Who here knows a lot about analog signal switching?

[Thomas P. Gootee]

Thanks very much for taking the time to read this.

I am looking for a way to implement a set of signal switches. The
design that I am modifying is a curve tracer that has three
connections to the device under test (DUT). I want to add the
capability to switch alternately between two devices under test.

The AC signals (sawtooth ramps, staircases, et al) that connect to the
DUT can range (according to user-selected switch settings) from just
above 0V p-p up to +/-15V, from about 0 Amps to 1.5 Amps, at
frequencies from 60 Hz to 22 kHz.

The user-selectable current-limiting resistance for the sweep-signal
path can be as low as 10 Ohms. So the switches' "on" resistances
should be "much smaller than" 10 Ohms.

Power supplies available are: two symmetric + and - 17.5V DC (could be
easily adjusted upward) and one fixed +5V DC. (Others "could" be
added.)

The switching rate(s): A "slow rate" would alternate between the two
DUTs on the order of once a second. But, if it's possible, it might be
desirable to also be able to user-select a "fast" switching rate, so
the devices' x-y displays would appear to be on the scope's screen
simultaneously. (And the user will also be able to select either DUT
by itself, i.e. without any automatic switching between them.)

So I'm wondering what the "best" implementation would be, for the
switches. I've looked at mechanical relays (e.g. reed type), "analog
switch" and multiplexer ICs, "solid state relays", and discrete MOSFET
or FET "transmission gate"-type analog AC switching circuits. Am I
just missing something else that's obvious? (I'm probably not familiar
with all of the device types that are out there.)

Relays would be cheap and easy. But I'm worried about their lifespan.
And the fast-rate switching option might be impossible, or at least "a
bad idea", with relays.

Analog switch types of ICs *sounded* good. But I haven't seen any
that have a low-enough "ON" resistance. Since the signal path's
resistance is user-selectable down to 10 Ohms, I'm thinking that the
switches' "ON" resistance should be no more than about 0.1 Ohms, with
less being better (and way less being desirable). And the analog
switch ICs that I've seen couldn't even begin to handle the currents
involved, anyway.

Solid sate relays look pretty good. But they seem pretty pricey
compared to just doing the same thing with discrete MOSFETs (or
FETs?).

So, I've been leaning toward trying to design some "transmission gate"
or bidirectional types of MOSFET or FET switches. I wasn't familiar
with them, and am actually not extremely familiar with MOSFET or FET
circuit design or analysis, although I've been learning a little more,
lately. ("Disclaimer": My BSEE degree is old (1978). And it had an
emphasis on automatic control instead of circuits. And I did nothing
but software from 1984 until 1998, when I started dabbling in circuits
again.)

Trying to get up to speed: I've seen various topologies for FET-type
switches, and am still slightly confused, maybe because many of them
seem to be for switching DC power instead of AC signals, and many were
for on-chip CMOS designs.

Would two same-channel devices that are basically back-to-back in
series work as a bidirectional analog AC signal switch, as I think
I've seen claimed? Or would we really need to use one N-channel and
one P-channel in parallel? (or was it "anti-parallel"?) And, if it's
the latter, how well would the device characteristics need to match?
Would it just be a matter of finding similarly-rated N and P devices
that had nearly equal RDSon values, and probably similar input
capacitances? (Not that I've actually FOUND any pairs like that,
yet...) And is it too much to hope for, that there might be available
some dual devices, with already-matched P an N channel devices in one
package? Or can anyone suggest some suitable N and P pairs that are
actually available (probably will buy from somewhere like
www.mouser.com)?

On the other hand, maybe I'm making this much more complicated than it
actually is. If there's a ready-made device that would be suitable, or
a standard way of doing this type of switching (it seems like there
must be), I'm all ears.

Thanks again, for reading this, and for any advice or ideas that you
can share.

Regards,

Tom Gootee

[email protected]

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

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

 

[John Woodgate]

I read in sci.electronics.design that Thomas P. Gootee
..com>) about 'Who here knows a lot about analog signal switching?', on
Fri, 16 Jul 2004:

I am looking for a way to implement a set of signal switches. The
design that I am modifying is a curve tracer that has three connections
to the device under test (DUT). I want to add the capability to switch
alternately between two devices under test.

