A-D front end - robust and for high voltage

[Gorilla Nerfball]

Can anyone shed some light on how multimeters, scopes and other
similar devices can measure anything from microvolts to 300V or more
without any moving parts (relays, etc.). Specifically, how do I design
the front end for an A-D capable of measuring up to 100V without
sacrificing performance at lower signal levels too much?

What about safety considerations? How do I isolate a 120V input from
the user, especially when that same front end has to measure
millivolts.

My application is a home brewed, flexible data logging/low freq
oscilloscope device. I want to be able to handle reasonably high
voltages (120V if possible), but at the same time be able to measure
millivolt waveforms via a high gain ins-amp at the front end. What if
I limit myself to something like 30V input, does that simplify things?

In the end, I need a high impedance input that's robust and can switch
between mV measurements and V measurements without any moving parts
(i.e. relays).

Any thoughts on where to start?

Thanks,

Chris

 

[Frank Miles]

Can anyone shed some light on how multimeters, scopes and other
similar devices can measure anything from microvolts to 300V or more
without any moving parts (relays, etc.). Specifically, how do I design
the front end for an A-D capable of measuring up to 100V without
sacrificing performance at lower signal levels too much?

What is "too much"? There are a lot of games that you can play with
attenuators (switching the low side, and relying to some extent on
input protection / fixed input limiting resistor) that fall apart when
you go from essentially DC to a much wider bandwidth.

What about safety considerations? How do I isolate a 120V input from
the user, especially when that same front end has to measure
millivolts.

Input resistor and low-leakage input protection.

My application is a home brewed, flexible data logging/low freq
oscilloscope device. I want to be able to handle reasonably high
voltages (120V if possible), but at the same time be able to measure
millivolt waveforms via a high gain ins-amp at the front end. What if
I limit myself to something like 30V input, does that simplify things?

What's the highest resolution that you will need? Can you attenuate
the input, and have variable gain thereafter?

In the end, I need a high impedance input that's robust and can switch
between mV measurements and V measurements without any moving parts
(i.e. relays).

Bandwidth? Resolution/accuracy for smallest inputs?

There are some really cool optically coupled semiconductor "relays"
(NAIS) but most of these have somewhat large leakage and capacitance
for these applications. The last I knew 'scope makers were still using
relays {I would be happy to hear if this information is out of date,
especially if that included how it was done...}

-frank
--

 

[[email protected]]

What is "too much"? There are a lot of games that you can play with
attenuators (switching the low side, and relying to some extent on
input protection / fixed input limiting resistor) that fall apart when
you go from essentially DC to a much wider bandwidth.

I think I'd like to get 8 bits of meaningful resolution at 1mV P-P, and
at least 12, ideally 16 bits at higher signal levels (10V P-P).
Bandwidth wouldn't need to be too high, ideally 100kS/s I'll be happy
with 10kS/s.

I understand what you mean about the input protection. If I was limited
to only a couple of volts, I'd just switch gain on an instrumentation
amp, letting it clip to its heart's content if the signal got too high.
But at 100V, it'd fry the whole thing.

Input resistor and low-leakage input protection.

I get the input resistor, the other thing I'm not familiar with. Assume
for a second that I have little or no experience with the practical
side of this type of electronics. Can you shed some more light on what
you mean by low-leakage input protection, perhaps in terms of a
circuit?

things?

What's the highest resolution that you will need? Can you attenuate
the input, and have variable gain thereafter?

16bits at 10V P-P, maybe 8 meaningful bits (after figuring noise, etc)
at 1mV P-P. If I attenuate the input, doesn't that inherently mean that
I attenuate my, already small, 1mV signal too?

Bandwidth? Resolution/accuracy for smallest inputs?

Again, 100kS/s ideally, 10kS/s would be ok.

There are some really cool optically coupled semiconductor "relays"
(NAIS) but most of these have somewhat large leakage and capacitance
for these applications. The last I knew 'scope makers were still using
relays {I would be happy to hear if this information is out of date,
especially if that included how it was done...}

Thanks for you help,

Chris

 

[Mark Jones]

What is "too much"? There are a lot of games that you can play with
attenuators (switching the low side, and relying to some extent on
input protection / fixed input limiting resistor) that fall apart when
you go from essentially DC to a much wider bandwidth.




Input resistor and low-leakage input protection.




What's the highest resolution that you will need? Can you attenuate
the input, and have variable gain thereafter?




Bandwidth? Resolution/accuracy for smallest inputs?

