Best way to get multiple voltage references?

T

Thomas P. Gootee

Guest
"Best way to get multiple voltage references?"

This may be obvious and trivial for those of you who are "real" analog
designers. But I don't have much experience with voltage references,
and no one else to ask.

The Situation/Background:

I have a self-designed piece of test equipment (a curve tracer), for
which I've almost completed re-designing the power amplifier's
gain-setting network. The amplifier only needs to have six different
0-to-peak output voltages: 0.5, 3, 6, 9, 12, and 15 volts, which are
controlled by a rotary switch.

The old design used an array of resistors, to set the gain for various
combinations of the controls' settings, to maintain the correct p-p
output voltage level. That was somewhat unsatisfactory, for several
reasons.

I now (will) use an AGC (automatic gain control) feedback control
loop, using three opamps, that derives (from the actual output) a DC
voltage equal to the 0-peak voltage and then subtracts a reference
voltage (DC, equal to the *desired* 0-peak amplitude) from it, to get
an "error signal", to drive the LED in a VACTROL (LED and photocell)
used as a variable resistor, to set the LM1875T amplifier's gain. (The
way the circuit is designed, I actually need negative reference
voltages.)

The Quest:

Everything is great, except that I am still trying to decide what is
the "best" way to derive the reference voltages. They wouldn't need to
be able to supply much current, since they will be tied to an opamp's
inverting input, through a 10K resistor. Needless to say, maybe: I
would like them to be as accurate and precise as possible, within my
cost ("low") and size ("small") constraints.

The unit's power supply produces + and - 15 to 18 volts (those CAN be
varied, although they "normally" wouldn't be), and +5 volts, all
usually with <= 10-20 mV of noise.

The available single-sided PCB space for the voltage references is
about 3/8-inch x 3 inches (about 0.95 cm x 7.62 cm), with other traces
right at each boundary. But I could probably squeeze the rest of the
layout to increase that to at least 1/2-inch x 3 1/2 inches (1.27 cm x
8.9 cm). The negative P.S. rail is available at one of the shorter
sides of that area. And ground is available all along one of the long
sides. Five pads, for wires to go to a rotary switch, also need to fit
in there.

I will probably just use the power supply's negative output voltage as
the reference-input for the highest (15 volts 0-peak) output level.
For the five lower voltages, my *first* thought was to just use a
79L12 TO-92 100mA negative 12-volt regulator and four resistive
voltage dividers running from the regulator's output. That would all
fit easily, on the PCB. But that probably wouldn't give precise-enough
reference voltages. I'm thinking that it would be good to have them as
accurate as possible, assuming a reasonably-low cost, but at LEAST
within 1-3 percent, and better if possible. And the 79L12 looks like
it's only good to within +/-5%, not to mention the resistors for the
dividers.

So, I did some research on "voltage references". There sure do seem to
be a lot of different possible ways to do it! I looked at using zener
diodes, opamp and transistor-based methods (with and without zener
diodes), and also those based on actual "precision voltage reference"
products (which seem to also use zener diodes).

I'm currently thinking that I should use something like four TL431CLP
(TO-92) "Adjustable Precision Voltage Reference" components (0.4%, I
think), for the -12, -9v, -6v, and the -3v levels, with a resistive
voltage-divider to derive the -0.5v from the -3v.

Is that the "best" way for me to do it? And, if so, should I choose
something other than a TL431CLP ($0.27 for qty 1 at www.jameco.com)?
And should I just configure each of them independently, or "stack"
them somehow, or what? Are there any decent (or "good enough")
alternatives to using actual "Voltage Reference" components, in this
case?

Assuming the use of "Voltage Reference"s:

Since I've never used "voltage reference" components before, are there
any "gotchas", or any "common practices", etc, that I might not be
aware of? Any good appnotes out there, for them? (I can't seem to
find very much. Maybe they're too trivial to use, so appnotes aren't
really needed?)

Also, for the -0.5v reference, is a resistive voltage-divider from the
-3v the best way to go? If so, it sure would be nice to get away with
using 1% resistors ($0.02 ea for qty 200, at www.mouser.com), unless
someone knows how I can get higher-precision resistors for less than
$1 or $2 each (although I will certainly pay for the 0.1% resistors,
if I have to, in order to almost match the accuracy and precision of
the other reference voltages).