Having read your whole article, I recommend reed switches. While you may
have a speed problem with the high-speed switching mode (depending on
some things you don't tell us), the difficulties for you with the other
techniques indicate a very long development time, if not ultimate
disappointment.

 

[Ben Bradley]

Thanks very much for taking the time to read this.

I am looking for a way to implement a set of signal switches. The
design that I am modifying is a curve tracer that has three
connections to the device under test (DUT). I want to add the
capability to switch alternately between two devices under test.

The AC signals (sawtooth ramps, staircases, et al) that connect to the
DUT can range (according to user-selected switch settings) from just
above 0V p-p up to +/-15V, from about 0 Amps to 1.5 Amps, at
frequencies from 60 Hz to 22 kHz.

The user-selectable current-limiting resistance for the sweep-signal
path can be as low as 10 Ohms. So the switches' "on" resistances
should be "much smaller than" 10 Ohms.

Power supplies available are: two symmetric + and - 17.5V DC (could be
easily adjusted upward) and one fixed +5V DC. (Others "could" be
added.)

The switching rate(s): A "slow rate" would alternate between the two
DUTs on the order of once a second. But, if it's possible, it might be
desirable to also be able to user-select a "fast" switching rate, so
the devices' x-y displays would appear to be on the scope's screen
simultaneously. (And the user will also be able to select either DUT
by itself, i.e. without any automatic switching between them.)

So I'm wondering what the "best" implementation would be, for the
switches. I've looked at mechanical relays (e.g. reed type), "analog
switch" and multiplexer ICs, "solid state relays", and discrete MOSFET
or FET "transmission gate"-type analog AC switching circuits. Am I
just missing something else that's obvious? (I'm probably not familiar
with all of the device types that are out there.)

Relays would be cheap and easy. But I'm worried about their lifespan.
And the fast-rate switching option might be impossible, or at least "a
bad idea", with relays.

Analog switch types of ICs *sounded* good. But I haven't seen any
that have a low-enough "ON" resistance. Since the signal path's
resistance is user-selectable down to 10 Ohms, I'm thinking that the
switches' "ON" resistance should be no more than about 0.1 Ohms, with
less being better (and way less being desirable). And the analog
switch ICs that I've seen couldn't even begin to handle the currents
involved, anyway.

Solid sate relays look pretty good. But they seem pretty pricey
compared to just doing the same thing with discrete MOSFETs (or
FETs?).

So, I've been leaning toward trying to design some "transmission gate"
or bidirectional types of MOSFET or FET switches. I wasn't familiar
with them, and am actually not extremely familiar with MOSFET or FET
circuit design or analysis, although I've been learning a little more,
lately. ("Disclaimer": My BSEE degree is old (1978). And it had an
emphasis on automatic control instead of circuits. And I did nothing
but software from 1984 until 1998, when I started dabbling in circuits
again.)

Trying to get up to speed: I've seen various topologies for FET-type
switches, and am still slightly confused, maybe because many of them
seem to be for switching DC power instead of AC signals, and many were
for on-chip CMOS designs.

I'm envisioning something like Jim Thomson posted on
alt.binaries.schematics.electronic on June 22 with the subject: "Re:
Design challenge - LoadLimiter.pdf". It may have rolled off your news
server (and Google Groups doesn't archive binary groups), but I can
repost it there or email it to you if you want.
You would use a beefed-up version of that, with several-amp power
rectifiers, and 1.5 amp current source and sink. To turn the 'switch'
off, just turn off the current source and sink (simultaneously). One
gotcha is the large amount of power dissipated by the current source
and sink circuits for 1.5 amps (or whatever the peak current into or
out of this device is) at +/-15 volts (or whatever the max voltage
this thing uses). This needs power transistors on decent-sized heat
sinks.

 

[Joerg]

Hi Thomas,

If you want to do this with semiconductors you probably need FETs that
are controlled with a voltage swing that is several volts larger than
your range. But you'd have to make sure you clamp the gate-source
voltage to whatever the spec says it should not exceed. Many common FETs
are limited to +/- 20V.

Reed relays are nice but often not rated as high as 1.5A.

Regards, Joerg

 

[John Fields]

Thanks very much for taking the time to read this.

I am looking for a way to implement a set of signal switches. The
design that I am modifying is a curve tracer that has three
connections to the device under test (DUT). I want to add the
capability to switch alternately between two devices under test.

