Skip to main content
Topic: Computer controlled PS/function gen/multimeter (Read 229245 times) previous topic - next topic

Re: Computer controlled PS/function gen/multimeter

Reply #165
The AC adapter isn't a worry, it can be sourced. I have a few here, and I'm pretty sure I can get one at the junk market down the street. What I would worry about on an AC design is any coils, transformers, or other odd parts we might need for the circuit board. These don't really seems to be standardized, and at our quantities I worry we'll be tracking down stock and changing the board constantly for that type of part. Maybe with the adapter no coils/transformers would be needed at all.

This is a complex project, and we've been really fortunate to have some really knowledgeable people contribute great ideas. My original goal was to put a digital trim-pot on an LM317 :)

Fen's design seems like the conclusion reached by lots of engineers who really study these things. It's cheap, well worn, and uses common parts. It seems like the path of least resistance for the regulation block.

It seems now like we're moved onto other details:
*how to supply the regulation block (SMPS, external SMPS, AC adapter, etc)
*how much ripple is acceptable
*regulation of low voltages (1.25 or .6 to 0 volt range)
*how to set the I and V (DAC, external DAC, PWM, etc)

Some of these details are easier on different regulator typologies (1.25-0volts), especially when we adhere to some of the original goals (DC only).

At this time, I would propose that we suggest which features and goals we can defer, just until we get a rough prototype PCB. This gives us time to actually play with hardware, and a test platform to try different stuff.

For example, I feel it is inevitable to do very low voltages. I want to use this as a testing device, and for that it will need to go the full range. The additional difficulty of AC seems to be a snag though. If we leave the "INPUT" section of the diagram off the rough prototype, would it still give us the opportunity to test different input supplies? I'd be willing to accept a gap in a DC-only prototype just to get familiar with the regulator and parts. But, we might still be able to use the rough regulator/sink prototype with different breadboarded power supplies to test negative voltages (and ripple from SMPS supply, etc).
Got a question? Please ask in the forum for the fastest answers.

Re: Computer controlled PS/function gen/multimeter

Reply #166
[quote author="ian"]What I would worry about on an AC design is any coils, transformers, or other odd parts we might need for the circuit board. These don't really seems to be standardized, and at our quantities I worry we'll be tracking down stock and changing the board constantly for that type of part. Maybe with the adapter no coils/transformers would be needed at all.[/quote]I don't think that you'll need any coils or transformers on the PCB, at least not if you can find an AC adaptor (with AC or DC output).  If you do, then you can research a few alternate options and design the PCB for multi-site.  That way, if you need to alter parts it won't require a PCB change.

I have designed with medium-sized SMD inductors for driving something like 170 LEDs from a 26V boost regulator, but each size had a different shape for the pads.  After the second or third alteration (mostly to save my client money on the production parts), I realized that I should have tried to figure out a way to multi-site, even if I had to overlay the different SMD pad shapes.

You could always design for through-hole, but, again, I think you won't need inductors.

Quote
The additional difficulty of AC seems to be a snag though. If we leave the "INPUT" section of the diagram off the rough prototype, would it still give us the opportunity to test different input supplies? I'd be willing to accept a gap in a DC-only prototype just to get familiar with the regulator and parts. But, we might still be able to use the rough regulator/sink prototype with different breadboarded power supplies to test negative voltages (and ripple from SMPS supply, etc).
An input section should be possible which accepts both DC and AC, although the negative voltages might not be so easy with DC input.  I haven't done a search yet, but I think Low Drop-Out diodes can be found which might handle enough current for a PS.

Re: Computer controlled PS/function gen/multimeter

Reply #167
[quote author="ian"]
Fen's design seems like the conclusion reached by lots of engineers who really study these things. It's cheap, well worn, and uses common parts. It seems like the path of least resistance for the regulation block.
[/quote]
Could you be more specific? I believe he posted a few linear designs based on HP lab power supplies, and a few ideas with various switch-mode controllers. I agree about using ideas that are widely used by smart people that have done a lot more research on this and have much more experience.

[quote author="ian"]
At this time, I would propose that we suggest which features and goals we can defer, just until we get a rough prototype PCB. This gives us time to actually play with hardware, and a test platform to try different stuff.
[/quote]
I think the regulation block (i.e. from the power input to the power input, plus the DACs) is the most interesting part of this project which is the easiest to screw up. The MCU side (unless it's involved in regulation) should be quite straightforward. So should the ADC inputs unless you want advanced features like auto ranging. That doesn't mean they can't go wrong, obviously. Isolated USB might be slightly challenging.