About the resistive voltage-divider to get the -0.5v ref from the -3v
ref: I read a newsgroup post that suggested using several resistors in
series, for the top of the divider, and several in parallel, for the
bottom of the divider. I think that they also mentioned that that is
how it's done on ICs. But, if so, what is the rationale for that?
(Higher precision? Redundancy? Lower noise? Or what?) I may be a
little short of PCB space, too. But if there's a good-enough reason
for doing it that way, I can try, and can most-likely fit it in.
(What?! Even MORE of those expensive 0.1% resistors?!)

Sorry that this is so lengthy, and for asking so many questions at
once. ANY help, advice, or ideas will be *GREATLY* appreciated.

Regards,

Tom Gootee

tomg AT fullnet.com

http://www.fullnet.com/u/tomg

P.S. There are photos of the curve tracer unit, and photos of the
prototype's screen-displays, etc, at
http://www.fullnet.com/u/tomg/gooteect.htm . (It's pretty neat!)
 
Thomas P. Gootee wrote:
"Best way to get multiple voltage references?"

This may be obvious and trivial for those of you who are "real" analog
designers. But I don't have much experience with voltage references,
and no one else to ask.

The Situation/Background:

I have a self-designed piece of test equipment (a curve tracer), for
which I've almost completed re-designing the power amplifier's
gain-setting network. The amplifier only needs to have six different
0-to-peak output voltages: 0.5, 3, 6, 9, 12, and 15 volts, which are
controlled by a rotary switch.

The old design used an array of resistors, to set the gain for various
combinations of the controls' settings, to maintain the correct p-p
output voltage level. That was somewhat unsatisfactory, for several
reasons.

I now (will) use an AGC (automatic gain control) feedback control
loop, using three opamps, that derives (from the actual output) a DC
voltage equal to the 0-peak voltage and then subtracts a reference
voltage (DC, equal to the *desired* 0-peak amplitude) from it, to get
an "error signal", to drive the LED in a VACTROL (LED and photocell)
used as a variable resistor, to set the LM1875T amplifier's gain. (The
way the circuit is designed, I actually need negative reference
voltages.)

The Quest:

Everything is great, except that I am still trying to decide what is
the "best" way to derive the reference voltages. They wouldn't need to
be able to supply much current, since they will be tied to an opamp's
inverting input, through a 10K resistor. Needless to say, maybe: I
would like them to be as accurate and precise as possible, within my
cost ("low") and size ("small") constraints.
I'd use just one. Then I'd only have to calibrate on this one.
Voltage references are usually measured with ppm/K. By using just one
reference and deriving the others from the one, you know the values
track each other by the offset drift of possible OpAmps.

A LM87Lxx is not a voltage reference, just a regulator.
A regulator is less for precision, but for current.

Rene
--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
 
Hi Thomas,

For a reference I'd look at the LMV431. At less than 20 cents a pop that
is pretty hard to beat in terms of budget. This will need some calibration
when creating a ladder.

For ladders you can buy resistor arrays that carry 4, 8 or 16 identical
resistors that should track pretty well. I believe CMD makes these.

Regards, Joerg
http://www.analogconsultants.com
 
In article <11915d6.0404260804.d5b18b3@posting.google.com>, Thomas P.
Gootee <tomg@fullnet.com> writes
"Best way to get multiple voltage references?"

This may be obvious and trivial for those of you who are "real" analog
designers. But I don't have much experience with voltage references,
and no one else to ask.

The Situation/Background:

I have a self-designed piece of test equipment (a curve tracer), for
which I've almost completed re-designing the power amplifier's
gain-setting network. The amplifier only needs to have six different
0-to-peak output voltages: 0.5, 3, 6, 9, 12, and 15 volts, which are
controlled by a rotary switch.

The old design used an array of resistors, to set the gain for various
combinations of the controls' settings, to maintain the correct p-p
output voltage level. That was somewhat unsatisfactory, for several
reasons.

I now (will) use an AGC (automatic gain control) feedback control
loop, using three opamps, that derives (from the actual output) a DC
voltage equal to the 0-peak voltage and then subtracts a reference
voltage (DC, equal to the *desired* 0-peak amplitude) from it, to get
an "error signal", to drive the LED in a VACTROL (LED and photocell)
used as a variable resistor, to set the LM1875T amplifier's gain. (The
way the circuit is designed, I actually need negative reference
voltages.)