The AC signals (sawtooth ramps, staircases, et al) that connect to the
DUT can range (according to user-selected switch settings) from just
above 0V p-p up to +/-15V, from about 0 Amps to 1.5 Amps, at
frequencies from 60 Hz to 22 kHz.

The user-selectable current-limiting resistance for the sweep-signal
path can be as low as 10 Ohms. So the switches' "on" resistances
should be "much smaller than" 10 Ohms.

Power supplies available are: two symmetric + and - 17.5V DC (could be
easily adjusted upward) and one fixed +5V DC. (Others "could" be
added.)

The switching rate(s): A "slow rate" would alternate between the two
DUTs on the order of once a second. But, if it's possible, it might be
desirable to also be able to user-select a "fast" switching rate, so
the devices' x-y displays would appear to be on the scope's screen
simultaneously. (And the user will also be able to select either DUT
by itself, i.e. without any automatic switching between them.)

So I'm wondering what the "best" implementation would be, for the
switches. I've looked at mechanical relays (e.g. reed type), "analog
switch" and multiplexer ICs, "solid state relays", and discrete MOSFET
or FET "transmission gate"-type analog AC switching circuits. Am I
just missing something else that's obvious? (I'm probably not familiar
with all of the device types that are out there.)

Relays would be cheap and easy. But I'm worried about their lifespan.
And the fast-rate switching option might be impossible, or at least "a
bad idea", with relays.

Analog switch types of ICs *sounded* good. But I haven't seen any
that have a low-enough "ON" resistance. Since the signal path's
resistance is user-selectable down to 10 Ohms, I'm thinking that the
switches' "ON" resistance should be no more than about 0.1 Ohms, with
less being better (and way less being desirable). And the analog
switch ICs that I've seen couldn't even begin to handle the currents
involved, anyway.

Solid sate relays look pretty good. But they seem pretty pricey
compared to just doing the same thing with discrete MOSFETs (or
FETs?).

So, I've been leaning toward trying to design some "transmission gate"
or bidirectional types of MOSFET or FET switches. I wasn't familiar
with them, and am actually not extremely familiar with MOSFET or FET
circuit design or analysis, although I've been learning a little more,
lately. ("Disclaimer": My BSEE degree is old (1978). And it had an
emphasis on automatic control instead of circuits. And I did nothing
but software from 1984 until 1998, when I started dabbling in circuits
again.)

Trying to get up to speed: I've seen various topologies for FET-type
switches, and am still slightly confused, maybe because many of them
seem to be for switching DC power instead of AC signals, and many were
for on-chip CMOS designs.

Would two same-channel devices that are basically back-to-back in
series work as a bidirectional analog AC signal switch, as I think
I've seen claimed? Or would we really need to use one N-channel and
one P-channel in parallel? (or was it "anti-parallel"?) And, if it's
the latter, how well would the device characteristics need to match?
Would it just be a matter of finding similarly-rated N and P devices
that had nearly equal RDSon values, and probably similar input
capacitances? (Not that I've actually FOUND any pairs like that,
yet...) And is it too much to hope for, that there might be available
some dual devices, with already-matched P an N channel devices in one
package? Or can anyone suggest some suitable N and P pairs that are
actually available (probably will buy from somewhere like
www.mouser.com)?

On the other hand, maybe I'm making this much more complicated than it
actually is. If there's a ready-made device that would be suitable, or
a standard way of doing this type of switching (it seems like there
must be), I'm all ears.

 

[colin]

Thanks very much for taking the time to read this.

I am looking for a way to implement a set of signal switches. The
design that I am modifying is a curve tracer that has three
connections to the device under test (DUT). I want to add the
capability to switch alternately between two devices under test.

The AC signals (sawtooth ramps, staircases, et al) that connect to the
DUT can range (according to user-selected switch settings) from just
above 0V p-p up to +/-15V, from about 0 Amps to 1.5 Amps, at
frequencies from 60 Hz to 22 kHz.

The user-selectable current-limiting resistance for the sweep-signal
path can be as low as 10 Ohms. So the switches' "on" resistances
should be "much smaller than" 10 Ohms.