[quote author="ian"]
For example, I feel it is inevitable to do very low voltages. I want to use this as a testing device, and for that it will need to go the full range.
[/quote]
I agree.

[quote author="ian"]
The additional difficulty of AC seems to be a snag though. If we leave the "INPUT" section of the diagram off the rough prototype, would it still give us the opportunity to test different input supplies? I'd be willing to accept a gap in a DC-only prototype just to get familiar with the regulator and parts. But, we might still be able to use the rough regulator/sink prototype with different breadboarded power supplies to test negative voltages (and ripple from SMPS supply, etc).
[/quote]
With AC you mean AC input? It's just an optional feature that might allow some more flexibility like generating extra voltages with just diodes and capacitors. I don't see how this is complicated, just pads for a bridge rectifier/diodes and a cap should be enough if we want to use it like DC, and would allow any polarity (at the cost of some voltage drop). They're easy to bridge if we want to skip it.

[quote author="rsdio"]
I don't think that you'll need any coils or transformers on the PCB, at least not if you can find an AC adaptor (with AC or DC output).
[/quote]
Agreed. The only magnetics might be in the buck converter (if we choose a SMPS), and possibly for filtering the output or PWM signals.

Re: Computer controlled PS/function gen/multimeter

Reply #168
I meant the linear designs using opamps. I forgot about the TL4xx design a few pages back though. The linear design seems well worn, and I tend to go with the flow when I'm not an expert (like in analog design).

Yes, I also meant AC input.  Can someone contribute a circuit for the AC input, and check the linear regulator design for any mods to use negative voltages?  Experience and knowledge are the practical barrier to getting this into a first prototype. A sketch or napkin scan is fine, though the Eagle schematic is somewhere around here:
http://dangerousprototypes.com/forum/in ... 29#msg9329
Got a question? Please ask in the forum for the fastest answers.

Re: Computer controlled PS/function gen/multimeter

Reply #169
[quote author="ian"]Yes, I also meant AC input.  Can someone contribute a circuit for the AC input, and check the linear regulator design for any mods to use negative voltages?[/quote]A rectifier circuit is dead simple, so I was thinking that finding a way to avoid the typical 2 V rectifier drop might be of use.  The NXP PMEG3010BEP might work, and the NXP PMEG3020BER handles 2 A continuous, 50 A peak, with a typical drop of only 460 mV, 520 mV max.

The regulator circuit design might be a little tricky, depending upon whether we can rely upon having a neutral or center tap from the AC source.  Otherwise, a center tap reference would have to created some other way.  Maybe jumpers?

Re: Computer controlled PS/function gen/multimeter

Reply #170
[quote author="ian"]
I meant the linear designs using opamps. I forgot about the TL4xx design a few pages back though. The linear design seems well worn, and I tend to go with the flow when I'm not an expert (like in analog design).[/quote]
I like the opamp design, too. Would like to get some feedback from tayken, since he appears to prefer the SMPS design, preferably with a dsPIC doing all the work. I'll try to spend more time on the op-amp design later. If the schematic is correct, the schematic will do 0V to somewhat below Vinput without any extra supplies. That would be cool, since most of these circuits (including the HP/Agilent E3620A) require a floating symmetrical supply around the output voltage. But I think further evaluation of Fen's circuit is a good idea (I think he suggested this, too). This is a pretty significant modification of the design used by HP/Agilent and tons of others, which use extra supplies for bias and the op-amps.

[quote author="ian"]
Yes, I also meant AC input.  Can someone contribute a circuit for the AC input, and check the linear regulator design for any mods to use negative voltages?  Experience and knowledge are the practical barrier to getting this into a first prototype. A sketch or napkin scan is fine, though the Eagle schematic is somewhere around here:
http://dangerousprototypes.com/forum/in ... 29#msg9329
[/quote]
My first idea would be to use two AC wall warts (unless something like a center tapped or double secondary will wart exists), that would be the easiest, but not very elegant. I think splitting the rectified AC voltage in two with something like an op-amp would be hard because it would need to sink and source up to 1.5A or so. Plus lots of dissipation.

Another idea might be to use a voltage doubler (fancy arrangement of diodes and large capacitors), like a singe Delon circuit. I'm not sure how happy the doubler is if you draw lots of current from the center and how stable its voltage is, but it would be a neat solution. Only downside is that you need twice the capacitance you'd normally need (probably about 2x2200uF for 1.5A), and obviously twice the current rating on the wall wart. This would definitely be harder with DC input (eg. laptop supply). I've attached a quick schematic of what I mean in gschem. The 7x12s represent the linear regulator blocks. I've not simulated or thought this through to check if it works at all, let alone works well.

