Looking for Ready made stand alone DIGITAL PID.

[Jamie]

As it said.
We need to replace some analog PID boards due to supply and
side effects the existing ones have.

The PID code that comes in digital drives and process controllers
are pretty much what we're looking for how ever, we only need just
a design that allows a input ref, reference feed back and a output
to drive what ever device.
This device should have it's own mini LCD interface to set the
values or at least a computer hook up to do so.

It should have the standard Proportional Gain, Integral and
derivative components to set the values.

Also, having an option to select unipolar and bipolar output
is a plus!..

I know there are various analog boards out there already that
do this how ever, We have found that they seem to exhibit side
effects and cause self oscillation under some conditions.

In the digital varieties, I have found this not to be a problem
because it's being done via software and thus these effects are not visible.

I have already made a few using a AVR/ATMega that employ the ADC
inputs and we simply buffer the output into a PWM filter and op-amp.
These units are programmed via a serial connection for the params.

How ever, my employer wants to buy units that are ready made so that
replacements can be had with out depending on in-house designs.

I have looked and I can't find a unit that has a build on LCD display
with a simple interface to do this..

We do have a SSD out board key pad unit that does do this function for
us along with some other little nifty things how ever, it appears that
this unit is no longer available.

Any idea's?

 

[Jon Kirwan]

<snip about PID>
Any idea's?

No ideas. Some comments.

I'm not clear about exactly how you are using these PID units, right
now. It almost sounds as though you are an off-the-shelf consumer of
PID controllers with a specific process to control -- except for the
fact that you talk about unipolar and bipolar operation being a plus,
which suggests that you have a variety of situations to deal with and
not just the one. That makes me imagine you more in the position of
not really knowing where the PID is being applied and that you want a
more general solution as an OEM supplier to end users (which implies
you aren't the end user of the PID.) Are you and end-user or do you
supply something to others that must include PID as a feature?

You also mention self-oscillation, which is a fairly common problem
I've seen with PID -- especially where there is a significant time
delay in the loop. One of the cheaper ways to handle that is to
simply shorten the delay. Even a brain-dead PID controller can do a
credible job on controlling a process where the loop time is shorter
and repeatable.

You mention having created a solution with an ATmega that works
already, but that your company doesn't want to use it. Instead
preferring to buy something on the outside. Why? Many would very
much like to be in control of something like this. You state the
desire for 3rd party suppliers of PID boxes and a lack of worrying
about supply problems in the future, etc. But isn't that exactly what
you get when you make it, yourself?

Is the real reason because they don't believe you've really got all
the "quirks" worked out and they don't want to take risks that there
will be a lot more time needed to shake out what you've already done?

Small or first-time suppliers will be "risky" from that point of view,
I suppose. So which larger manufacturers have you already looked at,
what devices, and why aren't they suitable? That would also help a
lot, for whomever tries to answer you effectively.

Tim Wescott may have a great deal to add to all this, but frankly I'm
in need of more information before having any idea what to recommend,
technically.

Jon

 

[Joerg]

As it said.
We need to replace some analog PID boards due to supply and
side effects the existing ones have.

The PID code that comes in digital drives and process controllers
are pretty much what we're looking for how ever, we only need just
a design that allows a input ref, reference feed back and a output
to drive what ever device.
This device should have it's own mini LCD interface to set the
values or at least a computer hook up to do so.

It should have the standard Proportional Gain, Integral and
derivative components to set the values.

Also, having an option to select unipolar and bipolar output
is a plus!..

Up to here it sounds like that is feasible with programmable logic
controllers (PLC). Allen-Bradley, Danaher, Omron, Siemens and many
others. But it depends how fast your loop needs to be.

I know there are various analog boards out there already that
do this how ever, We have found that they seem to exhibit side
effects and cause self oscillation under some conditions.

In the digital varieties, I have found this not to be a problem
because it's being done via software and thus these effects are not
visible.

