On Mon, 17 Mar 2008 14:06:35 +0100, lowcost <die.spam@invalid.com>
wrote:
>Robert Sneddon ha scritto:
>> Low cost is important in this project;
>
>here i am, my 2cents
>
>> (liquid bath temperature control in the range 35 deg C - 85 deg C).
>
>for such a narrow range NTC probes are useful: cheap, accurate and
>easily sampled by uC onchip ADC, without any expensive precision analog
>conditioning.
>
>> We'd like to bring this down below 50 bucks for the controller and sensor
>
>yes you can. looking at wall boiler controller or the like you will find
>similar product (probes+PID+front_end+actuators) well below 50 (parts cost).
>
>> We're not fixated on thermocouples but we do need +/- 0.5 deg C
>> accuracy without drift over time.
>
>mmmm, +- 0.5C onto K type thermocouples are +- 20uV , 0.2% for RTD.
>both need precision analog conditioning.
>regards
With thermocouples you also have to measure (and assure isothermal
conditions under operation conditions) a local temperature within that
error budget. So, if you're trying to do 0.5�C you might have to
measure the block temperature to within +/-0.2�C and have no more than
500nV/K drift and a few uV Vos to get close to your spec over a
moderate Ta range.
RTDs are relatively easy.
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Reply by lowcost●March 17, 20082008-03-17
Robert Sneddon ha scritto:
> Low cost is important in this project;
here i am, my 2cents
> (liquid bath temperature control in the range 35 deg C - 85 deg C).
for such a narrow range NTC probes are useful: cheap, accurate and
easily sampled by uC onchip ADC, without any expensive precision analog
conditioning.
> We'd like to bring this down below 50 bucks for the controller and sensor
yes you can. looking at wall boiler controller or the like you will find
similar product (probes+PID+front_end+actuators) well below 50 (parts cost).
> We're not fixated on thermocouples but we do need +/- 0.5 deg C
> accuracy without drift over time.
mmmm, +- 0.5C onto K type thermocouples are +- 20uV , 0.2% for RTD.
both need precision analog conditioning.
regards
Reply by Steve at fivetrees●March 16, 20082008-03-16
"Robert Sneddon" <fred@nospam.demon.co.uk> wrote in message
news:3v96ZGLu0Y3HFwby@nospam.demon.co.uk...
>
> The idea is to roll a neutral set of characteristics for the bath
> controller to preventing ringing, hunting and overshoot -- with no
> active cooling system for the bath it takes a long time for the temp
> to
> recover from an overshoot, one reason we're considering mineral oil or
> something with a lot less thermal capacity than water. If it takes
> longer to reach the target temperature but doesn't overshoot that
> doesn't matter in user terms. Overshoot in operation is *the* major
> worry as that will cause Bad Things to happen.
Quick reply: if you're worried about overshoot, but not about time to
reach target, you might want to consider reducing (or at least
tailoring) the heater power such that significant overshoot is hard to
do. Ideally, you want to be capable of delivering just enough power to
overcome losses, plus a reasonable margin, under worst case conditions.
Alternatively, some controllers have an output power limit, which can
achieve much the same thing.
You've probably already considered this; but I have seen cases of
massively overpowered heaters resulting in huge overshoots... esp.
during tuning.
Just a thought.
Steve
--
http://www.fivetrees.com
Reply by Robert Sneddon●March 16, 20082008-03-16
In message <yI2dnRXQQJ25lkDanZ2dnUVZ8tqinZ2d@pipex.net>, Steve at
fivetrees <steve@NOSPAMTAfivetrees.com> writes
>The temperature range you quote is the so-called "organic" range, and
>you may find it easier to use an RTD (aka PT100) sensor. These sensors
>*may* be a tad pricier than a type K t/c, but they are easier to read,
>either in 3-wire or 4-wire configurations, and you don't need a cold
>junction (but you do need a stable current source). Also the accuracy
>you're after may well be easier with an RTD.