With a symmetrical power supply, we would need a negative version of this regulator (just like the 78xx and 79xx pair), this is probably not worth the effort. With two separate windings, we could just use the same regulator twice (saves design effort and reduces BOM). This is what almost everyone including HP/Agilent does. But it would require a transformer with two secondary windings or two transformers, which I think would require two wall warts.

[quote author="rsdio"]
A rectifier circuit is dead simple, so I was thinking that finding a way to avoid the typical 2 V rectifier drop might be of use.  The NXP PMEG3010BEP might work, and the NXP PMEG3020BER handles 2 A continuous, 50 A peak, with a typical drop of only 460 mV, 520 mV max.[/quote]
Schottky rectifiers would help decrease the voltage drop, although I would be worried about the reverse voltage rating of 30V, especially if we go with a wall wart that delivers just enough voltage at 100V and almost too much at 250V, although it should be enough in a full bridge topology (where it will only see half of Vpp as long as none of them fails open).

Re: Computer controlled PS/function gen/multimeter

Reply #171
It seems like everybody wants to go with the linear design. :) I got the circuits but couldn't get my hands on an o-scope yet to make voltage ripple tests, but Dave's video gives some ideas. If fenugrec's circuit will not be a routing nightmare then it is way to go as everybody wants to use that. Plus voltage ripple is not a problem with the linear one.

Two wall-warts is impractical and expensive, I think. But Delon circuit can probably be used, I don't see any reason why it cannot be used.

Do we really need to use a bridge rectifier? Maybe we should use four separate Shottky diodes if a Shottky rectifier does not exist. But I couldn't understand why we cannot use a regular rectifier.

But if this way cannot be realized, we can use an inverting SMPS to supply negative voltages?

I think PWM with an LC filter can be used to set V and I references if an RC filter will give too much noise.

Re: Computer controlled PS/function gen/multimeter

Reply #172
I'm following this thread for a while now but I have to say I have no idea where the project is attm :(

Few hints (that might already be in the thread but that I missed - apologize if I bring back something already discussed but I went trough all 10 pages now again and I didn't find any of this)

- wrt "reverse input protection" - if the input to the "device" is DC the way I usually solve the "reverse protection" is not by using 4 diodes (and lose 2V + produce heat) but one diode across the input and fast fuse so if you attach source in reverse the diode will make "short" and fuse will blow. If there's no room for the fuse, then I use only one diode on one input lead so if you switch +/- on the input the device will just not get any power

- wrt "precission/calibration" - there are fairly cheap, very precise voltage references out there - 1V, 1.6V, 2V, 2.4V, 2.5V ... so imho using proper voltage reference would allow for simple calibration (as you have a "known" value to compare everything else to

- wrt ADC precision, 0-12V sampled at 8bit gives 0.05V precision, 10bit gives 0.012V ... I believe anything over 10bit will just bring in noise, nothing else, and "any" dsPic can handle 10bit adc at high sampling speed

- wrt DAC, a friend used LTC1590 (2x12bit dac) with great results (I attached his schematic but it is not finished, it works but he want to add much more stuff to it but might give someone some ideas)

- wrt reading the values from shunts / dividers, programmable gain op amps are imho the answer, they solve the "auto ranging" problem without losing precision and without adding noise, check out
  PGA: http://www.microchip.com/wwwproducts/De ... e=en010483
  R2R PGA: http://www.microchip.com/wwwproducts/De ... e=en021311

- wrt input - for 0-30V AC2DC non switching regulator I usually suggest AC transformer with 6/12/24/30V output and a relay that would switch to the proper input depending on the output so you decrease power dissipation on the power transistors (or whatever you use to drive the output)... since this project looks like 0-12V under 2A it is not that important (I even thing there's too much philosophy as initial idea with digital resistor was actually going to work, maybe using 12bit dac instead of digital resistor would be even nicer but with current below 2A and voltage below 12V it's basically very simple ... why make it so complex?)

Re: Computer controlled PS/function gen/multimeter

Reply #173
[quote author="alm"]Schottky rectifiers would help decrease the voltage drop, although I would be worried about the reverse voltage rating of 30V, especially if we go with a wall wart that delivers just enough voltage at 100V and almost too much at 250V, although it should be enough in a full bridge topology (where it will only see half of Vpp as long as none of them fails open).[/quote]Would that "voltage doubler" circuit work better with a full-wave rectifier?  i.e. smoother output from the capacitor and regulator?