I have already made a few using a AVR/ATMega that employ the ADC inputs
and we simply buffer the output into a PWM filter and op-amp.
These units are programmed via a serial connection for the params.

How ever, my employer wants to buy units that are ready made so that
replacements can be had with out depending on in-house designs.

Why? Put an LCD and buttons on them, have them produced, use as many as
needed for your purposes and sell the rest like hot cakes, make oodles
of money. Ask your employer for a 2-3% revenue sharing incentive, buy a
tropical island, a LearJet, a few crates of margarita mix, then retire.

I have looked and I can't find a unit that has a build on LCD display
with a simple interface to do this..

We do have a SSD out board key pad unit that does do this function for
us along with some other little nifty things how ever, it appears that
this unit is no longer available.

Any idea's?

PLC can do all that if your loop timing isn't super fast. But they cost
a pretty penny and usually every ever so little extra such as LCD,
ADC/DAC or keypad costs extra. Oh, you want a steering wheel with that
car? That's be another $100. Maybe you can find a budget version that
has at least a tiny LCD and 4-5 buttons.

Boxes geared more towards PID:

http://www.alliedelec.com/Images/Products/Datasheets/BM/OMRON_ELECTRONICS_INC/834-0360.PDF
http://news.thomasnet.com/fullstory/453366

 

[Joerg]

That's exactly what we have:

http:linnix.com/pid.jpg

LCD, LM358 A2D, 2 or 4 buttons, 10 pins external interface.

Does it have a DAC as well so you can control the analog plant input?

We get several hundred hertz closed-loop responses.

That is sufficient for many if not most process control jobs.

We did so too. But AVRs are too expensive.


We can customize it for the OP.

We sell them for less than $10. It would need huge qty to afford the
island, jet, etc.

Now you've piqued my interest. Is there a datasheet of the unit and
maybe a photo with housing? Maybe a little more in focus, the one in the
link looks like it was taken after a meeting with Jose Cuervo ;-)

 

[Joerg]

Tim Wescott wrote:

[...]

I only know of two dedicated embedded PID controllers, including the
linnix one. The one example of this that I've worked with was pretty
pedestrian; in many cases you could just read my PID article and roll
your own -- anything you lost in PID expertise you'd gain back with an
intimate knowledge of how your own system.

I think, frankly, that the reason you don't see more is because it _is_
so easy to tuck a PID controller into the corner of a larger general-
purpose embedded controller -- so why go buy one special?

The market would be companies like Jamie's. For whatever reason,
liabitity, procurement, whatever, they don't want to roll their own.
Other times there just isn't anyone who can program a uC. But most of
all people want this:

Something that drops in, does not have potmeters that can get screwed
up, can be pre-programmed and shipped, and has mounting screw holes as
well as screw terminals. For high rollers maybe even a clickable DIN
rail version.

It'd be interesting to see what linnix has to offer, if the site would
cough up more than a few lousy pictures.

He said they'll have discussions, probably how much and when to release
info. Eagerly awaiting that ...

 

[Jamie]

Up to here it sounds like that is feasible with programmable logic
controllers (PLC). Allen-Bradley, Danaher, Omron, Siemens and many
others. But it depends how fast your loop needs to be.



Why? Put an LCD and buttons on them, have them produced, use as many as
needed for your purposes and sell the rest like hot cakes, make oodles
of money. Ask your employer for a 2-3% revenue sharing incentive, buy a
tropical island, a LearJet, a few crates of margarita mix, then retire.



PLC can do all that if your loop timing isn't super fast. But they cost
a pretty penny and usually every ever so little extra such as LCD,
ADC/DAC or keypad costs extra. Oh, you want a steering wheel with that
car? That's be another $100. Maybe you can find a budget version that
has at least a tiny LCD and 4-5 buttons.

