Serial Control of Adjustable Voltage Regulators?

[Frank Bemelman]

We recently did a board that needed eight separate controllable power
sources for strain gage excitation. We used LM1117 ldo regulators and
applied the control voltages to their adj pins (open loop, so the
outputs were 1.25 volts above the control inputs.) That sort of uses
the regulator as a power amp with all the nice characteristics of a
voltage reg chip.

Anyway, I tested the 1117's for damage modes associated with driving
the adj pin from a stiff source and simultaneously abusing the inputs
and outputs various ways. I couldn't find a reasonable way to damage
the regulator, so it looks pretty safe. I didn't test an LM317, but
the 1117 is generally a nicer part anyhow.

Okay, thanks, then it's perhaps not a problem. I just wondered,
because in the usual application, the adj-input is always below
the output, even with a shorted output. I would certainly test
it again if a LM317 were used, in particular for a bench/test
supply.

 

[Frank Bemelman]

This is a fairly basic variation on the above:

Please view in a fixed-width font such as Courier.



R2
Vout=(1 + --) x K x Vref
+-----------+ R3
Vin --+--->| Regulator |----+-------------------+-->
| +-----+-----+ | |
| |Adj | 240 \
| | +--/\/\-+---+ R2
| | | | \
| +------------------+ c /
| \| |
| npn |-----+
| /| |
| e \
| | R3
\ e \
R1 K x Vref |/ /
\ +-----| pnp |
/ | |\ |
| | c-------+
| | |
| | ---
| | Gnd
| Vref ( < 5V ) |
+-------------------------|----------------+
| | |Rh
| | +-----+
| | Radj | | Variable
/---/ zener or +-------------| | Pot
/ \ TL431 | |
--- 2-wire <===============| |
| interface +-----+
| |Rl
--- |
Gnd

Looks nice, but it that Vout formula still correct? I figured
something like this:

R2
Vout=(1 + --) x K x (Vref+1.2)
R3

Be gentle, I'm not drunk, but did have a few.

 

[Costas Vlachos]

Yes, exactly. It may or may not change it. If it doesn't change it, that's
fine. If it does (and it will at some point), then ringing and noise may
appear at the PSU output.


The regulator is analogue and the voltage variations will be very smooth.
Not like the abrupt LSB jumps that the counter and R2R ladder will generate.
Also, the regulator's reaction time is in the order of microseconds (orders
of magnitude faster than a typical ADC --> comparator --> R2R ladder path).
For these reasons I don't think it's a good idea to have the digital loop
constantly monitoring/correcting the PSU output.

Since arriving at the desired output voltage depends on the digital
servo slewing the control voltage until the numerical inputs to the
comparator are equal, there's no way (at least initially) the output
voltage can be correctly set without the digital servo being in the
loop. Once it's been decided that "Yes, the output's really at 12.34V",
(A=B for a while) then it would be possible to put the servo in watchdog
status, while leaving the R2R ladder set and allowing the analog stuff
in the regulator to play by itself.

Yes, of course you need a closed-loop to make sure the output follows the
set point! The problem is that you're using a digital (and hence slow and
quantized) loop for this. If you break the loop once you receive a few A=B
samples and freeze the digital servo until the user changes the set point,
then yes, it would be more acceptable. But having the digital loop always
monitoring/correcting the PSU output may cause the problems I described.
Conceptually it's an attractive idea - I just wanted to point out the
problems in case the OP tries to build it.

cheers,
Costas

 

[Frank Bemelman]

Yes, exactly. It may or may not change it. If it doesn't change it, that's
fine. If it does (and it will at some point), then ringing and noise may
appear at the PSU output.



The regulator is analogue and the voltage variations will be very smooth.
Not like the abrupt LSB jumps that the counter and R2R ladder will generate.
Also, the regulator's reaction time is in the order of microseconds (orders
of magnitude faster than a typical ADC --> comparator --> R2R ladder path).
For these reasons I don't think it's a good idea to have the digital loop
constantly monitoring/correcting the PSU output.