The Quest:

Everything is great, except that I am still trying to decide what is
the "best" way to derive the reference voltages. They wouldn't need to
be able to supply much current, since they will be tied to an opamp's
inverting input, through a 10K resistor. Needless to say, maybe: I
would like them to be as accurate and precise as possible, within my
cost ("low") and size ("small") constraints.

The unit's power supply produces + and - 15 to 18 volts (those CAN be
varied, although they "normally" wouldn't be), and +5 volts, all
usually with <= 10-20 mV of noise.

The available single-sided PCB space for the voltage references is
about 3/8-inch x 3 inches (about 0.95 cm x 7.62 cm), with other traces
right at each boundary. But I could probably squeeze the rest of the
layout to increase that to at least 1/2-inch x 3 1/2 inches (1.27 cm x
8.9 cm). The negative P.S. rail is available at one of the shorter
sides of that area. And ground is available all along one of the long
sides. Five pads, for wires to go to a rotary switch, also need to fit
in there.

I will probably just use the power supply's negative output voltage as
the reference-input for the highest (15 volts 0-peak) output level.
For the five lower voltages, my *first* thought was to just use a
79L12 TO-92 100mA negative 12-volt regulator and four resistive
voltage dividers running from the regulator's output. That would all
fit easily, on the PCB. But that probably wouldn't give precise-enough
reference voltages. I'm thinking that it would be good to have them as
accurate as possible, assuming a reasonably-low cost, but at LEAST
within 1-3 percent, and better if possible. And the 79L12 looks like
it's only good to within +/-5%, not to mention the resistors for the
dividers.

So, I did some research on "voltage references". There sure do seem to
be a lot of different possible ways to do it! I looked at using zener
diodes, opamp and transistor-based methods (with and without zener
diodes), and also those based on actual "precision voltage reference"
products (which seem to also use zener diodes).

I'm currently thinking that I should use something like four TL431CLP
(TO-92) "Adjustable Precision Voltage Reference" components (0.4%, I
think), for the -12, -9v, -6v, and the -3v levels, with a resistive
voltage-divider to derive the -0.5v from the -3v.

Is that the "best" way for me to do it? And, if so, should I choose
something other than a TL431CLP ($0.27 for qty 1 at www.jameco.com)?
And should I just configure each of them independently, or "stack"
them somehow, or what? Are there any decent (or "good enough")
alternatives to using actual "Voltage Reference" components, in this
case?

Assuming the use of "Voltage Reference"s:

Since I've never used "voltage reference" components before, are there
any "gotchas", or any "common practices", etc, that I might not be
aware of? Any good appnotes out there, for them? (I can't seem to
find very much. Maybe they're too trivial to use, so appnotes aren't
really needed?)

Also, for the -0.5v reference, is a resistive voltage-divider from the
-3v the best way to go? If so, it sure would be nice to get away with
using 1% resistors ($0.02 ea for qty 200, at www.mouser.com), unless
someone knows how I can get higher-precision resistors for less than
$1 or $2 each (although I will certainly pay for the 0.1% resistors,
if I have to, in order to almost match the accuracy and precision of
the other reference voltages).

About the resistive voltage-divider to get the -0.5v ref from the -3v
ref: I read a newsgroup post that suggested using several resistors in
series, for the top of the divider, and several in parallel, for the
bottom of the divider. I think that they also mentioned that that is
how it's done on ICs. But, if so, what is the rationale for that?
(Higher precision? Redundancy? Lower noise? Or what?) I may be a
little short of PCB space, too. But if there's a good-enough reason
for doing it that way, I can try, and can most-likely fit it in.
(What?! Even MORE of those expensive 0.1% resistors?!)

Sorry that this is so lengthy, and for asking so many questions at
once. ANY help, advice, or ideas will be *GREATLY* appreciated.

Regards,

Tom Gootee

tomg AT fullnet.com

http://www.fullnet.com/u/tomg

P.S. There are photos of the curve tracer unit, and photos of the
prototype's screen-displays, etc, at
http://www.fullnet.com/u/tomg/gooteect.htm . (It's pretty neat!)
common voltage is 3V
use opamps and 2:1 dividers in feedback loops to generate voltages
independent of resistor tolerances.

