Isolated variable resistor function?

[Joerg]

Hehe. I was just needing a faster PWM (wanted to use a very
cheap RC, 20db/decade rolloff, low-pass, but needed an audio
spectrum. Couldn't get there with 1/62500 PWM period of
standard MSP430. So I looked and... cripes!! They had
something else!

"Timer D also includes a dual capture mode reducing loading
of capture operation by half. In addition, the combining
compare blocks help to control both rising and falling edges
of the PWMoutput signal. The Timer-Event-Control-block offers
external triggering options as well as internal
synchronization of timer instances."

See the MSP430F51x2 parts.

http://www.ti.com/product/msp430f5172&DCMP=msp430&HQS=430timer

http://www.ti.com/litv/pdf/slau208k

Quote, page 511, Time_D Operation: "The timer input clock frequency can
be within the range of 8 MHz to 16 MHz if 16x is selected (TDHMx = 00)
or 8 MHz to 25 MHz if 8x (TDHMx = 01) is selected".

That would be an order of magnitude less than 256MHz. Or am I missing
something here?

 

[Joerg]

You can make a nice fake resistor from an MDAC. The cheap DACs on most
uPs will work as single-quadrant MDACs. Do you have a uP on the
isolated side? You could send it serial UART messages through an opto.

There is only a uC on the low voltage side but nothing on the mains side
where the controlled resistor is needed.

If not, an MDAC or a digital pot is probably the answer. One dual
optoisolator, or even a single, can be enough to get SPI data up
there, with a little glue logic.

Well, since I am not going to use one from M.... I think I am stuck with
the MCP series or I2C stuff. The MCPs only have up/down command so you
can't command them to go to a particular position. But that's ok, I can
let the uC quickly peg it and go from there. It's just that this method
is, how shall I say, a bit pedestrian.

Too bad nobody makes analog multipliers any more, or at least
affordable ones.

Sure they do. For example, this one can be had for around 30c:

http://www.onsemi.com/pub_link/Collateral/MC1496-D.PDF

 

[Joerg]

John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"

So changing the value can be slow, but the "resistor" must be a good
resistor to several MHz?

Yup. Pretty much like the gain control in an IF amplifier. It won't have
to react faster than the modulation of the signals after a narrowband
filter, typically a few kHz. But the amp has to work at several MHz.

I might just do the digital potmeter thing. Not pretty but works and is
cheap.

 

[[email protected]]

John Larkin wrote: [snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.

Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"

Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)

 

[[email protected]]

The signal going into this divider can't be distorted more than 5% and
with a FET-optocoupler it would be.
MDAC-opamp-FET



Sure, but then I might as well just use my solution "b", a digital
potmeter. Reduces that function to one part costing 30-40c, outside the
optocoupler to xfer the data lines which I'd need either way.

I had problems with power dissipation of DPots. There should be a way of
eliminating the MDAC, too.

 

[Jon Kirwan]

http://www.ti.com/litv/pdf/slau208k

Quote, page 511, Time_D Operation: "The timer input clock frequency can
be within the range of 8 MHz to 16 MHz if 16x is selected (TDHMx = 00)
or 8 MHz to 25 MHz if 8x (TDHMx = 01) is selected".

That would be an order of magnitude less than 256MHz. Or am I missing
something here?

Yes, probably you are missing something. They include a x16
multiplier. But the PWM clocking system can't handle anything
faster than 256MHz. It dies after. But the CPU can run at
25MHz. So if you use the x16 multiplier, you can't run the
CPU above 16MHz or else you exceed the 256MHz spec on the PWM
counter chain.

I think, anyway.

Jon

 

[Joerg]

John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"

Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)

But not da big ones.

 

[Jon Kirwan]

In case that earlier answer wasn't clear enough, look at
Figure 19.2 in the family user's guide where they talk about
the high resolution generator. It can multiply by 8 or by 16.
And start reading the text around there and below. I think
you will see what is going on. (Atmel does it on some of
their parts, too.)

See also:

http://www.ti.com/lit/an/slaa497/slaa497.pdf

Jon

 

[legg]

Hi Folks,

Does anyone know a part that is small, can be used to command a certain
resistance and can be talked to across the usual mains-isolated barrier
of office equipment and the like? I need to set resistor values on the
mains side from a uC on the low voltage side. Has to be:

Linearity error from 0V to about 1V excitation should be <5%.
Absolute resistance tolerance <20% if possible.
Bandwidth for resistor value changes up to a few kHz.
No limit on the number of changes (meaning no EEPROM).
Addressable across isolation barrier.