Yes, of course you need a closed-loop to make sure the output follows the
set point! The problem is that you're using a digital (and hence slow and
quantized) loop for this. If you break the loop once you receive a few A=B
samples and freeze the digital servo until the user changes the set point,
then yes, it would be more acceptable. But having the digital loop always
monitoring/correcting the PSU output may cause the problems I described.
Conceptually it's an attractive idea - I just wanted to point out the
problems in case the OP tries to build it.

I agree with that. An R-2R ladder implies 8 bit max, which gives 20mV
steps at 5V, not very pleasant. Analog means infinite steps. But the
idea was nice

 

[Costas Vlachos]

What are you, some kind of drunken retard? That was a programmable
current source.

A constant-current PSU is a current source. I hope you agree with that. Plus
his circuit also provides voltage set point programmability. I suggest you
check the circuit again. The way I see it, his circuit is a constant-voltage
constant-current PSU, although it may not be optimised for constant-voltage
operation (perhaps Win can comment on that). My posting was not intended as
a direct solution to the OP's problem, only a source of ideas on clever uses
of the LM317, esp. in terms of voltage/current set point programmability.

Costas

 

[Winfield Hill]

Costas Vlachos wrote...

... The way I see it, his circuit is a constant-voltage constant-
current PSU, although it may not be optimised for constant-voltage
operation (perhaps Win can comment on that).

.. _______
.. | | Rs 0 to 1.0A current source
.. Vin --+----| LM317 |-+-- 1.25 ohms --+---+-D S--+--+--> Iout
.. RAW | |_______| | | | G | |
.. 33k | | R2 250 '--' Q2 | |
.. | ,---| ---+ | IRF510 | _|_+
.. LM385 | |/ '----|---------------+ etc | --- Cout
.. -1.2 +---| Q3 | | | |
.. \_|_ |\v | Cs 10pF | | gnd
.. /_\ | | ,----||----+-| ----------'
.. | 100 | | | |
.. gnd | ,----| -+-R6 500k -' |
.. Vmax gnd | | | Iout = Vp -------
.. 0 to 30V | _| | Rs R1 R4
.. ---, +--|- \ 1n4148 |
.. +5V | Vp | | >----|<|------+ provided,
.. pot <---| -|+_/ | R2 R3 1.25V
.. | R5 | | ----- = -----
.. gnd 100k gnd | R1 R4 5.0V
.. | |
.. gnd | R6
.. 0 to 5V | Vout(max) = Vp {1 + --)
.. Ip --- R4 97.6k--5k--+- R3 20k-, | R5
.. Iout program pot | __ | |
.. zero '--|- \ | |/
.. | >--| --|
.. +5V -- R4 100k --+--|+_/ | |\V Q1
.. Vcc | | | 2n4401
.. R3 LM358 '-----+
.. 20k LT1013 |
.. | R1 200
.. gnd |
.. gnd

My circuit was optimized as a voltage-programmable current source,
but it does include a voltage-programmability aspect (intended as
a voltage-compliance limit for the current source). The voltage
servo can likely use a little feedback-loop tightening, added as
capacitor Cs, say about 10pF for 33kHz a loop zero. I designed a
much better LM317-family circuit for voltage-programmable supplies
and I'll try to find time to post it in the next day or two.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Frank]

Fred Bloggs wrote:
Eliminating that Vref short on the first diagram- looks like this:

Please view in a fixed-width font such as Courier.

Fred, what is CF for? I admit I'm more of a digital geek and the
analog stuff has **much** to be desired. Pardon any stupid
conclusions, but is the Op Amp operating as an integrator??? It
doesn't make sense to me if it is.

Sorry for all the questions. I'm not asking for you guys to give me
the values for all of these parts, but like I said, to me it's either
on or off...

With that, I'm assuming...

R1 is set to give the Zener enough current to do it's thing based on
Vin. The value would be set based on the Zener's data sheet. Prety
basic.

Rb is set to sink Iadj (acording to the LM317 data sheet, 50-100uA).

CF is for ???

R2/R3 sets the maximum output voltage.

K is a new one. From the schematic, K would be the ratio of the high
and low sides of the pot (i.e, if a 10k pot was at it's midpoint,
K=0.5).

Time to break out the breadboard and parts bin!