There are some really cool optically coupled semiconductor "relays"
(NAIS) but most of these have somewhat large leakage and capacitance
for these applications. The last I knew 'scope makers were still using
relays {I would be happy to hear if this information is out of date,
especially if that included how it was done...}

-frank

Hmm. I bought an older ('93) Iwatsu DS6121 DSO from military surplus.
It works well and has a series of small relays inside that switch the
active preamp. They barely even make a "click" sound. (It also has two
slots full of "74F" series logic chips comprising the RAM - could
probably save a couple hundred watts if they were changed to CMOS!)


There are several ways to accomplish "auto-ranging", here's one idea.
I have not made this device nor tested it, so take these words with a
big grain of salt. It is also not linear in any way. It is just an
idea to get you thinking. Perhaps it is useful, perhaps it is not.

"Smart" electronics usually require the use of a processor somewhere,
and this is such an idea. Imagine that an input voltage is divided
between "taps" on a resistor ladder and the voltage is read relative
to that divisor. i.e.,


[Vin]------+-------o Tap 0 (Vin * 1)
|
/
\R1
/6.4M 1%
\
| _
+-------o Tap 1 (Vin * 2.06)
|
/
\R2
/3.2M 1%
\
|
+-------o Tap 2 (Vin * 4.428571429)
|
/
\R3
/1.6M 1%
\
| _
+-------o Tap 3 (Vin * 10.3)
|
/
\R4
/800k 1%
\
|
+-------o Tap 4 (Vin * 31.0)
|
/
\R4
/400k 1%
\
|
-----
---
-

Those values can surely be tweaked to provide better numbers, but
frankly I ain't got that kind of time.

Now you would run all these taps to something like a 4066 digital
switch (provided one of these can handle at least +15v in the "on"
condition, but they may not like +155v in the "off" condition - please
check its datasheet. A series of opto-isolators and transistor
"switches" might work in place of the 4066.)

Okay before one gets too confused, here's what I'm getting at. For a
155v DC input voltage:

Tap 0 = 155.00v
Tap 1 = 75.00v
Tap 2 = 35.00v
Tap 3 = 15.00v
Tap 4 = 5.00v

Your device reads tap 4. It is within 0-5v (at 5.00v.)
Your device reads tap 3. It is outside 0-5v, testing halted.
Tap 4 = "multiply value read by 31 to get volts present"
5.00 * 31 = 155.00v


For a 32.0v input voltage:

Tap 0 = 32.00v
Tap 1 = 15.50v
Tap 2 = 7.22v
Tap 3 = 3.10v
Tap 4 = 1.03v

Your device reads tap 4. It is within 0-5v (at 1.03v.)
Your device reads tap 3. It is within 0-5v (at 3.10v.)
Your device reads tap 2. It is outside 0-5v, testing halted.
Tap 3 gives higher resolution than tap 4, so use that.
Tap 3 = "multiply value by 10.32258065 to get volts"
3.10 * 10.32258065 = 32.00v


For a 8.55v input voltage:

Tap 0 = 8.55v
Tap 1 = 4.14v
Tap 2 = 1.93v
Tap 3 = 0.827v
Tap 4 = 0.276v

Your device reads tap 4, 3, 2, 1, and stops at 0.
Tap 1 gives highest resolution, so use that.
Tap 1 = "multiply value by 2.065217391 to get volts"
4.14 * 2.065217391 = 8.55v

Now if you wanted mV resolution and below, turn on an analog x2 or x4
multiplier at tap 0 (as long as tap 0 voltage is /2 or /4 VanalogVcc.)

For that matter, instead of "peeking" and "poking" a ladder this way,
a better idea might be to "R2R ladderize" the feedback loops of a
division-mode op-amp and a multiplication-mode op-amp connected in
series. Say a 4-bit R2R ladder for each, so an 8-bit microcontroller
can control the whole shebang with one port. (i.e., high nibble =
multiplier amp, low nibble = divider amp, cycle through all range
combinations until measurement closest but under 5.00v is found, then
compute volts * (all multipliers) / (all divisors).)

Hmm, calibration sounds tricky though. And temperature stability-
ehhhh!

Again, just ideas. Besides, experimentation is good for you.