--
ddwyer
 
Rene Tschaggelar <none@none.net> wrote in message news:<408d38f0$0$717$5402220f@news.sunrise.ch>...
Thomas P. Gootee wrote:
"Best way to get multiple voltage references?"
<snipped>
I'd use just one. Then I'd only have to calibrate on this one.
Voltage references are usually measured with ppm/K. By using just one
reference and deriving the others from the one, you know the values
track each other by the offset drift of possible OpAmps.

A LM87Lxx is not a voltage reference, just a regulator.
A regulator is less for precision, but for current.

Rene
------------

Rene,

Thanks for the quick response!

OK. That sounds reasonable. So there would be only one reference
voltage. Off the top of my head, it seems like it would then need to
either: A) be scaled with some stable type of opamp circuit with
multiple gain-setting trimpots that would be selected by the existing
rotary "voltage" switch, or: B) be the highest reference voltage
needed and get divided by multiple resistive voltage-dividers, each
with a fixed resistor and a trimpot. Any suggestions about the
details of actually implementing either of those ways? And, are there
alternative ways to derive the other reference voltages from the one?
And does anyone have opinions about what would probably be the *best*
way, in this case?

Thanks!

Regards,

Tom Gootee

------------
 
Thomas P. Gootee wrote:
Rene Tschaggelar <none@none.net> wrote in message news:<408d38f0$0$717$5402220f@news.sunrise.ch>...

Thomas P. Gootee wrote:

"Best way to get multiple voltage references?"


snipped

I'd use just one. Then I'd only have to calibrate on this one.
Voltage references are usually measured with ppm/K. By using just one
reference and deriving the others from the one, you know the values
track each other by the offset drift of possible OpAmps.

A LM87Lxx is not a voltage reference, just a regulator.
A regulator is less for precision, but for current.

Rene


------------

Rene,

Thanks for the quick response!

OK. That sounds reasonable. So there would be only one reference
voltage. Off the top of my head, it seems like it would then need to
either: A) be scaled with some stable type of opamp circuit with
multiple gain-setting trimpots that would be selected by the existing
rotary "voltage" switch, or: B) be the highest reference voltage
needed and get divided by multiple resistive voltage-dividers, each
with a fixed resistor and a trimpot. Any suggestions about the
details of actually implementing either of those ways? And, are there
alternative ways to derive the other reference voltages from the one?
And does anyone have opinions about what would probably be the *best*
way, in this case?
A quad Opamp to the reference gives you another 4 voltages.
References tend to output low voltages, 1.25V, 2.5V, 5V, 7V, 10V.
TempCo's down to 5ppm/K are payable, meaning those more stable become
much more expensive.

You didn't specify well enough. What are the requirements in terms
of current, accuracy, temperature stability, noise, number of units
to be built.

Rene
--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
 
Thomas P. Gootee wrote:
"Best way to get multiple voltage references?"

This may be obvious and trivial for those of you who are "real" analog
designers. But I don't have much experience with voltage references,
and no one else to ask.

The Situation/Background:

I have a self-designed piece of test equipment (a curve tracer), for
which I've almost completed re-designing the power amplifier's
gain-setting network. The amplifier only needs to have six different
0-to-peak output voltages: 0.5, 3, 6, 9, 12, and 15 volts, which are
controlled by a rotary switch.

The old design used an array of resistors, to set the gain for various
combinations of the controls' settings, to maintain the correct p-p
output voltage level. That was somewhat unsatisfactory, for several
reasons.

I now (will) use an AGC (automatic gain control) feedback control
loop, using three opamps, that derives (from the actual output) a DC
voltage equal to the 0-peak voltage and then subtracts a reference
voltage (DC, equal to the *desired* 0-peak amplitude) from it, to get
an "error signal", to drive the LED in a VACTROL (LED and photocell)
used as a variable resistor, to set the LM1875T amplifier's gain. (The
way the circuit is designed, I actually need negative reference
voltages.)

The Quest:

Everything is great, except that I am still trying to decide what is
the "best" way to derive the reference voltages. They wouldn't need to
be able to supply much current, since they will be tied to an opamp's
inverting input, through a 10K resistor. Needless to say, maybe: I
would like them to be as accurate and precise as possible, within my
cost ("low") and size ("small") constraints.

The unit's power supply produces + and - 15 to 18 volts (those CAN be
varied, although they "normally" wouldn't be), and +5 volts, all
usually with <= 10-20 mV of noise.