....anyone know of any good books on Op Amp design?

Thanks everyone for all the ideas, tips, and comments so far!

-Frank

 

[Frank]

Ditch the OpAmp design book question! I just saw the same thread
several messages up!

 

[Michael]

---
"Tiny" is, say, one LSB which, depending on the width of the ADC and the
comparator, may or may not cause the output(s) of the comparator to
change.
---


---
That's really no different from the up/down "jumps" the regulator has to
make on its own in order to keep the output voltage constant.
---


---
Since arriving at the desired output voltage depends on the digital
servo slewing the control voltage until the numerical inputs to the
comparator are equal, there's no way (at least initially) the output
voltage can be correctly set without the digital servo being in the
loop. Once it's been decided that "Yes, the output's really at 12.34V",
(A=B for a while) then it would be possible to put the servo in watchdog
status, while leaving the R2R ladder set and allowing the analog stuff
in the regulator to play by itself.

If you don't mind loosing 1bit of resolution on the A2D, the jitter
problem from the A2D can be solved completely by ignoring all of the
even, or all of the odd codes.

 

[Frank Bemelman]

If you don't mind loosing 1bit of resolution on the A2D, the jitter
problem from the A2D can be solved completely by ignoring all of the
even, or all of the odd codes.

Nah, that would have to be 'ignoring all of the even if the code
is odd' and 'all of the odd codes if the code is even'. You have
to add the appropriate logic to deal with that

ADC's always 'jitter', no matter how many bits you sacrifice.

 

[Tony Williams]

. _______
. | | Rs 0 to 1.0A current source
. Vin --+----| LM317 |-+-- 1.25 ohms --+---+-D S--+--+--> Iout
. RAW | |_______| | | | G | |
. 33k | | R2 250 '--' Q2 | |
. | ,---| ---+ | IRF510 | _|_+
. LM385 | |/ '----|---------------+ etc | --- Cout
. -1.2 +---| Q3 | | | |
. \_|_ |\v | Cs 10pF | | gnd
. /_\ | | ,----||----+-| ----------'
. | 100 | | | |
. gnd | ,----| -+-R6 500k -' |
. Vmax gnd | | | Iout = Vp -------
. 0 to 30V | _| | Rs R1 R4
. ---, +--|- \ 1n4148 |
. +5V | Vp | | >----|<|------+ provided,
. pot <---| -|+_/ | R2 R3 1.25V
. | R5 | | ----- = -----
. gnd 100k gnd | R1 R4 5.0V
. | |
. gnd | R6
. 0 to 5V | Vout(max) = Vp {1 + --)
. Ip --- R4 97.6k--5k--+- R3 20k-, | R5
. Iout program pot | __ | |
. zero '--|- \ | |/
. | >--| --|
. +5V -- R4 100k --+--|+_/ | |\V Q1
. Vcc | | | 2n4401
. R3 LM358 '-----+
. 20k LT1013 |
. | R1 200
. gnd |
. gnd

A couple of typos have crept in during the editing.

Vp1 R2 R3
Iout = --- x -- x -- where Vp1 is the voltage applied to R4.
Rs R1 R4

What happens to any of these suggested circuits if they
are connected to a large capacitive load, set to full
output, and then suddenly programmed-down?

a) How fast (slow) will the output voltage come down?

b) Does anything get damaged by the reverse current from
the capacitor?

As a long-time user of programmable system power supplies
I prefer to use a programmable supply which has an active
pull-down capability, even if just a transient capability.

 

[Fred Bloggs]

Looks nice, but it that Vout formula still correct? I figured
something like this:

R2
Vout=(1 + --) x K x (Vref+1.2)
R3

Be gentle, I'm not drunk, but did have a few.