Re: Computer controlled PS/function gen/multimeter

Reply #174
[quote author="arhi"]
- wrt input - for 0-30V AC2DC non switching regulator I usually suggest AC transformer with 6/12/24/30V output and a relay that would switch to the proper input depending on the output so you decrease power dissipation on the power transistors (or whatever you use to drive the output)... since this project looks like 0-12V under 2A it is not that important (I even thing there's too much philosophy as initial idea with digital resistor was actually going to work, maybe using 12bit dac instead of digital resistor would be even nicer but with current below 2A and voltage below 12V it's basically very simple ... why make it so complex?)
[/quote]
An AC transformer with multiple taps would be nice, but since we want to avoid bringing the user into contact with mains, a wall wart with AC output currently seems the best option. If you can suggest an enclosed transformer (like a wall wart) with multiple taps, that would be great.

The issue with simpler solutions is usually that you can't regulate the voltage below a certain value (eg. 1.25V) and that you don't have variable current limiting unless you make it almost as complex as this. I did propose a dual LM317 scheme (something like this), but the op-amp idea is more tried and true (at least the original version before Fen's modifications).

[quote author="tayken"]
Do we really need to use a bridge rectifier? Maybe we should use four separate Shottky diodes if a Shottky rectifier does not exist. But I couldn't understand why we cannot use a regular rectifier.
[/quote]
There are Schottky bridge rectifiers, the issue is finding one that has enough current capacity, reverse voltage range, and is cheap. The only issue with silicon rectifiers is the voltage drop, which could be an issue if we use a relatively low voltage wall wart. I think the Schottky rectifier came along when we were planning for both DC and AC input. 2V voltage drop for a 12V DC wall wart is pretty significant. I don't think it's as much of an issue with 24VAC, which is about 33Vpp. It might be an issue if we go with the almost too low at 100V, almost too high at 250V solution.

[quote author="tayken"]
But if this way cannot be realized, we can use an inverting SMPS to supply negative voltages?
[/quote]
In that case I would vote for both supplies to be switching, to keep the design simple. But I think switching supplies (at least the transformerless designs like buck and boost) have exactly the same issue: For a symmetrical output, you either need two supplies of opposite polarity (one inverting and one non-inverting, or one positive and one negative), or two floating inputs so you can connect the negative output terminal of the first regulator to the positive output terminal of the second. I don't think any of the common SMPS topologies can both sink and source current (which you would need if you stacked two positive supplies).

[quote author="rsdio"]
Would that "voltage doubler" circuit work better with a full-wave rectifier?  i.e. smoother output from the capacitor and regulator?
[/quote]
Yes, but the catch is that that one requires a center tap (at least the version I'm aware of), so we're back to square one ;). Might as well directly use that center tap, since we don't really care about doubling the voltage. A switchable voltage doubler would be a nice way to handle the 115V/230V issue, although we couldn't use it for creating a center tap at the same time. This one is actually using the full wave of the AC signal (except the part between +/- one diode drop), so it should be pretty efficient on the transformer if you load both power supplies. The positive part goes into the top capacitor, and the negative part into the lower one. But both sides are only driven from a half wave, so you need double the capacitance to keep the ripple from increasing.

I just did an LTspice simulation (circuit attached) of the Delon voltage doubler, and it behaved just as expected. Ripple is about 5Vpp each with a pulsating 1.5A load (about .75A avg) with 2200uF capacitance per side with an ESR of .05ohm (which seems a pretty typical value for non-low-ESR types). With 1.5A constant load on each side, the ripple is about 10Vpp, so we would need extra capacitance for 1.5A capacity. Increasing capacitance to 2x4700u decreases the ripple at 2x1.5A to 5Vpp again, which would give us 18V minimum at the input of the regulators with 24Vpp input. More capacitance wouldn't hurt, but would probably be expensive and bulky. I used 50Hz since this is the worst case scenario, ripple at 60Hz will be less. I used SS24 Schottkys because these were the first 2A diodes I found, I don't think Schottky is necessary here, any diode with enough current capacity (more than 2A would probably be better) should be fine.