The available single-sided PCB space for the voltage references is
about 3/8-inch x 3 inches (about 0.95 cm x 7.62 cm), with other traces
right at each boundary. But I could probably squeeze the rest of the
layout to increase that to at least 1/2-inch x 3 1/2 inches (1.27 cm x
8.9 cm). The negative P.S. rail is available at one of the shorter
sides of that area. And ground is available all along one of the long
sides. Five pads, for wires to go to a rotary switch, also need to fit
in there.

I will probably just use the power supply's negative output voltage as
the reference-input for the highest (15 volts 0-peak) output level.
For the five lower voltages, my *first* thought was to just use a
79L12 TO-92 100mA negative 12-volt regulator and four resistive
voltage dividers running from the regulator's output. That would all
fit easily, on the PCB. But that probably wouldn't give precise-enough
reference voltages. I'm thinking that it would be good to have them as
accurate as possible, assuming a reasonably-low cost, but at LEAST
within 1-3 percent, and better if possible. And the 79L12 looks like
it's only good to within +/-5%, not to mention the resistors for the
dividers.

So, I did some research on "voltage references". There sure do seem to
be a lot of different possible ways to do it! I looked at using zener
diodes, opamp and transistor-based methods (with and without zener
diodes), and also those based on actual "precision voltage reference"
products (which seem to also use zener diodes).

I'm currently thinking that I should use something like four TL431CLP
(TO-92) "Adjustable Precision Voltage Reference" components (0.4%, I
think), for the -12, -9v, -6v, and the -3v levels, with a resistive
voltage-divider to derive the -0.5v from the -3v.

Is that the "best" way for me to do it? And, if so, should I choose
something other than a TL431CLP ($0.27 for qty 1 at www.jameco.com)?
And should I just configure each of them independently, or "stack"
them somehow, or what? Are there any decent (or "good enough")
alternatives to using actual "Voltage Reference" components, in this
case?

Assuming the use of "Voltage Reference"s:

Since I've never used "voltage reference" components before, are there
any "gotchas", or any "common practices", etc, that I might not be
aware of? Any good appnotes out there, for them? (I can't seem to
find very much. Maybe they're too trivial to use, so appnotes aren't
really needed?)

Also, for the -0.5v reference, is a resistive voltage-divider from the
-3v the best way to go? If so, it sure would be nice to get away with
using 1% resistors ($0.02 ea for qty 200, at www.mouser.com), unless
someone knows how I can get higher-precision resistors for less than
$1 or $2 each (although I will certainly pay for the 0.1% resistors,
if I have to, in order to almost match the accuracy and precision of
the other reference voltages).

About the resistive voltage-divider to get the -0.5v ref from the -3v
ref: I read a newsgroup post that suggested using several resistors in
series, for the top of the divider, and several in parallel, for the
bottom of the divider. I think that they also mentioned that that is
how it's done on ICs. But, if so, what is the rationale for that?
(Higher precision? Redundancy? Lower noise? Or what?) I may be a
little short of PCB space, too. But if there's a good-enough reason
for doing it that way, I can try, and can most-likely fit it in.
(What?! Even MORE of those expensive 0.1% resistors?!)

Sorry that this is so lengthy, and for asking so many questions at
once. ANY help, advice, or ideas will be *GREATLY* appreciated.

Regards,

Tom Gootee

tomg AT fullnet.com

http://www.fullnet.com/u/tomg

P.S. There are photos of the curve tracer unit, and photos of the
prototype's screen-displays, etc, at
http://www.fullnet.com/u/tomg/gooteect.htm . (It's pretty neat!)
It is not clear from your description how that sweep works- I assume the
end product is a linear power ramp traversing 0->Vpk within an
adjustable sweep time. If you place the power amp inside a feedback loop
with input driven by Vin and a feedback developed from the output of
K*dVout/dt so that in sweep time T the output linearly attains
Vin/K*T=Vpk, then Vin must be adjusted with the T and Vpk settings to be
Vin=K*Vpk/T or =K*Vpk*Fsweep. This makes it clear that if you gang K
with Fsweep in your switches then Vpk is just an independent input to
Vin, i.e. when Fsweep increases by a factor of 10, then the feedback
differentiator gain K is reduced by a factor of 10, so that for a given
Vpk setting the sweep will stay the same 0-Vpk volts-and this amounts to
a switch selecting a voltage divider tap into the amplifier. What kind
of sweep rates are you working with? What is their range from minimum to
maximum frequency?
 
