put pairs together.

[c4757p]

This image is brought to you by the letter I, for "god damn post from China is slow" - er, I mean "impatience".

(And yeah, that's a shitty specimen, but it is "good enough")

[c4757p]

Quick meter prototype. A few things need to be adjusted (arc radius is obviously wrong, for instance, and text positions need to be tweaked), but I was more concerned with it looking at least semi-professional, and I think it does.

These little $1.20 meters are actually pretty good. They are small, which I didn't realize, but I like the size. Very linear, and even though the null screw is under the cover, they seem to hold their null well. The resistor inside them looks to have been soldered by a three-year-old (it probably was), but those are coming out anyway. Fairly high current (900uA full scale); that could be good or bad. Acceptable here.

(Initial accuracy was ghastly. 10% out. I suspect they're supposed to be 1mA full scale, not 900uA, but that's what I measured.)

The low voltage version will allow higher current (equal power), so that will have a different meter face. Meter faces will be printed on white card and sprayed with UV-resistant fixative.

[rdl]

Okay, I'm not trying to be a jerk or anything because it looks pretty good, but think about that "100 milliamps". What does the meter actually display? This actually something car manufacturers do all the time and it has always bugged me.

The tach in my truck reads from 0 to 6, but the label says "RPM x 1000". Does this mean the engine only goes from 0 to .06 RPM?

I had a Chevelle in the old days, and a Mitsubishi more recently, where the tach had numbers from 0 to 70 and was labeled "RPM / 100", which actually makes sense. I don't know, I guess it's just a pet peeve of mine.

[c4757p]

I think it is correct. The unit is "100 mA". When it's pointing at 2, that is two units of 100 mA, or 200mA., in the same way that when it points at 20 on the top scale, it is indicating twenty units of 1V.

I could understand confusion (or.... fake confusion ) if I used a multiplication or division sign, but I did not.

Edit: Just finished a test and resistorectomy on all ten, and holy crap these have a bell curve to rival Big Ben. 2.5% my ass! That's OK, they were getting trimmers anyway.

[Thor-Arne]

Quick meter prototype. A few things need to be adjusted (arc radius is obviously wrong, for instance, and text positions need to be tweaked), but I was more concerned with it looking at least semi-professional, and I think it does.

Looking good. 
Perhaps a little to wide lines on the scale if I have to comment on something.

[c4757p]

I agree, I made them a bit thinner. I still like them thicker than the usual, though - easy to read at a distance.

[codeboy2k]

I think it is correct. The unit is "100 mA". When it's pointing at 2, that is two units of 100 mA, or 200mA., in the same way that when it points at 20 on the top scale, it is indicating twenty units of 1V.

I could understand confusion (or.... fake confusion ) if I used a multiplication or division sign, but I did not.

Edit: Just finished a test and resistorectomy on all ten, and holy crap these have a bell curve to rival Big Ben. 2.5% my ass! That's OK, they were getting trimmers anyway.

Meter looks great! yes, one you fix the arc angle it will be even better ! 

Regarding the comments about 100 milliamp on the bottom scale.. I have a suggestion.. you could but 1, 2, 3, 4, 500mA , i.e. just label the last one as 500mA and the rest are single digits.  Or put 1,2,3,4,500 and just put the text at the bottom middle of the meter as "MILLIAMPS" not "100 MILLIAMPS"

Although I do agree with your thinking re: not putting the divide or multiply sign.. that just confuses one way or the other.

Was that PC board you made etched or routed? Looks almost like it could be routed but the right edge in the picture looks like it was etched, so I was not sure.

I like your silk screen with the sections grouped and labeled !  good work!

[c4757p]

Thanks for the kind words.

I considered labeling the meter like that, but I decided that the overly literal perfectionist side of me would hate that... And it's too late now, I just made a big batch of 3 x 18 meter faces, all printed, coated and cut out.

The PCB was etched. I stopped the etch early for the sake of the couple finely spaced traces (I did not design this board for home etching...... it didn't work very well) and then touched it up manually with a knife, so that might be why it looks a bit odd.

Both PCB and schematic are still being changed a bit. I'm making a few subtle changes as I build and exercise the circuits (and perhaps one not-so-subtle change - the line regulation isn't quite what I hoped, might have to change the local regulator to fix that). Also, I put the trimmers for the meter on-board, switched to a bigger TVS diode (one that can absolutely, without a doubt handle the full impulse of popping the fuse under some fairly heavy overload conditions), and added more of the off-board wiring to the schematic.

Speaking of off-board wiring - it occurs to me that not a lot of people know this trick, so I thought I'd share: In KiCad, if you want to place a component that does not exist in the PCB (for off-board wiring diagrams, symbols, etc.), just add a # symbol at the beginning of its reference designator. It will then be omitted from the netlist and PCB. Then, if you like, you can give it a "different", non-# refdes by marking the real one invisible and using one of the custom attributes.

[Thor-Arne]

I'm looking forward to the next update on this project, really promising.

Have you designed the pcb to a specific enclosure?

Speaking of off-board wiring - it occurs to me that not a lot of people know this trick, so I thought I'd share: In KiCad, if you want to place a component that does not exist in the PCB (for off-board wiring diagrams, symbols, etc.), just add a # symbol at the beginning of its reference designator. It will then be omitted from the netlist and PCB. Then, if you like, you can give it a "different", non-# refdes by marking the real one invisible and using one of the custom attributes.