With a symmetrical input (like the voltage doubler), we would have to design a negative version of whatever regulator we use (for example the final linear design from Fen), replacing all transistors by those of opposite polarity, reversing all diodes, move the connections to the op-amps around and turn the electrolytic caps around. Plus we would have to test it separately, since NPN transistors are not identical to PNP, and op-amps behavior different in inverting and non-inverting configurations. Having two similar but inverted layouts might also increase the risk of assembly errors (like the one LED on the Bus Pirate v3). Just doing the same regulator twice would be easier and less risky, but would require to separate secondaries. I'm not actually aware of any negative polarity lab supplies (you usually just stack two positive regulators, since they're floating anyway).

PS: Ian, any chance you can add .asc to the list of allowed file types?

Re: Computer controlled PS/function gen/multimeter

Reply #175
The dual LM317 with the digital pot was my starting point :) I take it you could do the same with a LM317 and op-amp.

I added .asc to the allowed file types.
Got a question? Please ask in the forum for the fastest answers.

Re: Computer controlled PS/function gen/multimeter

Reply #176
[quote author="ian"]
The dual LM317 with the digital pot was my starting point :) I take it you could do the same with a LM317 and op-amp.
[/quote]
I think you'd need two op-amps, and would basically be using the LM317 as a fancy power transistor, and be just as complex as the current idea. You do need some kind of OR circuit of the current and voltage regulation. Only two of the op-amps in Fen's circuit (IC1B and IC1C) are actually critical for regulation, I think. IC1D looks to be just for current measurement, and IC1A to amplify the signal from the DAC to the voltage regulator.

About the symmetrical power supply, I just remembered a commercial design of a lab supply with negative polarity: the Tektronix PS503A (and probably a few hundred others I'm not aware of). It uses both a positive and negative power supply, with opposite polarity transistors, but the two parts are not identical, some changes were probably necessary to compensate for different behaviors. It's not terribly stable, I've got it to oscillate by adding about a meter of wires (inductance) between the input filter cap and the circuit.

One issue with making Fen's design negative polarity is that the LM324 output voltage range is 0V to V+ - 1.5V. So in a negative polarity supply, it wouldn't be able to go below something like 1.5V - Vbe. But if we have symmetrical inputs, we could use the opposite polarity as 'negative' potential.

I think I found an error in Fen's schematic (he didn't present it as a finished product), Q1 was a Darlington in the Agilent design (not indicated in the schematic), and I don't see how a regular MJE3055 will have enough gain for the base current for R2 to be enough. I'm also not a fan of the 'constant current source' provided by (Vin-Vb(Q1))/R2, which will vary with output voltage. Can't imagine how Spice approved of this, unless the output voltage was low and it was optimistic about the Hfe of Q1. Either that, or I'm wrong, which is possible. Will try to study this in more detail and come up with an updated schematic.

Re: Computer controlled PS/function gen/multimeter

Reply #177
According to LTspice, the beta of the MJE3055T at ~350mA collector current is 160, I find this really optimistic, it's even higher than the typical curve from the Fairchild datasheet at its highest point (200mA), let alone worst case (which you should design for).

My suggestion is to move closer to the original Agilent schematic again. It appears to me that with the symmetrical input and AC, we should be able to copy the symmetrical supply around Vout for the op-amps.
- Change Q1 to a Darlington, probably from two discrete transistors since I believe these work better than monolithic Darlingtons.
- Bring back the symmetric power supply. I think we can do this with just linear regulators, without using the extra transformer windings. We create the positive voltage (for the positive regulator) from the input voltage with for example a 5V LDO, we connect the ground terminal to the output voltage (just like label A2 in the Agilent schematic). This would require Vin to be at least about 5V + dropout voltage + Vbe above Vout. Will have to check if this setup is stable enough. This might increase dissipation or limit current/voltage, we might be able to use a voltage multiplier trick for this, it's only low current, so small cap. The negative voltage comes from the opposite polarity (negative) input, with a linear regulator (preferably the same model as used for the negative voltage for the negative polarity regulator, but LDO not necessary). Voltage range might be an issue with this regulator, since it will see Vout+ - Vin- at the input. We might be able to avoid the negative voltage for the positive regulator (the LM324's output voltage range includes the negative rail), but for simplicity sake, I'd like to keep both designs as close as possible. I'll draw a diagram if this explanation doesn't make sense.

We need at least a positive supply for the negative regulator (LM324 can't come close to positive rail). We need something like the negative supply for the negative regulator (and positive supply for the positive regulator) for the current source, although we could also use a transistor-based current source in its place, which would require less voltage overhead. In my opinion, it's easier to feed op-amps with supply rails that are far from any input or output voltages, no worry about running into rails. It makes our work easier and less risky by allowing us to remain closer to the (tried and true) Agilent schematic. It might increase the number of passives, but this shouldn't increase costs by too much (all can be low power SMD).