"Rene Tschaggelar" <none@none.net> wrote in message
news:408e374a$0$711$5402220f@news.sunrise.ch...
Thomas P. Gootee wrote:
Rene Tschaggelar <none@none.net> wrote in message
news:<408d38f0$0$717$5402220f@news.sunrise.ch>...


OK. That sounds reasonable. So there would be only one reference
voltage. Off the top of my head, it seems like it would then need
to
either: A) be scaled with some stable type of opamp circuit with
multiple gain-setting trimpots that would be selected by the
existing
rotary "voltage" switch, or: B) be the highest reference voltage
needed and get divided by multiple resistive voltage-dividers,
each
with a fixed resistor and a trimpot. Any suggestions about the
details of actually implementing either of those ways? And, are
there
alternative ways to derive the other reference voltages from the
one?
And does anyone have opinions about what would probably be the
*best*
way, in this case?

A quad Opamp to the reference gives you another 4 voltages.
References tend to output low voltages, 1.25V, 2.5V, 5V, 7V, 10V.
TempCo's down to 5ppm/K are payable, meaning those more stable
become
much more expensive.

You didn't specify well enough. What are the requirements in terms
of current, accuracy, temperature stability, noise, number of units
to be built.

Rene
Perhaps use a single reference and an LTC2620 octal DAC with on-board
buffers (gross overkill, but actually pretty cheap and small).

Regards
Ian
 
Fred Bloggs <nospam@nospam.com> wrote in message news:<408E58A5.7000704@nospam.com>...
Thomas P. Gootee wrote:
"Best way to get multiple voltage references?"
SNIPPED

It is not clear from your description how that sweep works- I assume the
end product is a linear power ramp traversing 0->Vpk within an
adjustable sweep time. If you place the power amp inside a feedback loop
with input driven by Vin and a feedback developed from the output of
K*dVout/dt so that in sweep time T the output linearly attains
Vin/K*T=Vpk, then Vin must be adjusted with the T and Vpk settings to be
Vin=K*Vpk/T or =K*Vpk*Fsweep. This makes it clear that if you gang K
with Fsweep in your switches then Vpk is just an independent input to
Vin, i.e. when Fsweep increases by a factor of 10, then the feedback
differentiator gain K is reduced by a factor of 10, so that for a given
Vpk setting the sweep will stay the same 0-Vpk volts-and this amounts to
a switch selecting a voltage divider tap into the amplifier. What kind
of sweep rates are you working with? What is their range from minimum to
maximum frequency?
The sweeps come from another circuit board, which I don't want to
change, right now. They can be sawtooth ramps from -Vpk to +Vpk,
triangle ramps from -Vpk to +Vpk, and (integrated) quasi-sines from
-Vpk to +Vpk.

The sawtooth ramps can be at frequencies of 60 Hz, 200 Hz, 750 Hz, 4.5
kHz, 12 kHz, and 22 kHz, selectable by a rotary switch. The triangle
and sine, if selected, are always at half the selected frequency.

There is a toggle switch to select either sawtooth or triangle, and
another toggle switch to select whether the integrator is used or not,
which is normally used only when the triangle is selected, to convert
it to a quasi-sine.

Due to the way the signal generator on the other PCB was designed, the
amplitude of the generated sweep varies with both the frequency and
the waveform type, such that the gain of the amplifier needs to be
different for every combination of frequency and waveform-type, in
order for the output's amplitude to match what the "Volts +/-Peak"
switch is set to.

So, RATHER than having multiple-pole rotary switches to select a
gain-setting resistance from among two 6x6 arrays of resistors (One
array for sawtooth. One array for triangle. And one dimension of each
array for the voltage setting and one dimension for the frequency
setting, with two standard-value resistors used to make up each
resistance, i.e. 144 resistors), I am going to use, instead, six DC
reference voltages and a feedback loop around the amplifier that will
use a VACTROL voltage-controlled resistor (LED and photocell) to
change the gain until the output's peak voltage matches the reference
voltage, which will be selected by the "Volts +/-Peak" rotary switch.

My main concern, at this point, is/was to find the "best" way to
generate the DC reference voltages.