Great tip, thanks. 

[c4757p]

Have you designed the pcb to a specific enclosure?

I have not. Nice, chunky metal enclosures are often rather expensive, but it's usually pretty easy to find something suitable cheap if you don't need a certain, exact style. If I had built this around one enclosure the cost would have gone up quite a bit.

Speaking of off-board wiring - it occurs to me that not a lot of people know this trick, so I thought I'd share: In KiCad, if you want to place a component that does not exist in the PCB (for off-board wiring diagrams, symbols, etc.), just add a # symbol at the beginning of its reference designator. It will then be omitted from the netlist and PCB. Then, if you like, you can give it a "different", non-# refdes by marking the real one invisible and using one of the custom attributes.

Great tip, thanks. 

You're welcome! This is how it internally excludes power symbols from the netlist, so it's not too hackish - I don't expect it to break with an update or anything.

[Thor-Arne]

I tend to source my enclosures from AliExpress, especially the plastic ones. I have good experiences with this store.

I guess you have been reading through the KiCad source to find useful tings like this?

[Jebnor]

A couple noted regarding the manual uploaded 09 January 2014.  (If you're still accepting constructive criticism that is.

S1.2 Par.5: "Both the nominal 650mV threshold voltage and the 300 330(ohm)..."
S1.2 Par.5: ~middle ~middle "A typical A typical"
S1.2 Following the Error calculations:  The paragraph begins "This 2mA is drawn..."  It's a little odd as the calculation was finished with 130uA, then you jump back a calculation and talk about 2mA. Might I suggest moving the reasoning and details of the 2mA for ZD1 to a paragraph before the error calculations.  Something like:  "The VEE regulator is more complicated" -> Paragraph about ZD1 and mA requirements and stability -> Current source Paragraph -> Error Calcs -> Paragraph on error effects.
S1.2 Final Par: Change "It" to "Q3" in Sentence 2.  "It" is ambiguous.

S1.3 P1: "input power rejection compounds" comes in from no where. I think it needs some explanation.
S1.3 Table: Needs a title/label.

General:  pick a consistent scheme for your diagrams.  Fig1.1 used blue and Red with dotted and dashed lines, where Fig1.4 uses red and Green with all dashed lines.   Add a legend to Fig 1.4, which will then eliminate the need for the italics note while referencing Fig1.4 (any such notes, if required, should be with the figure.) In my experience, green is a difficult colour to get right in documents, getting a green dark enough such that it prints well in B+W makes it hard to distinguish from blue.  I stick with Red/Blue.  Further, dotted lines (such as those in Fig1.1) are more difficult to follow with the eye and don't scale well when printed, dashed (------) and dash-dot (-.-.-.-.-.-) lines are best IMNSHO.

Nits/style:
Fig1.1: 1) The {Rectifier -> Regulator -> OUT} Main power path is difficult to follow, too many 90s in the line. Possibly make the Regulator more "inline" from Rectifier to out. i.e. left to right. 2) The Power line isn't easily distinguished from the 'low power' lines such as those driving the current sources off Vcc.

Table 1: Spacing... (As I'm sure you already know.)


One final thing.  LaTeX FTW! 

PS: I'm learning a lot from this.  Keep up the good work.

[c4757p]

I'm still accepting criticism, but some is going to be outdated, because I've done a lot of work on the manual that hasn't been posted yet.

However, you seem to have noticed mostly things I haven't.

300 -> 330 ohm: got it
A typical A typical -> got it
2mA/130uA -> Good point, thank you!
it -> Q3 -> Another good point.
compounds -> And another.
Table title -> And another.

Diagrams: Yes, the colors need to be standardized. I used red/blue in the first diagram for +ve/-ve, but then as I started using blue for annotations, I dropped that. I'll fix this. And you're right about the dotted lines.

Block diagram power path: Yes, the main power path should be better marked.

Table spacing: Noted.

One final thing.  LaTeX FTW! 

FTW indeed And XCircuit for diagrams

TBH I've used LaTeX almost exclusively for long enough that I don't even know how I'd manage a somewhat large document in a traditional word processor. ToC, standardization of styles, figure positions, etc... And I'm significantly less productive on a keyboard without vim editing commands.

[c4757p]

Transient response. First 25V then 10V at a 500mA switched load. The voltage never spikes by more than 35mV or dips by more than 50mV, response in 25us.

Load regulation is within 65ppm from 0-500mA.

Voltage stability is good enough that with no cover, I can watch the setpoint drift (by about 500uV on 10V) as I move and the air around me blows.

[Andy Watson]

On a practical note, what happens if the sense feedback becomes disconnected?

[c4757p]

There is a pulldown resistor on the sense input, so the output will drop. It may not drop all the way to zero, but at least to 3V or so.

Also, I recommend installing 10-100 ohm resistors linking sense to output on the back of the binding posts to provide an even better default. Actual cables will have a lower impedance and "override" the resistors.

[sync]

With a pull down resistor on + sense the output should rise.

Don't forget a diode to protect the 10-100ohm resistor from over current, e.g. connecting + sense to negative terminal. Then the output rise and the output voltage is across the resistor.

[Andy Watson]

There is a pulldown resistor on the sense input, so the output will drop. It may not drop all the way to zero, but at least to 3V or so.

Will it ?
 R24? Spotted that, but, what happens if the demand is more than half? I think the output will go to full bananas!