Boxes geared more towards PID:

http://www.alliedelec.com/Images/Products/Datasheets/BM/OMRON_ELECTRONICS_INC/834-0360.PDF

http://news.thomasnet.com/fullstory/453366

We have high end process loop controllers how ever, these are intended
for user operation and is not what we're looking for..

using a process controller configured out side it's normal use does
not work well or not at all and most of them I've seen do not provide a
unipolar/bipolar outputs with differential inputs with a ref source.

The keyboard interface we're looking for is a very simply one. Just a
M, UP/DOWN pin switches with mini LCD screen edged soldered..

These units are intended for one time calibration and to be mounted
internally.

We have a variety of other devices that are DIN Rail mounted that
employs their own little mini LCD screen with a minimum set of buttons
to program them.

It's just that I can not find A PID unit in such a configuration for
DIN rail mounting.

PLC's are way expensive for this simple application.

But I'll take your suggestions of manufacturing the existing units we
have made in house as a solution for 200 or so to be made.
We have already looked into Display supported Uc's and the use of
simple serial LCD displays.

 

[Jamie]

Actually, it says we are too busy to make fancy web pages. The main
page log into our email system only. Our customers know were to look
for subfolders. We will update more for this later.

[...]

I only know of two dedicated embedded PID controllers, including the
linnix one. The one example of this that I've worked with was pretty
pedestrian; in many cases you could just read my PID article and roll
your own -- anything you lost in PID expertise you'd gain back with an
intimate knowledge of how your own system.

I think, frankly, that the reason you don't see more is because it _is_
so easy to tuck a PID controller into the corner of a larger general-
purpose embedded controller -- so why go buy one special?

The market would be companies like Jamie's. For whatever reason,
liabitity, procurement, whatever, they don't want to roll their own.
Other times there just isn't anyone who can program a uC. But most of
all people want this:

Unlike Flash, OTP development cost is very high. I don't think many
people want to deal with it. But volume production cost is much
lower. The entire BOM for this board is $2 to $3, same as the 10K
price for the Flash based LCD AVR chip alone.

Something that drops in, does not have potmeters that can get screwed
up, can be pre-programmed and shipped, and has mounting screw holes as
well as screw terminals. For high rollers maybe even a clickable DIN
rail version.

Give me more time, please.

He said they'll have discussions, probably how much and when to release
info. Eagerly awaiting that ...

Right now, I am busy trying to solve the external integrator problem.
Our customer needs better closed-loop response with hardware
integrator. At the end of a cycle, I need a high resistant switch to
discharge the capacitor. The MOSFET I tried did not isolate well
enough. Mechanical relay works, but size is a problem. I need a
SOT-23-5 sized semiconductor relay. Any suggestions?

You mean something like this?
http://www.rsinlibya.com/catalogue/category/meder-smt-micro-reed-relay/PCSR_292214.aspx

 

[Joerg]

Tim Wescott wrote:

[...]


Give me more time, please.

Ok, time's up now.

Just kidding ...

Right now, I am busy trying to solve the external integrator problem.
Our customer needs better closed-loop response with hardware
integrator. At the end of a cycle, I need a high resistant switch to
discharge the capacitor. The MOSFET I tried did not isolate well
enough. Mechanical relay works, but size is a problem. I need a
SOT-23-5 sized semiconductor relay. Any suggestions?

What's the leakage requirement? In many older analyzers this was done
with small JFETs. But if you only have +3V or something like this it'll
be more of a challenge.

 

[Joerg]

Tim Wescott wrote: [...]

It'd be interesting to see what linnix has to offer, if the site would
cough up more than a few lousy pictures.
Give me more time, please.

Ok, time's up now.

Just kidding ...

He said they'll have discussions, probably how much and when to release
info. Eagerly awaiting that ...
Right now, I am busy trying to solve the external integrator problem.
Our customer needs better closed-loop response with hardware
integrator. At the end of a cycle, I need a high resistant switch to
discharge the capacitor. The MOSFET I tried did not isolate well
enough. Mechanical relay works, but size is a problem. I need a
SOT-23-5 sized semiconductor relay. Any suggestions?