And thanks to the helpful, knowledgable people, here, I now have at
least two good ways to do it. It looks like using a Voltage Reference
IC (in a TO-92 package), set for 3V output, with opamps to multiply
the 3V to get the other voltages, will fit nicely into the PCB space
that I have left. (I even finally realized why those little square
cermet trimpots have their pins spaced 0.1" apart, hehe.)

Thanks, everyone!

- Tom
 
"Ian Buckner" <Ian_Buckner@agilent.com> wrote in message news:<1083144078.671706@cswreg.cos.agilent.com>...
"Rene Tschaggelar" <none@none.net> wrote in message
news:408e374a$0$711$5402220f@news.sunrise.ch...
Thomas P. Gootee wrote:
Rene Tschaggelar <none@none.net> wrote in message
news:<408d38f0$0$717$5402220f@news.sunrise.ch>...


OK. That sounds reasonable. So there would be only one reference
voltage. Off the top of my head, it seems like it would then need
to
either: A) be scaled with some stable type of opamp circuit with
multiple gain-setting trimpots that would be selected by the
existing
rotary "voltage" switch, or: B) be the highest reference voltage
needed and get divided by multiple resistive voltage-dividers,
each
with a fixed resistor and a trimpot. Any suggestions about the
details of actually implementing either of those ways? And, are
there
alternative ways to derive the other reference voltages from the
one?
And does anyone have opinions about what would probably be the
*best*
way, in this case?

A quad Opamp to the reference gives you another 4 voltages.
References tend to output low voltages, 1.25V, 2.5V, 5V, 7V, 10V.
TempCo's down to 5ppm/K are payable, meaning those more stable
become
much more expensive.

You didn't specify well enough. What are the requirements in terms
of current, accuracy, temperature stability, noise, number of units
to be built.

Rene

Perhaps use a single reference and an LTC2620 octal DAC with on-board
buffers (gross overkill, but actually pretty cheap and small).

Regards
Ian

Ian,

Good idea! But it looks like the LTC2620 is surface-mount only, and
only goes up to 5.5v. I've not used surface-mount parts, yet. I guess
I'm kind of afraid to try them. (But it sure would be nice to not have
to drill all of those tiny HOLES!)

And I do need outputs up to at least 12 volts. Or am I missing
something, regarding the LTC2620?

But, in general, and noting that my practical knowledge of them is
extremely limited, a DAC IC (digital-to-analog converter) sounds like
it COULD be a very good way to generate multiple DC reference
voltages.

I don't even remember, at the moment, what voltage ranges are
available, in DACs. But...

Let's see. A 12-bit DAC would have 4,096 output levels. IF those were
spread from 0 to 15 volts, the resolution would be about 0.003662
volts.

That's probably better than I need. I see that DACS are available in
at least 8, 10, 12, 14, and 16-bit versions.

A 10-bit DAC would provide 1,024 different output levels. If those
went from 0-15 volts, that would give a per-level (i.e. LSB) change of
about 0.01465 volts. That would work, well-enough, for my case. Even
the 0.5-volt reference would be very close. i.e. 34 x 0.01465 =
0.4981.

An 8-bit DAC would provide 256 levels, which, if spread over 0-15
volts, would give a change of about 0.0586 volts per level. That
might not be quite good enough. But if an 8-bit DAC were run from
0-12V, it would give 0.046875V/step, which is ALMOST good enough for
my smallest (0.5V) reference requirement.

I guess I could also just use a voltage-divider to get the 0.5 from
one of the higher reference voltages. So *maybe* an 8-bit DAC would
work.

I haven't thought about using a DAC, much, until just now. So, how
could I control one? The 8-DAC 16-bit IC mentioned at first seems
like it would need some other (digital) ICs, to work for this case.

Off the top of my head, it seems like it would be very nice if I could
just "hardwire" the binary input bytes needed for the four (or five or
six) reference voltages (3, 6, 9, 12 (and maybe 0.5 and 15)). Maybe I
could use a *single* DAC and use one pole of a rotary switch to switch
its inputs to one of the pre-set input bytes, somehow. It sounds like
that *might* be very easy to do. But nothing comes immediately to
mind. Any good ideas?

- Tom
 
Ian Buckner wrote:

"Rene Tschaggelar" <none@none.net> wrote in message
news:408e374a$0$711$5402220f@news.sunrise.ch...