Power supplies available are: two symmetric + and - 17.5V DC (could be
easily adjusted upward) and one fixed +5V DC. (Others "could" be
added.)

The switching rate(s): A "slow rate" would alternate between the two
DUTs on the order of once a second. But, if it's possible, it might be
desirable to also be able to user-select a "fast" switching rate, so
the devices' x-y displays would appear to be on the scope's screen
simultaneously. (And the user will also be able to select either DUT
by itself, i.e. without any automatic switching between them.)

So I'm wondering what the "best" implementation would be, for the
switches. I've looked at mechanical relays (e.g. reed type), "analog
switch" and multiplexer ICs, "solid state relays", and discrete MOSFET
or FET "transmission gate"-type analog AC switching circuits. Am I
just missing something else that's obvious? (I'm probably not familiar
with all of the device types that are out there.)

Relays would be cheap and easy. But I'm worried about their lifespan.
And the fast-rate switching option might be impossible, or at least "a
bad idea", with relays.

Analog switch types of ICs *sounded* good. But I haven't seen any
that have a low-enough "ON" resistance. Since the signal path's
resistance is user-selectable down to 10 Ohms, I'm thinking that the
switches' "ON" resistance should be no more than about 0.1 Ohms, with
less being better (and way less being desirable). And the analog
switch ICs that I've seen couldn't even begin to handle the currents
involved, anyway.

Solid sate relays look pretty good. But they seem pretty pricey
compared to just doing the same thing with discrete MOSFETs (or
FETs?).

So, I've been leaning toward trying to design some "transmission gate"
or bidirectional types of MOSFET or FET switches. I wasn't familiar
with them, and am actually not extremely familiar with MOSFET or FET
circuit design or analysis, although I've been learning a little more,
lately. ("Disclaimer": My BSEE degree is old (1978). And it had an
emphasis on automatic control instead of circuits. And I did nothing
but software from 1984 until 1998, when I started dabbling in circuits
again.)

Trying to get up to speed: I've seen various topologies for FET-type
switches, and am still slightly confused, maybe because many of them
seem to be for switching DC power instead of AC signals, and many were
for on-chip CMOS designs.

Would two same-channel devices that are basically back-to-back in
series work as a bidirectional analog AC signal switch, as I think
I've seen claimed? Or would we really need to use one N-channel and
one P-channel in parallel? (or was it "anti-parallel"?) And, if it's
the latter, how well would the device characteristics need to match?
Would it just be a matter of finding similarly-rated N and P devices
that had nearly equal RDSon values, and probably similar input
capacitances? (Not that I've actually FOUND any pairs like that,
yet...) And is it too much to hope for, that there might be available
some dual devices, with already-matched P an N channel devices in one
package? Or can anyone suggest some suitable N and P pairs that are
actually available (probably will buy from somewhere like
www.mouser.com)?

On the other hand, maybe I'm making this much more complicated than it
actually is. If there's a ready-made device that would be suitable, or
a standard way of doing this type of switching (it seems like there
must be), I'm all ears.

Thanks again, for reading this, and for any advice or ideas that you
can share.

Regards,

Tom Gootee

[email protected]

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

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

main problem with using discrete paralell n and p chanel mosfets is they
always seem to have antiparalel diode wich screws it up, puting 2 n chanel
in series wld be ok, n chanel devices are generaly beter choice anyway. u
have to consider leakege capacitance though, if its too much u can have 2
switches in series and a third wich shorts the junction between them to
ground. (that makes 5 mosfets in total)

reeds wold be easier of course if they were fast enough, theyr pretty
reliabe.

Colin =^.^=

 

[[email protected]]

Check out the solild state relays from Clare. They're small, up to a
few amps, TTL drive, cheap.

 

[Tony Williams]

What I'd be seriously tempted to do would be to consider your
driver(s) as signal sources and use them to drive parallel power
op-amps driving your DUTs.

I've just been simulating some LT1010 ciruits in
LTSpice. Using that device might be the way to
go there.

 

[Thomas P. Gootee]

I read in sci.electronics.design that Thomas P. Gootee
.com>) about 'Who here knows a lot about analog signal switching?', on
Fri, 16 Jul 2004:

Having read your whole article, I recommend reed switches. While you may
have a speed problem with the high-speed switching mode (depending on
some things you don't tell us), the difficulties for you with the other
techniques indicate a very long development time, if not ultimate
disappointment.