So far I've considered:

a. FET optocouplers such as the H11F1. A bit pricey at over $1 but most
of all has the problem that the linearity goes to pots as soon as there
is more than 20-30mV across the FET, plus it needs a ton of drive power
for low resistance:

http://www.fairchildsemi.com/ds/H1/H11F1M.pdf

b. Digital potmeters such as the MCP4011 series. These are nice, tons of
bandwidth, cheap and small. But they do need an additional two-channel
optocoupler to get the command data across and the absolute resistance
value tolerance is borderline at 20%:

http://ww1.microchip.com/downloads/en/DeviceDoc/21978c.pdf

c. Servoed FET plus opto. The servo gets rid of the non-linearity. Done
it before but this gets old because it needs a lot of parts and real estate.

d. LDR plus LED: Very nice, but fairly large and most of all LDRs are
banned in some countries on account of their Cadmium content.


Any other miracle parts out there?

Give a non-isolated example that does absolutely everything you need,
then we'll figure out how to do it, isolated.

I'm sick to death of weaseling.

RL

 

[[email protected]]

Jim Thompson wrote:
John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"

Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)

But not da big ones.

Yeah, one is a "tiny" 400W boost. I cheated, though, and it's four-phase. ;-)
Most of the others are in the 15-20W class.

 

[legg]

On Sat, 25 Aug 2012 19:41:35 -0700 (PDT),

This is another one of your zero context questions. Sounds like you're
working on a amplifier with uC controlled gain, and some crazy
demod/mod scheme with the couple hundred kiloHz control frequency?

Or just replacing a pot with a plasic shaft, to go vroom vroom.

When you see resistance-controlled interface, you have to think 'OLD'.
You get the gold star if you can do it some other way.

A resistor seldom has to do everything:

- It may be polarized, in it's place.

- It may be used with a fixed voltage - actually modulating branch
current.

- It may be used at a fixed current - actually modulating a node
voltage.

If a variable resistor did work, then there were acceptable limits of
linearity, resolution, minimum and maximum compliance values that
would be expressed and passed on.

If it didn't actually work, then there's little point in promoting a
substitute that attempts the same degree of basic malfunction.

I hope this isn't another frigging lamp dimmer, with power factor
correction. These will likely never see significant North American
manufacture, no matter what brand name gets stamped on them.

RL

 

[Joerg]

On Sat, 25 Aug 2012 19:41:35 -0700 (PDT),

The above describes quite accurately what's needed. Of course I can't
post a schematic.

Or just replacing a pot with a plasic shaft, to go vroom vroom.

When you see resistance-controlled interface, you have to think 'OLD'.
You get the gold star if you can do it some other way.

Only if it costs less.

A resistor seldom has to do everything:

- It may be polarized, in it's place.

Yes, it can be.

- It may be used with a fixed voltage - actually modulating branch
current.

Unfortunately it can't :-(

- It may be used at a fixed current - actually modulating a node
voltage.

I wish, but no.

If a variable resistor did work, then there were acceptable limits of
linearity, resolution, minimum and maximum compliance values that
would be expressed and passed on.

This was stated in my original post: <5% linearity error, <20% absolute
resistance tolerance, 1V max excitation. I thought that was quite clear.
Resolution can be as low as 6 bits, maybe even 5 bits.

If it didn't actually work, then there's little point in promoting a
substitute that attempts the same degree of basic malfunction.

I hope this isn't another frigging lamp dimmer, with power factor
correction. These will likely never see significant North American
manufacture, no matter what brand name gets stamped on them.

You may be very mistaken there

 

[Joerg]

In case that earlier answer wasn't clear enough, look at
Figure 19.2 in the family user's guide where they talk about
the high resolution generator. It can multiply by 8 or by 16.
And start reading the text around there and below. I think
you will see what is going on. (Atmel does it on some of
their parts, too.)

See also:

http://www.ti.com/lit/an/slaa497/slaa497.pdf

Amazing! Maybe I should use a F5 series some day.

 

[Joerg]

Jim Thompson wrote:
John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"
Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)

But not da big ones.