What's the leakage requirement? In many older analyzers this was done
with small JFETs. But if you only have +3V or something like this it'll
be more of a challenge.

It would be approx. 10K R and 0.01uF C with 3.5V charging and
discharging. If I connect a 2N7002 across the C, it wouldn't charge
at all. I guess the MOSEFT off-state leakage is larger than the 10K
R.

That's initially 350uA. A lot. The 2N7002 sits at 1uA for off. Maybe too
much flux or something?

But the 2N7002 is too fat for this and at higher temps leakage really
hits. These are usually better for this app although not spec'f at 0V:

http://www.nxp.com/acrobat_download/datasheets/BSS83_N_3.pdf

Anything from the SD-sries would probably be too expensive for your project.

 

[Fred Bartoli]

linnix a écrit :

linnix wrote:
Tim Wescott wrote:
[...]
It'd be interesting to see what linnix has to offer, if the site would
cough up more than a few lousy pictures.
Give me more time, please.
Ok, time's up now.
Just kidding ...
He said they'll have discussions, probably how much and when to release
info. Eagerly awaiting that ...
Right now, I am busy trying to solve the external integrator problem.
Our customer needs better closed-loop response with hardware
integrator. At the end of a cycle, I need a high resistant switch to
discharge the capacitor. The MOSFET I tried did not isolate well
enough. Mechanical relay works, but size is a problem. I need a
SOT-23-5 sized semiconductor relay. Any suggestions?
What's the leakage requirement? In many older analyzers this was done
with small JFETs. But if you only have +3V or something like this it'll
be more of a challenge.
It would be approx. 10K R and 0.01uF C with 3.5V charging and
discharging. If I connect a 2N7002 across the C, it wouldn't charge
at all. I guess the MOSEFT off-state leakage is larger than the 10K
R.

That's initially 350uA. A lot. The 2N7002 sits at 1uA for off. Maybe too
much flux or something?

I know it doesn't make sense. I wired up a test circuit with the gate
grounded. Just hooking up Drain and Source stops the capacitor from
charging.

Supply voltages, bias conditions,...?
Is your output bipolar WRT to minus input?
Is your 7002 still OK? (ESD)

 

[Joerg]

linnix a écrit :

linnix wrote:
linnix wrote:
Tim Wescott wrote:
[...]
It'd be interesting to see what linnix has to offer, if the site would
cough up more than a few lousy pictures.
Give me more time, please.
Ok, time's up now.
Just kidding ...
He said they'll have discussions, probably how much and when to release
info. Eagerly awaiting that ...
Right now, I am busy trying to solve the external integrator problem.
Our customer needs better closed-loop response with hardware
integrator. At the end of a cycle, I need a high resistant switch to
discharge the capacitor. The MOSFET I tried did not isolate well
enough. Mechanical relay works, but size is a problem. I need a
SOT-23-5 sized semiconductor relay. Any suggestions?
What's the leakage requirement? In many older analyzers this was done
with small JFETs. But if you only have +3V or something like this it'll
be more of a challenge.
It would be approx. 10K R and 0.01uF C with 3.5V charging and
discharging. If I connect a 2N7002 across the C, it wouldn't charge
at all. I guess the MOSEFT off-state leakage is larger than the 10K
R.
That's initially 350uA. A lot. The 2N7002 sits at 1uA for off. Maybe too
much flux or something?
I know it doesn't make sense. I wired up a test circuit with the gate
grounded. Just hooking up Drain and Source stops the capacitor from
charging.

Supply voltages, bias conditions,...?

Single 4V supply. Source of 2N7002 and negative side of capacitor at
virtual ground (minus input of LMV358). Drain of 2N7002 and positive
of cap to LMV358 output. Gate of 2N7002 to groung with/without 10K
resistor.

If I understand it right than this sounds ok. You just can't ever go
reverse because then the FET's body diode will kick in.

Tried some 2N7000 (TO-92). Same result.

Something is wrong there. But hard to say from far away.