John,

Thanks, very much. That is one of the types of advice that I was
looking for, since I DO have to worry about balancing the benefits and
pleasures of learning new things, even if the solution could be
somewhat better, against the need to push this particular project to
completion within a reasonable time.

Thanks again.

Regards,

Tom Gootee

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

 

[Thomas P. Gootee]

I'm envisioning something like Jim Thomson posted on
alt.binaries.schematics.electronic on June 22 with the subject: "Re:
Design challenge - LoadLimiter.pdf". It may have rolled off your news
server (and Google Groups doesn't archive binary groups), but I can
repost it there or email it to you if you want.
You would use a beefed-up version of that, with several-amp power
rectifiers, and 1.5 amp current source and sink. To turn the 'switch'
off, just turn off the current source and sink (simultaneously). One
gotcha is the large amount of power dissipated by the current source
and sink circuits for 1.5 amps (or whatever the peak current into or
out of this device is) at +/-15 volts (or whatever the max voltage
this thing uses). This needs power transistors on decent-sized heat
sinks.

Ben,

Thanks for the response.

It sounds interesting. And I can't find it. If you would email it to
me, I would appreciate it!

Thanks again!

Regards,

Tom Gootee

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

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

 

[Thomas P. Gootee]

main problem with using discrete paralell n and p chanel mosfets is they
always seem to have antiparalel diode wich screws it up, puting 2 n chanel
in series wld be ok, n chanel devices are generaly beter choice anyway. u
have to consider leakege capacitance though, if its too much u can have 2
switches in series and a third wich shorts the junction between them to
ground. (that makes 5 mosfets in total)

reeds wold be easier of course if they were fast enough, theyr pretty
reliabe.

Colin =^.^=

Colin,

Thanks! Good info. I'm leaning toward using reed relays, I think.

- Tom

 

[Thomas P. Gootee]

Check out the solild state relays from Clare. They're small, up to a
few amps, TTL drive, cheap.

Thanks much! I'll take a look at them.

Regards,

Tom Gootee

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

 

[Thomas P. Gootee]

John,

Thanks for the reply!

Yes, I had thought of that. I had originally just casually dismissed
the idea because it would require a duplicate of my power amplifier
circuit board (which basically IS just a power opamp circuit, along
with output-level selection and precision control circuitry), which (I
tend to think) would cost a lot more than just "switches". (But maybe
it wouldn't, since the power amp board's design work is already DONE,
and the "switches" are becoming problematical for me.) It seems like
it would also require a way to have precise matching of the two power
amplifiers' output levels.

But it DOES seem like it would be a lot easier and better to just
switch the two amps' low-level INPUTS on and off, instead of trying to
switch one amp's higher-power output between two DUTs.

The output-levels-matching trick might have been difficult, with my
OLDER power amplifier design. But the new one uses a feedback control
loop that compares a DC level that's derived from its *actual* output
amplitude to a (user-selected) DC reference voltage that's derived
from a fairly-high-precision 2.5V voltage reference (accurate to <=
0.4%), that corresponds to the user's selected/desired output
amplitude, and uses their arith-metic *difference* (i.e. feedback
loop's "error signal") to drive a VTL5C2 Vactrol (analog optical
isolator; essentially an LED encapsulated with a photocell; i.e. a
current-controlled resistor with none of the "quirks" of using an FET
as one), which continuously adjusts the amplifier's gain to always try
to drive the error-signal to zero.

The VTL5C2 Vactrol seems like a pretty handy device: Its resistance
goes from something like 1 or 2 MegOhms, with 0 mA through its LED, to
200 Ohms, with 40 mA. (Of course, compared to a solid state device,
the vactrol is very, VERY slow to respond. But that's OK, for this
application.) And they're only $0.50 each, at www.bgmicro.com. (A
person could probably very-easily make their own equivalent, too.)

My power amplifier's gain can vary from 1 to at least 101. And it is
quite accurate at producing the desired output amplitude, for voltages
from +/-0.5v to +/- 15v, while amplifying signals that have a
frequency of from 30 Hz to at least 22 kHz (uses a
frequency-compensated peak-detector/rectifier circuit, etc etc), while
the current it's pushing varies anywhere from 0 to about 3 or 4 amps
(which is limited to 1.5 amps max, in my curve tracer product).