Your formula is correct...it is actually a thermometric regulator
requiring A/D feedback- much more than the OP asked for

 

[Costas Vlachos]

Costas Vlachos wrote...
. _______
. | | Rs 0 to 1.0A current source
. Vin --+----| LM317 |-+-- 1.25 ohms --+---+-D S--+--+--> Iout
. RAW | |_______| | | | G | |
. 33k | | R2 250 '--' Q2 | |
. | ,---| ---+ | IRF510 | _|_+
. LM385 | |/ '----|---------------+ etc | --- Cout
. -1.2 +---| Q3 | | | |
. \_|_ |\v | Cs 10pF | | gnd
. /_\ | | ,----||----+-| ----------'
. | 100 | | | |
. gnd | ,----| -+-R6 500k -' |
. Vmax gnd | | | Iout = Vp -------
. 0 to 30V | _| | Rs R1 R4
. ---, +--|- \ 1n4148 |
. +5V | Vp | | >----|<|------+ provided,
. pot <---| -|+_/ | R2 R3 1.25V
. | R5 | | ----- = -----
. gnd 100k gnd | R1 R4 5.0V
. | |
. gnd | R6
. 0 to 5V | Vout(max) = Vp {1 + --)
. Ip --- R4 97.6k--5k--+- R3 20k-, | R5
. Iout program pot | __ | |
. zero '--|- \ | |/
. | >--| --|
. +5V -- R4 100k --+--|+_/ | |\V Q1
. Vcc | | | 2n4401
. R3 LM358 '-----+
. 20k LT1013 |
. | R1 200
. gnd |
. gnd

My circuit was optimized as a voltage-programmable current source,
but it does include a voltage-programmability aspect (intended as
a voltage-compliance limit for the current source). The voltage
servo can likely use a little feedback-loop tightening, added as
capacitor Cs, say about 10pF for 33kHz a loop zero. I designed a
much better LM317-family circuit for voltage-programmable supplies
and I'll try to find time to post it in the next day or two.

Thanks,
- Win

Thanks for that Win, looking forward to seeing the better circuit. Please
post it if you find the time. I'm especially interested in the set point
programmability aspect.

cheers,
Costas

 

[Winfield Hill]

Costas Vlachos wrote...

Thanks for that Win, looking forward to seeing the better circuit.
Please post it if you find the time. I'm especially interested in
the set point programmability aspect.

Argghh! I usually get up early, even as early as 4am, and work
at my computer or dabble a design on a quadrille pad. Yesterday
I brought from work a schematic for a design of mine that we've
been using a few years, to adapt it and make ASCII-drawing posts
as promised. This morning I examined this design and the scene,
and discovered a few issues or shortcomings.

In my design, there's an opamp driving an LM317 or other three-
terminal adjustable chip. Here are the issues:

First, this opamp is powered from the output! of the '317, which
solves the issue of having a high enough supply-voltage rating.
This was fine for my design, and it'll work down to say 0.5V out,
but perhaps not for a power supply that's adjustable down to 0V.
Higher-voltage opamps could be used and connected to the LM317's
input if this voltage is limited to say 35 or 38V.

Second, I used a Burr-Brown rail-to-rail micropower opamp that's
slow and doesn't mind having a 1000pF load on its output. This
is OK, if you don't need to rapidly change the output voltage.

Using another opamp would require careful analysis and testing,
just as I did for the present design.

Third, my design had no current limit, other than that intrinsic
to the '317, and its thermal limiting. This could mean putting
more current than you'd like into a faulty load. I sketched a
solution to this, but it adds quite a few additional parts, and
circuitry to draw. :>( Of course, the LM317 doesn't mind being
mistreated with excessive currents and dissipation, that's why
we're using it instead of a pass transistor in the first place!

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Costas Vlachos]

Costas Vlachos wrote...

Argghh! I usually get up early, even as early as 4am, and work
at my computer or dabble a design on a quadrille pad. Yesterday
I brought from work a schematic for a design of mine that we've
been using a few years, to adapt it and make ASCII-drawing posts
as promised. This morning I examined this design and the scene,
and discovered a few issues or shortcomings.

In my design, there's an opamp driving an LM317 or other three-
terminal adjustable chip. Here are the issues:

First, this opamp is powered from the output! of the '317, which
solves the issue of having a high enough supply-voltage rating.
This was fine for my design, and it'll work down to say 0.5V out,
but perhaps not for a power supply that's adjustable down to 0V.
Higher-voltage opamps could be used and connected to the LM317's
input if this voltage is limited to say 35 or 38V.