P.S. most "C" compilers for microcontrollers can do floating-point
math. If C isn't your thing, may I suggest a peek at Ziya Erdimir's
exponent-mantissa port to JAL,
http://groups.yahoo.com/group/jallist/files/Ziya Erdemir/
and my handy-dandy JAL floating-point-to-LCD-formatting front-end,
http://groups.yahoo.com/group/jallist/files/heliosstudios/


-- "One day, personal computers are going to be labeled just as
addictive as narcotics." MCJ 200405

 

[Apostrophe Police]

Hmm. I bought an older ('93) Iwatsu DS6121 DSO from military surplus.
It works well and has a series of small relays inside that switch the
active preamp. They barely even make a "click" sound. (It also has two
slots full of "74F" series logic chips comprising the RAM - could ^^^^^^^^^^
probably save a couple hundred watts if they were changed to CMOS!)

An Apostrophe Police Medal of Honor, to the one who is the first I've
seen use that word properly in over 25 years.

 

[Keith Williams]

An Apostrophe Police Medal of Honor, to the one who is the first I've
seen use that word properly in over 25 years.

Your kidding.

 

[Frank Miles]

I think I'd like to get 8 bits of meaningful resolution at 1mV P-P, and
at least 12, ideally 16 bits at higher signal levels (10V P-P).
Bandwidth wouldn't need to be too high, ideally 100kS/s I'll be happy
with 10kS/s.

Ouch! Let's assume for the moment that you need an input R of 1megohm
(standard for 'scope), and that the 100kS/s is roughly comparable to
50kHz bandwidth. Just to make things really simple. This translates
to a noise voltage (at room temp) of ~27uV rms, assuming that your
electronics is perfectly noiseless -- this just including the source
resistance. 8 bits with 1mV p-p translates to a step size of ~4uV.
That's pretty absurd if you're looking for 'scope-like operation.

You're not going to get there without changing something pretty
drastic. If you can narrow the bandwidth (a lot) you might get somewhere;
you can't lower in input resistor much without getting into trouble,
unless you use relays and decrease the input impedance of your device.

Of course, there is a more complex and expensive way of doing this.
If you were willing to put in a high voltage voltage-follower on the input,
you could do some cute things with some "high voltage" diode bridges
at relatively low impedances. If cost is no object this might be possible.
(I actually have done something similar for a special biomedical app).
A relay or two is much cheaper.

I understand what you mean about the input protection. If I was limited
to only a couple of volts, I'd just switch gain on an instrumentation
amp, letting it clip to its heart's content if the signal got too high.
But at 100V, it'd fry the whole thing.


I get the input resistor, the other thing I'm not familiar with. Assume
for a second that I have little or no experience with the practical
side of this type of electronics. Can you shed some more light on what
you mean by low-leakage input protection, perhaps in terms of a
circuit?

I think (without actually looking at any DMM schematics) that they rely on
an initial fixed input resistor, and do all their switching closer to ground.
That way the switches don't have to operate at high voltages. The penalties
are twofold: (1) input protection leakage currents must be very small, since
they will contribute substantially (I*R) to errors; and (2) you don't benefit
from low source resistances to reduce the noise when the signals aren't so
big. Of course, with a 'scope-like input I assume that you don't want
the input capacitance to be large, either, unlike DMMs!

I suggest you rethink your specs, especially why you cannot tolerate relays.

16bits at 10V P-P, maybe 8 meaningful bits (after figuring noise, etc)
at 1mV P-P. If I attenuate the input, doesn't that inherently mean that
I attenuate my, already small, 1mV signal too?

[snip]

HTH...

-frank
--

 

[Apostrophe Police]

Your kidding.

This isn't a sentence either. ;-)

 

[Bradley1234]

Sure, use ohms law and a resistive divider network.

Imagine 2 resistors in series, a 10 Megohm and a 100 ohm. The total is
about 10M

Ohms law approximates voltage EMF to be = current (I) x resistance (R)
and if youre hooked into 300V, with 10M then you need to find I which is I
= E divided by R, so 300/10M I think thats 3E-5 Amps (3E2/1E-7)

Or take the ratio of 10M/100 which may be 100,000 to 1, if you read across
the 100ohm resistor its a small voltage

Now you can tailor the divider in such a way to get the input swing voltage
on your measurement device

 

[John Larkin]

Can anyone shed some light on how multimeters, scopes and other
similar devices can measure anything from microvolts to 300V or more
without any moving parts (relays, etc.). Specifically, how do I design
the front end for an A-D capable of measuring up to 100V without
sacrificing performance at lower signal levels too much?

DVMs have big manually-operated rotary switches, as do analog scopes.
Most digital scopes do in fact have relays; you can hear them click as
you change vertical ranges.