At any rate, with the feedback output-amplitude control system that's
in place on the power amp board, it might be fairly straightforward to
modify it slightly so that one power amplifier board could be
designated as the "master" and another one the "slave", with the
slave's feedback control loop comparing its output amplitude to the DC
references from the *master*, instead of to its own on-board DC
references, which might be slightly different.

I'd still like to NOT have to duplicate the power amp board. So I
might have to just settle for having only the slow-rate switching,
using something like reed relays, if I can't do it with mosfets or
fets, and if solid state relays are too expensive (They may really NOT
be "too expensive", though, I'm guessing/hoping, compared to
duplicating the power amp board, etc., at least in terms of production
costs alone. But I will have to see.).

Thanks again! Sorry to have blathered-on for so long about all of
that!

Regards,

Tom Gootee

 

[John Fields]

John,

Thanks for the reply!

Yes, I had thought of that. I had originally just casually dismissed
the idea because it would require a duplicate of my power amplifier
circuit board (which basically IS just a power opamp circuit, along
with output-level selection and precision control circuitry), which (I
tend to think) would cost a lot more than just "switches". (But maybe
it wouldn't, since the power amp board's design work is already DONE,
and the "switches" are becoming problematical for me.) It seems like
it would also require a way to have precise matching of the two power
amplifiers' output levels.

But it DOES seem like it would be a lot easier and better to just
switch the two amps' low-level INPUTS on and off, instead of trying to
switch one amp's higher-power output between two DUTs.

The output-levels-matching trick might have been difficult, with my
OLDER power amplifier design. But the new one uses a feedback control
loop that compares a DC level that's derived from its *actual* output
amplitude to a (user-selected) DC reference voltage that's derived
from a fairly-high-precision 2.5V voltage reference (accurate to <=
0.4%), that corresponds to the user's selected/desired output
amplitude, and uses their arith-metic *difference* (i.e. feedback
loop's "error signal") to drive a VTL5C2 Vactrol (analog optical
isolator; essentially an LED encapsulated with a photocell; i.e. a
current-controlled resistor with none of the "quirks" of using an FET
as one), which continuously adjusts the amplifier's gain to always try
to drive the error-signal to zero.

The VTL5C2 Vactrol seems like a pretty handy device: Its resistance
goes from something like 1 or 2 MegOhms, with 0 mA through its LED, to
200 Ohms, with 40 mA. (Of course, compared to a solid state device,
the vactrol is very, VERY slow to respond. But that's OK, for this
application.) And they're only $0.50 each, at www.bgmicro.com. (A
person could probably very-easily make their own equivalent, too.)

My power amplifier's gain can vary from 1 to at least 101. And it is
quite accurate at producing the desired output amplitude, for voltages
from +/-0.5v to +/- 15v, while amplifying signals that have a
frequency of from 30 Hz to at least 22 kHz (uses a
frequency-compensated peak-detector/rectifier circuit, etc etc), while
the current it's pushing varies anywhere from 0 to about 3 or 4 amps
(which is limited to 1.5 amps max, in my curve tracer product).

At any rate, with the feedback output-amplitude control system that's
in place on the power amp board, it might be fairly straightforward to
modify it slightly so that one power amplifier board could be
designated as the "master" and another one the "slave", with the
slave's feedback control loop comparing its output amplitude to the DC
references from the *master*, instead of to its own on-board DC
references, which might be slightly different.

I'd still like to NOT have to duplicate the power amp board. So I
might have to just settle for having only the slow-rate switching,
using something like reed relays, if I can't do it with mosfets or
fets, and if solid state relays are too expensive (They may really NOT
be "too expensive", though, I'm guessing/hoping, compared to
duplicating the power amp board, etc., at least in terms of production
costs alone. But I will have to see.).

---
Interesting description, but I'm kind of at a loss here because I
don't know exactly what you're trying to do.

I had thought that you wanted to exercise two DUTs with the same
forcing function and then plot their responses on a Cartesian display
in ±y as some function of ±x, x being common to them both and also
driving the display in ±x.

 

[Ben Bradley]

Ben,

Thanks for the response.

It sounds interesting. And I can't find it. If you would email it to
me, I would appreciate it!