Thomas P. Gootee wrote:

Rene Tschaggelar <none@none.net> wrote in message

news:<408d38f0$0$717$5402220f@news.sunrise.ch>...


OK. That sounds reasonable. So there would be only one reference
voltage. Off the top of my head, it seems like it would then need

to

either: A) be scaled with some stable type of opamp circuit with
multiple gain-setting trimpots that would be selected by the

existing

rotary "voltage" switch, or: B) be the highest reference voltage
needed and get divided by multiple resistive voltage-dividers,

each

with a fixed resistor and a trimpot. Any suggestions about the
details of actually implementing either of those ways? And, are

there

alternative ways to derive the other reference voltages from the

one?

And does anyone have opinions about what would probably be the

*best*

way, in this case?

A quad Opamp to the reference gives you another 4 voltages.
References tend to output low voltages, 1.25V, 2.5V, 5V, 7V, 10V.
TempCo's down to 5ppm/K are payable, meaning those more stable

become

much more expensive.

You didn't specify well enough. What are the requirements in terms
of current, accuracy, temperature stability, noise, number of units
to be built.

Rene


Perhaps use a single reference and an LTC2620 octal DAC with on-board
buffers (gross overkill, but actually pretty cheap and small).
The LT26xx series is great. I use some of the them in mixed signal
projects where a lot of settable control voltages are required.


Rene
 
Thomas P. Gootee wrote:

"Ian Buckner" <Ian_Buckner@agilent.com> wrote in message news:<1083144078.671706@cswreg.cos.agilent.com>...

Perhaps use a single reference and an LTC2620 octal DAC with on-board
buffers (gross overkill, but actually pretty cheap and small).

Good idea! But it looks like the LTC2620 is surface-mount only, and
only goes up to 5.5v. I've not used surface-mount parts, yet. I guess
I'm kind of afraid to try them. (But it sure would be nice to not have
to drill all of those tiny HOLES!)

And I do need outputs up to at least 12 volts. Or am I missing
something, regarding the LTC2620?

But, in general, and noting that my practical knowledge of them is
extremely limited, a DAC IC (digital-to-analog converter) sounds like
it COULD be a very good way to generate multiple DC reference
voltages.

I don't even remember, at the moment, what voltage ranges are
available, in DACs. But...

Let's see. A 12-bit DAC would have 4,096 output levels. IF those were
spread from 0 to 15 volts, the resolution would be about 0.003662
volts.

That's probably better than I need. I see that DACS are available in
at least 8, 10, 12, 14, and 16-bit versions.

A 10-bit DAC would provide 1,024 different output levels. If those
went from 0-15 volts, that would give a per-level (i.e. LSB) change of
about 0.01465 volts. That would work, well-enough, for my case. Even
the 0.5-volt reference would be very close. i.e. 34 x 0.01465 =
0.4981.

An 8-bit DAC would provide 256 levels, which, if spread over 0-15
volts, would give a change of about 0.0586 volts per level. That
might not be quite good enough. But if an 8-bit DAC were run from
0-12V, it would give 0.046875V/step, which is ALMOST good enough for
my smallest (0.5V) reference requirement.

I guess I could also just use a voltage-divider to get the 0.5 from
one of the higher reference voltages. So *maybe* an 8-bit DAC would
work.

I haven't thought about using a DAC, much, until just now. So, how
could I control one? The 8-DAC 16-bit IC mentioned at first seems
like it would need some other (digital) ICs, to work for this case.

Off the top of my head, it seems like it would be very nice if I could
just "hardwire" the binary input bytes needed for the four (or five or
six) reference voltages (3, 6, 9, 12 (and maybe 0.5 and 15)). Maybe I
could use a *single* DAC and use one pole of a rotary switch to switch
its inputs to one of the pre-set input bytes, somehow. It sounds like
that *might* be very easy to do. But nothing comes immediately to
mind. Any good ideas?

Yes, a microcontroller is required. Plus one OpAmp per channel over 5V,
to extend the range.

Rene
--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
 
Hi Tom,

When using multiple opamps keep an eye on offset errors. I'd probably use one opamp to
obtain the highest voltage, adjust that and then divide down with resistor arrays where
the individual resistors track well enough. Those arrays come in flavors down to 0.1% or
so in relative (not necessarily absolute) tolerance.

Regards, Joerg
http://www.analogconsultants.com
 

Welcome to EDABoard.com

Sponsor

Back
Top