I'll try to find time to create a schematic in LTspice this weekend, will have to spend some time really understanding the Agilent schematic first. If anyone notices anything stupid or violently disagrees, feel free to point that out before I waste any time ;).

Re: Computer controlled PS/function gen/multimeter

Reply #178
- I liked the idea of using precise voltage sources which was suggested by arhi. We might also use the PGA idea suggested by him.

- wrt negative voltage SMPS idea before: I first thought of suggesting a voltage inverter but there might be current issues related to them. But MC34063 can be used to generate negative voltages, so we can design the linear regulator stage with + and - supplies and outputs. We can also route these positive and negative supplies to Vref pins of the ADC and get signed measurements. Any objections to this idea?

Re: Computer controlled PS/function gen/multimeter

Reply #179
[quote author="tayken"]
- I liked the idea of using precise voltage sources which was suggested by arhi. We might also use the PGA idea suggested by him.
[/quote]
For the ADC? Sure, as long as it's significantly better than the built-in reference in the PIC and the rest of the circuitry makes the extra accuracy worthwhile. Not much point in using a voltage reference much more accurate than the voltage divider. PGA sounds like a good idea, as long as the voltages are within the PGA's input range (although we could put a fixed divider in front of the PGA), and it's not too expensive. Isn't this PGA programmed by connecting various pins to ground/Vcc? But I'm sure we can hook PIC GPIO pins to the programming pins. It would save us (relative) precision resistive dividers, if it's indeed accurate enough.

For the power supplies, it's probably not worth the effort, but you'd probably have to replace the regulators for the op-amp power supplies by more stable versions, and replace the op-amps by something with less offset and noise.

[quote author="tayken"]
- wrt negative voltage SMPS idea before: I first thought of suggesting a voltage inverter but there might be current issues related to them. But MC34063 can be used to generate negative voltages, so we can design the linear regulator stage with + and - supplies and outputs. We can also route these positive and negative supplies to Vref pins of the ADC and get signed measurements. Any objections to this idea?
[/quote]
I don't understand how this fits in with the current idea. Do you mean the symmetric power supply for the op-amps? Or the input for the negative regulator? The former idea would work fine, I guess, as long as PSRR of the op-amps is enough to reject the noise. But I'm not sure if it's necessary, since we already have a supply of opposite polarity for the other regulator. A boost converter for the positive supply above the output voltage (for positive regulator) might be a good idea, but I'd prefer a solution with lower parts count (something like diode, capacitor and linear regulator). These are low-current supplies (only op-amps and base currents, probably <<100mA or so), so simplicity and parts count are the most important, and efficiency is irrelevant.

I've made a rough schematic of what I meant in my previous post about symmetrical supplies for the op-amps. Q1/Q2 are the pass transistors (other transistors and op-amps omitted because I'm lazy^W^W^Wfor simplicity sake). D2/C2/D3/C3 are the voltage doubler that produce a symmetrical power supply (with a significant amount of ripple, since we want to keep our caps as small as possible for price, board real-estate and transformer loading). V+out and V-out are the outputs from the supply. V+cc/V+ee/V-cc/V-ee are symmetrical supplies around this potential for the op-amps. The input for the V+cc and V-ee regulators is about 1 diode drop below the peak input voltage, since there's very little loading, there's almost no ripple (as long as the capacitor is not too small). The 7x12 are arbitrary, they could probably be a much lower voltage (+/- 5V?), and could be LDO. As always, completely untested ;). But I think it should work since there's a significant voltage drop over the pass transistors (Darlington's can't saturate) and a significant ripple, so the max Ve is probably significantly below the peak output voltage from the doubler. But feel free to point out flaws and come up with alternative solutions.

The V+cc and V-ee might not be necessary, if we just use a current source for the base current, will have to spend more time with both the Agilent schematic and Fen's modifications to figure that out, didn't have time this weekend. Will also have to find a better Spice model for at least the power transistor. Don't feel obligated to wait for me, will probably take another week or so. Feel free to come up with other solutions.

If we use DC input, I see no way around a SMPS, but for AC, the voltage doubler seems like a nice solution. Can't imagine SMPS to be much cheaper than the extra capacitance, although it would be smaller. Price would be extra noise and complexity. Both are fair solutions depending on trade-offs.

Edit: just noticed that I missed the connection between C2/C3 and common. Can't edit it now, will fix it in next revision. But I wanted to mention it in case anyone gets confused.