Emailed with the thread text from a.b.s.e included.

 

[Thomas P. Gootee]

---
Interesting description, but I'm kind of at a loss here because I
don't know exactly what you're trying to do.

I had thought that you wanted to exercise two DUTs with the same
forcing function and then plot their responses on a Cartesian display
in ±y as some function of ±x, x being common to them both and also
driving the display in ±x.

John,

You are correct. I am either not explaining it well-enough or I have
misunderstood something.

Briefly:

I have designed and built a simple curve tracer, with three
connections to the DUT.

It has a main circuit board that generates a sweep signal for the DUT.
The generated sweep signal is "low-level", i.e. just the output of a
small opamp (OP275).

The main board also generates a synchronized staircase waveform, such
that at the end of each sweep cycle (usually a symmetrically-bipolar
sawtooth ramp; but can be user-switched to triangle or sine) it
triggers a transition to the next stairstep level. [The staircase
generator is made with a 4-bit "0-to-15" binary "up"-counter IC that
is clocked with an end-of-sweep pulse. Its four output lines are fed
through a simple R-2R resistive-ladder digital-to-analog converter and
into an opamp, which outputs the staircase. (And the counter
automatically resets to 0000 if a pulse is applied when it's at
1111.)]

The main board also contains everything else that is needed for it to
be a very basic standalone curve tracer, except for the power supplies
and the power amplification needed if you want to beef up the current
and voltage that are used to exercise the DUT. i.e. It also has the
current-limiting resistors (10 Ohms to 5 MegOhms in 12 steps). And it
has the voltage and current sensing and display circuitry (with x and
y outputs to go to an oscilloscope), which mainly consists of some
instrumentation amps, inverting amps, current-sensing resistors, and
switching, such that I can use switches to display any of the three
DUT currents versus the voltage between any two DUT leads, with the
ability to flip the displayed polarity on both the x and y axes.

To be able to push more current through the DUT (and to add a p-p
voltage selector), I designed an amplifier for the sweep signal. It
is on a separate circuit board, which has its input and output wired
in series with a point just upstream of the main board's
current-limiting resistors and the DUT "+" output. It uses an LM1875T
amplifier IC circuit. It also has a precision voltage reference
(TL1431IZ) and some opamps to create the DC reference voltages
corresponding to the user-selectable p-p output-voltage amplitudes,
and the previously-described feedback control loop that varies the
amplifier's gain to maintain the desired output-voltage amplitude. Due
to the +/-15 v p-p maximum and the 10 Ohms minimum, the maximum
current should be +/-1.5 Amps.

(I also put together a switchmode power supply that provides variable
+/-V vdc and fixed +5 vdc, where V is normally set to 17.5 or a little
more, so everything can swing to +/-15.)

**NOW**, I want to be able to run *two* DUTs, "simultaneously", so
their displays can be more-easily compared. The plan was to switch the
three DUT connections back and forth between the two DUTs, and, if
possible, at a variable rate, or with more than one selectable rate.

(Disregarding the staircase base/gate-drive signal As I understand
the discussion so far: I could either put the sweep's switching (to
select between the two DUTs) just before the DUTs i.e. AFTER the sweep
amplifier (probably using reed relays?), OR, use two identical sweep
amplifiers and switch the low-level sweep signal to their inputs,
alternately, which is what I thought you were suggesting.

If I can, I would like to not have to use two sweep amplifiers. I
would like to be able to just add another set of three DUT terminals
and have some type of switches that would automatically alternate
between each pair of the two sets of three.

Tom Gootee

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

 

[sycochkn]

They do make AC relays with fets instead of contacts and they would probably
work very well for you.

Bob

 

[Thomas P. Gootee]

Bob,

Thanks for the response!

I see that www.mouser.com has an NEC solid-state MOSFET relay
(PS710A-1A) with an RDSon of only 0.2 Ohms, in a DIP-6 package, that
will handle 1800 mA at 60v, for $4.52 for qty 1. THAT might WORK!