John

 

[Tom Del Rosso]

Now you would run all these taps to something like a 4066 digital
switch (provided one of these can handle at least +15v in the "on"
condition, but they may not like +155v in the "off" condition - please
check its datasheet. A series of opto-isolators and transistor
"switches" might work in place of the 4066.)

I think this part of the issue is the part that the OP was asking about, and
I don't think any of the respondents really addressed it. Obviously a 4066
would fry.

My Fluke meter has no relays (and damn few of any type of components). It
does have what looks like a resistor network on a 1cm x 4cm substrate with
several taps, but nothing that looks like a solid state relay. There are
several SMT transistors, but only one or two TO92s.

 

[[email protected]]

Perhaps it's time for me to spend some money on a multimeter and a good
pry bar...

Thanks for all the posts, they've given me a bit to work on. In
general, I think I'm going to end up making special gain stages
separate from the main DAQ. Connectors are cheap and it won't take much
to snap on a high gain or high attenuation module to an otherwise
ordinary +/-10V DAQ front end.

Chris

 

[Tom Del Rosso]

DVMs have big manually-operated rotary switches, as do analog scopes.
Most digital scopes do in fact have relays; you can hear them click as
you change vertical ranges.

But how do autoranging DVMs do it with only a few SMT transistors?

 

[John Larkin]

But how do autoranging DVMs do it with only a few SMT transistors?

Big resistors. That's easy for DC or low-frequency AC. Scopes are
wideband and can't tolerate large unbypassed resistors, so need
switched dividers. Solid-state switches still have too much
capacitance and can't tolerate overvoltages, so most scopes still use
a relay or two in the sront-end.

John

 

[Tom Del Rosso]

Big resistors. That's easy for DC or low-frequency AC. Scopes are
wideband and can't tolerate large unbypassed resistors, so need
switched dividers. Solid-state switches still have too much
capacitance and can't tolerate overvoltages, so most scopes still use
a relay or two in the sront-end.

Doesn't the "big resistor" divider have to be switched too?

 

[keith]

Doesn't the "big resistor" divider have to be switched too?

Sure, but "big resistor" isn't so sensitive to "small switch resistance"
and "low bandwith" doesn't have to have the high capacitance of the
FET switches compensated for. It's all about bandwidth.

 

[Rich Grise]

DVMs have big manually-operated rotary switches, as do analog scopes.
Most digital scopes do in fact have relays; you can hear them click as
you change vertical ranges.

I've heard relays click as an autoranging DVM autoranged.

But about scopes and knobs, if you have your scope set at 1 mV/div, and
accidentally hit the mains with it, what do they use for proteciton there?
Just a high resistance, and a couple of diodes?

In the old days, they used the fact that if the grid didn't arc over, the
tube would be OK. ;-)

Thanks,
Rich

 

[Tom Del Rosso]

Sure, but "big resistor" isn't so sensitive to "small switch resistance"
and "low bandwith" doesn't have to have the high capacitance of the
FET switches compensated for. It's all about bandwidth.

I'm sorry I haven't been very explicit about what has me baffled. What I
don't understand is how those small switches can take the high voltage
sometimes present. For low ranges I think there has to be a switch near the
top of the divider, right? If so, why doesn't it get fried when there's a
high-voltage input?

 

[Mark Jones]

I'm sorry I haven't been very explicit about what has me baffled. What I
don't understand is how those small switches can take the high voltage
sometimes present. For low ranges I think there has to be a switch near the
top of the divider, right? If so, why doesn't it get fried when there's a
high-voltage input?

200Vce bipolars/fets?

-- Kzizz to Ikzik: "Hey Ik... didn't we pass that barn an hour ago? YOU IDIOT,
WE'RE FLYING IN CROP CIRCLES!!!"

 

[Frank Miles]

I'm sorry I haven't been very explicit about what has me baffled. What I
don't understand is how those small switches can take the high voltage
sometimes present. For low ranges I think there has to be a switch near the
top of the divider, right? If so, why doesn't it get fried when there's a
high-voltage input?

Imagine a 2-resistor attenuator:

Vin ------ R1 ------+----- Vo
|
R2
|
Sw
|
Gnd

If switch Sw is open (for low-voltage signals), there's no attenuation.
Sensitive operation. If Sw is closed, you get attenuation. Sw never has
to see the full input voltage. To protect the ADC measuring Vo, you need
a low-leakage protecting device of some kind. The source is protected from
the clamp by R1.

Expand to get more ranges.

-frank

--