Second, I used a Burr-Brown rail-to-rail micropower opamp that's
slow and doesn't mind having a 1000pF load on its output. This
is OK, if you don't need to rapidly change the output voltage.

Using another opamp would require careful analysis and testing,
just as I did for the present design.

Third, my design had no current limit, other than that intrinsic
to the '317, and its thermal limiting. This could mean putting
more current than you'd like into a faulty load. I sketched a
solution to this, but it adds quite a few additional parts, and
circuitry to draw. :>( Of course, the LM317 doesn't mind being
mistreated with excessive currents and dissipation, that's why
we're using it instead of a pass transistor in the first place!

Thanks,
- Win

whill_at_picovolt-dot-com

No need to bother too much about it, I was just curious to see it and learn
from it. I'm sure you have much better things to do than drawing ASCII
circuits for s.e.d.

I did a voltage/current programmable PSU design a while back using discretes
(op-amps, pass elements, and a PIC+DAC for the user interface & set points).
Works very well so far (it even passed the file/rasp test that someone in
s.e.d suggested), but I'm sure there must be room for improvement as I'm
just a hobbyist in electronics - not my main job. The core of the design is
based around this idea:

http://www.national.com/an/LB/LB-28.pdf

I like the fact that there's no large output capacitor, which keeps the
impedance high in constant-current mode and provides some degree of
protection to the load.

cheers,
Costas

 

[John Larkin]

In my design, there's an opamp driving an LM317 or other three-
terminal adjustable chip. Here are the issues:

First, this opamp is powered from the output! of the '317,

Cool. I'll have to remember that one.

solves the issue of having a high enough supply-voltage rating.
This was fine for my design, and it'll work down to say 0.5V out,
but perhaps not for a power supply that's adjustable down to 0V.
Higher-voltage opamps could be used and connected to the LM317's
input if this voltage is limited to say 35 or 38V.

Second, I used a Burr-Brown rail-to-rail micropower opamp that's
slow and doesn't mind having a 1000pF load on its output. This
is OK, if you don't need to rapidly change the output voltage.

Tried the LM8261? R-R i/o, and it tolerates any capacitive load. I'm
using it to drive an LM1117 to make a very accurate +10 volt supply
from VMEbus +12.

Using another opamp would require careful analysis and testing,
just as I did for the present design.

Third, my design had no current limit, other than that intrinsic
to the '317, and its thermal limiting. This could mean putting
more current than you'd like into a faulty load. I sketched a
solution to this, but it adds quite a few additional parts, and
circuitry to draw. :>( Of course, the LM317 doesn't mind being
mistreated with excessive currents and dissipation, that's why
we're using it instead of a pass transistor in the first place!

Vregs make great power amps. Somebody should make an equivalent part
where Vout = Vadj, a zero-voltage regulator.

John

 

[Costas Vlachos]

First, this opamp is powered from the output! of the '317, which
solves the issue of having a high enough supply-voltage rating.
This was fine for my design, and it'll work down to say 0.5V out,
but perhaps not for a power supply that's adjustable down to 0V.
Higher-voltage opamps could be used and connected to the LM317's
input if this voltage is limited to say 35 or 38V.

Interesting setup. About powering from the unregulated input, a similar
situation occurred in my design, where the unregulated voltage can go up to
55V at worst case (no load and maximum mains line voltage). What I did was
to drop it to around 36V with a simple Zener regulator and then power the
op-amps from that (and a -5V negative supply). I'm using the LT1013A/14A
which can take up to 44V I think.

Costas

 

[Winfield Hill]

John Larkin wrote...

Tried the LM8261? R-R i/o, and it tolerates any capacitive load.
I'm using it to drive an LM1117 to make a very accurate +10 volt
supply from VMEbus +12.

That's a nice opamp, with 9V/us slew and 20MHz extrapolated GBW.
But it's only rated for 32V total supply voltage. While this is
more than most small SOT-package RRIO parts, it's less than the
36V rating of the OPA2251 that I'm using. But I'd really like
to find a 40 or 44V part for this application.

Thanks,
- Win

whill_at_picovolt-dot-com