And its Ton and Toff are rated at 3 ms and 1 ms, which seems like it
"should" allow for switching at up to something like, say, 50 times
per second. Let's see: 50 cycles per second would give 20 msec per
cycle, displaying two devices, i.e. each device for half(??) of each
cycle, or 10 msec, with 4 msec eaten up by turn-on and turn-off
(unless I could overlap them a little bit...), giving 6 msec "on" time
for each DUT. I wonder if the display would flicker too much, or be
too dim. I guess I could also try it at, say, 33 cycles per second,
which would extend each cycle to 30 ms, giving each DUT 11 msec
on-time out of 15 msec per half cycle. Or, if I used a FULL cycle for
each device's "on" time, it might be much better, even at 50 Hz.

I'll probably just have to wait and see what the display looks like at
different rates and see what looks acceptable. It seems like it
probably won't be very good at all for the lower sweep frequencies,
since it wouldn't even allow enough time for one entire cycle of the
base or gate staircase sequence, although it would probably pick up
the entire sequence over multiple cycles, possibly even rapidly enough
to produce a good display. Or, maybe I could control the timing so
that the switching is synchronized with the staircase cycles. It'll
be fun to see what works, at any rate (excuse the pun).

I'm also kind of wondering how much of a problem the transients might
be.

I see that www.mouser.com also has some Axicom miniature DPDT PCB
relays; three models with "expected mechanical life" ratings of 15
million operations, 100 million operations, and 1 billion operations,
for $1.47, $1.87, and $2.23 for qty 1. The 1 billion ops model is
only rated for 1 Amp. The 100 million and 15 million ops models are
rated at 5 amps and 3 amps.

MAYBE the 100-million-operations-lifetime model would be good enough,
if I ran it at only 1 or 2 ops/second. If I'm calculating correctly,
that should be an expected life of something between 25,000 and 30,000
hours, if cycling at 1 op/second, which is over 1,000 24-hour days,
which is 3,000 8-hour days, giving an expected life that probably over
10 years of continuous switching for 8 hours a day, 5 days a week.

Thanks again, Bob!

Regards,

Tom Gootee

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

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

 

[Thomas P. Gootee]

Bob,

Thanks for the response!

I see that www.mouser.com has an NEC solid-state MOSFET relay
(PS710A-1A) with an RDSon of only 0.2 Ohms, in a DIP-6 package, that
will handle 1800 mA at 60v, for $4.52 for qty 1. THAT might WORK!

And its Ton and Toff are rated at 3 ms and 1 ms, which seems like it
"should" allow for switching at up to something like, say, 50 times
per second. Let's see: 50 cycles per second would give 20 msec per
cycle, displaying two devices, i.e. each device for half(??) of each
cycle, or 10 msec, with 4 msec eaten up by turn-on and turn-off
(unless I could overlap them a little bit...), giving 6 msec "on" time
for each DUT. I wonder if the display would flicker too much, or be
too dim. I guess I could also try it at, say, 33 cycles per second,
which would extend each cycle to 30 ms, giving each DUT 11 msec
on-time out of 15 msec per half cycle. Or, if I used a FULL cycle for
each device's "on" time, it might be much better, even at 50 Hz.

I'll probably just have to wait and see what the display looks like at
different rates and see what looks acceptable. It seems like it
probably won't be very good at all for the lower sweep frequencies,
since it wouldn't even allow enough time for one entire cycle of the
base or gate staircase sequence, although it would probably pick up
the entire sequence over multiple cycles, possibly even rapidly enough
to produce a good display. Or, maybe I could control the timing so
that the switching is synchronized with the staircase cycles. It'll
be fun to see what works, at any rate (excuse the pun).

I'm also kind of wondering how much of a problem the transients might
be.

I see that www.mouser.com also has some Axicom miniature DPDT PCB
relays; three models with "expected mechanical life" ratings of 15
million operations, 100 million operations, and 1 billion operations,
for $1.47, $1.87, and $2.23 for qty 1. The 1 billion ops model is
only rated for 1 Amp. The 100 million and 15 million ops models are
rated at 5 amps and 3 amps.

MAYBE the 100-million-operations-lifetime model would be good enough,
if I ran it at only 1 or 2 ops/second. If I'm calculating correctly,
that should be an expected life of something between 25,000 and 30,000
hours, if cycling at 1 op/second, which is over 1,000 24-hour days,
which is 3,000 8-hour days, giving an expected life that probably over
10 years of continuous switching for 8 hours a day, 5 days a week.

Thanks again, Bob!

Regards,

Tom Gootee

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

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