Yeah, one is a "tiny" 400W boost. I cheated, though, and it's four-phase. ;-)
Most of the others are in the 15-20W class.

2MHz at 400W? That's pretty good. How did you get that to pass EMC and

be efficient?

 

[Joerg]

MDAC-opamp-FET

Sure, but that's quite a lot of parts. I was planning on doing it with a
single device plus optocoupler.

I had problems with power dissipation of DPots. There should be a way of
eliminating the MDAC, too.

The MCP4011 series is listed at 15uA when operated at 2.7V. I could live
with that

 

[Joerg]

Actually, that's the balanced modulator version. The MC1495 - no
longer made - had the extra circuitry to make it a more or less
accurate multiplier, which slowed it down a bit.

It wasn't anything like as good as the Analog Devices parts are now,
and it was a bit if pig to use - like the MC4024/MC4044 parts for
which Jim Thompson also admits responsibility - but it was cheap, as
it had to be to sell at all against the much better Analog Devices and
Burr-Brown parts.

That was just one example. There are many more, for example RF mixers
that can go down to zero even though the datasheet often don't mention
it. Typically well under a Dollar. AD makes good multipliers but that's
more for precision scientific instruments where cost doesn't really
matter. I don't need any precision here.

 

[Joerg]

Give a non-isolated example that does absolutely everything you need,
then we'll figure out how to do it, isolated.

I gave one, see "b". You can get digital potmeters that are more
accurate than 20% and cost more. But it's not isolated. No problem to
communicate with one via optocouplers but the whole purpose of this post
was to see if there's a more simple one-chip solution. Like the
optocouplers that contain a 2nd diode so you can servo out the transfer
function tolerance (but that's not resistive).

I'm sick to death of weaseling.

Then why did you answer?

 

[[email protected]]

[email protected] wrote:

Jim Thompson wrote:
John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"
Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)

But not da big ones.

Yeah, one is a "tiny" 400W boost. I cheated, though, and it's four-phase. ;-)
Most of the others are in the 15-20W class.

2MHz at 400W? That's pretty good. How did you get that to pass EMC and

be efficient?

I just got the first boards back a few weeks ago, so haven't, yet. ;-)
However, efficient should help. An extruded aluminum case should help, too
(the final product will be in a casting).

As I said, it's cheating but it's really four phases, each at 500kHz. It's an
advanced development project. One of the groundrules is that switching
regulators must be above 2MHz (AM band). This cheating was the only way I
found to get there. On another project, I have a couple of other smaller
2-phase buck regulators (2-phase, to get the power/efficiency up where its
needed).

 

[[email protected]]

Sure, but that's quite a lot of parts. I was planning on doing it with a
single device plus optocoupler.

We don't always get our wishes.

The MCP4011 series is listed at 15uA when operated at 2.7V. I could live
with that

Isn't the tolerance of the 4011 an issue?

 

[Joerg]

[email protected] wrote:

Jim Thompson wrote:
John Larkin wrote:
[snip]
Any chance you can PWM a resistor?

Would be nice but the part where the resulting signal goes into is
blazingly fast (has to be). So the PWM would cause an undesired modulation.

But it is an idea, maybe I can slow something down in there. And go in
with several meggeehoitzes. The uC we have would moan and groan though.
Please define "blazingly fast"?

In this case several MHz. In the world of switch mode power supplies
that counts as "blazingly fast"
Blazingly? Every switching supply I've designed in the last nine months has
had a Fsw > 2MHz. ;-)
But not da big ones.
Yeah, one is a "tiny" 400W boost. I cheated, though, and it's four-phase. ;-)
Most of the others are in the 15-20W class.

2MHz at 400W? That's pretty good. How did you get that to pass EMC and

be efficient?

I just got the first boards back a few weeks ago, so haven't, yet. ;-)
However, efficient should help. An extruded aluminum case should help, too
(the final product will be in a casting).

As I said, it's cheating but it's really four phases, each at 500kHz. It's an
advanced development project. One of the groundrules is that switching
regulators must be above 2MHz (AM band). This cheating was the only way I
found to get there. On another project, I have a couple of other smaller
2-phase buck regulators (2-phase, to get the power/efficiency up where its
needed).

If you run them at 500kHz that's a bog-standard switcher range. That's
not a 2MHz switcher. You won't be above the AM band with that unless you
have a super-precise balancing system in there.