100V common mode diff front end amp?

[Grant]

Looks nice but I don't think the (creapage) distance between the pins
is large enough to actually work reliably at 200V (or more). I also
like to add a capacitor between the inputs of the instrumentation amp
to filter unwanted signals.

Well creepage distance is for diff voltage which is usually lower but
can be accidentally high, but unlikely over 100V for my stuff.

Conformal coating for creepage distance?

Or perhaps the safer three opamp topology using the inverting inputs for
high voltage? That's the other way to do it. Use single opamps each nulled
out, and 0.1% resistors instead of the laser trimmed ones in the INA117...

Response is already down to 200kHz, how much less? I'm already suffering
loss of bandwidth

Grant.

 

[Nico Coesel]

Well creepage distance is for diff voltage which is usually lower but
can be accidentally high, but unlikely over 100V for my stuff.

The common mode voltage is also close to the supply pins!

Conformal coating for creepage distance?

Or perhaps the safer three opamp topology using the inverting inputs for
high voltage? That's the other way to do it. Use single opamps each nulled
out, and 0.1% resistors instead of the laser trimmed ones in the INA117...

I like to use an instrumentation amp like the INA332 or INA333 and
have an external divider. If you use 3 seperate opamps the precision
resistors will cost you more than a single instrumentation amp.

 

[[email protected]]

Which reminds me. In thinking delta-sigma modulators... I just got an
inquiry... I vaguely recall of a trivial way to turn sigma-delta into
a Manchester-like stream, then restore easily on the receiving side...
nice for passing thru opto-couplers.

Something resembling SACD ?

 

[Grant]

On Sun, 24 Apr 2011 07:09:07 +0300, [email protected] wrote:

[snip]

At the differential output, run an other amplifier stage driving an
optoisolator. To linearize the opto output, put an identical
optoisiolator in series on the input side and use the output from it
as feedback to the second amplifier stage.

[snip]

I did that with isolated sensing on switching power supplies in the
early '80's. I vaguely remember issues with linearity over a large
dynamic range.

Yeah, back then with slow signals I had better luck transferring the
analog value as a duty cycle though the optocoupler, transfer was very
accurate.

I suppose now you'd call the technique self-clocked delta-sigma, or
something like that? Signal into an integrator that forced output
duty cycle to match, therefore recovery on other side of opto was
very close to original signal.

I used CMOS inverters for rail-to-rail buffers to get the accuracy.

Grant.

Which reminds me. In thinking delta-sigma modulators... I just got an
inquiry... I vaguely recall of a trivial way to turn sigma-delta into
a Manchester-like stream, then restore easily on the receiving side...
nice for passing thru opto-couplers.

Well if you clocked a flipflop inside the signal path? But it wouldn't
be strict Manchester-like until you forced AC into the data stream to
make the average zero requirement.

I didn't so that, and decoding Manchester has always remained a mystery
to me, used a black box that does it Which of course makes me wonder
if you forced Manchester in the feedback path, would the analog data force
correct digital sort of directly, automagically?

When you first posted, I thought of using current probes... I have
some good from DC to 50MHz, but the front-end would end up being power
hungry :-(

I'd like a decent current probe, Tek want anything north of $900 for one,
so I made one from something I read on the 'net, it let me see a waveform,
but of course no calibration. Some help, at the time I was investigating
the nasty switching MOSFET drain to gate current injection, before I knew
it was normal behaviour.

Grant.

 

[Grant]

I like to use an instrumentation amp like the INA332 or INA333 and
have an external divider. If you use 3 seperate opamps the precision
resistors will cost you more than a single instrumentation amp.

Thank you. Hmm, INA333 4000 hours operation from LI coin cell!
Better than +/-15V

Grant.

 

[Grant]

You just xor the data stream with the clock, no? Recovery is almost as
simple.

You have a block diagram for recovery? I never got my head around the
clock extraction, how it worked.

And, you have to take care about the data edge setup time (right term?
make sure you don't glitch output, latching moving input data) on the
encoder?

Grant.

 

[Grant]

Each of the five BNC inputs (four channels and trigger) is fully
isolated. You can use any probes, or no probes.

Yes, I looked at datasheet yesterday, wont happen in my world. They did
quote that 30V limit for ground leads, but they do that now for handheld
multimeter leads I bought recently. Maybe some country has a silly law?

Grant.

 

[[email protected]]

You have a block diagram for recovery? I never got my head around the
clock extraction, how it worked.

The Manchester encoded signal has a transition in the middle of the cycle
(falling edge of the clock). If that transition is positive, it's a '0', if
negative, it's a '1'. The corollary to this is that if there is a transition
at the beginning of the "cell", D(n) = D(n-1). A phase lock loop can pretty
easily find the clock.

 

[josephkk]

Yes, I looked at datasheet yesterday, wont happen in my world. They did
quote that 30V limit for ground leads, but they do that now for handheld
multimeter leads I bought recently. Maybe some country has a silly law?

Grant.

It seems like we have a global population explosion of silly laws
recently. Decent people have let down their guard, or maybe it was
battered down by insane courts.