We thought about Pt100 sensors but they're intrinsically more expensive
than "simple" thermocouples -- we may have to supply sets of different
sensors built for different physical layouts within the bath. Worst case
we roll our own sensors using regular twin-core K-series thermocouple
wire, insulate them and encapsulate them in thin-walled copper or
stainless tubing to prevent water ingress and corrosion. Other
possibilities include thermistors but they have their own problems like
individual calibration, drift, insensitivity etc. Nothing we can't fix
but it's more work again.
>
>Re cost: I've designed temperature controllers (albeit for high-volume
>production) with way tighter budgets than that ;). Have you considered
>some of the 1/16th DIN controllers from e.g. CAL or West Instruments?
We're using a DIN PID module at the moment in the prototype, but unless
you know of anyone selling such modules under fifty bucks we're screwed.
Best price we've been quoted for small quantities is about a hundred
bucks each, and more typically they're about 150-200 bucks. They're also
over-complex for naive users to configure, with too many things that can
be programmed wrong for the job it's meant to do.
>Also, have you considered how you'll amortise the development costs? A
>PID algorithm is not rocket science, but there are a number of wrinkles
>(e.g. integral wind-up etc etc) that newcomers tend to have to learn the
>hard way...
Yah, one reason we prototyped using an off-the-shelf PID to start with.
There's a secondary R&D project running with the existing prototype
which we're also planning to make some money out of but we're looking to
sell units to end-users as well. For those end-users PID *is* rocket
science, just about. I've already had to reprogram the PID once after my
assistant managed to disable the PID autolearn function which we were
relying on to deal with variable masses of water in the bath. This was
after he managed to lose the programming instructions which were written
in Janglish anyway.
The idea is to roll a neutral set of characteristics for the bath
controller to preventing ringing, hunting and overshoot -- with no
active cooling system for the bath it takes a long time for the temp to
recover from an overshoot, one reason we're considering mineral oil or
something with a lot less thermal capacity than water. If it takes
longer to reach the target temperature but doesn't overshoot that
doesn't matter in user terms. Overshoot in operation is *the* major
worry as that will cause Bad Things to happen.
--
To reply, my gmail address is nojay1 Robert Sneddon
Reply by Steve at fivetrees●March 16, 20082008-03-16
"Robert Sneddon" <fred@nospam.demon.co.uk> wrote in message
news:sQKMmTGw6E3HFwaa@nospam.demon.co.uk...
>
> or b) a pre-built single-board computer (SBC) with on-board circuitry
> for interfacing a K-type thermocouple.
Unlikely.
> Low cost is important in this project; our current proof-of-concept
> prototype cost us about 300 bucks in parts and is too big physically
> to
> do the job we want (liquid bath temperature control in the range 35
> deg
> C - 85 deg C). It's also too complex to use in the real world. We'd
> like
> to bring this down below 50 bucks for the controller and sensor (which
> must be submersible) and simplify the front-end controls a lot.
>
> We're not fixated on thermocouples but we do need +/- 0.5 deg C
> accuracy without drift over time. The info on the Dallas DS1620
> suggests
> it could do the job as a sensor/digitiser but we'd have to encapsulate
> it to make it submersible in fluid (water, mineral oil etc.).
> Thermocouples are available off-the-shelf already packaged and proven
> for this sort of application.
The temperature range you quote is the so-called "organic" range, and
you may find it easier to use an RTD (aka PT100) sensor. These sensors
*may* be a tad pricier than a type K t/c, but they are easier to read,
either in 3-wire or 4-wire configurations, and you don't need a cold
junction (but you do need a stable current source). Also the accuracy
you're after may well be easier with an RTD.
Re cost: I've designed temperature controllers (albeit for high-volume
production) with way tighter budgets than that ;). Have you considered
some of the 1/16th DIN controllers from e.g. CAL or West Instruments?
Also, have you considered how you'll amortise the development costs? A
PID algorithm is not rocket science, but there are a number of wrinkles
(e.g. integral wind-up etc etc) that newcomers tend to have to learn the
hard way...
Steve
--
http://www.fivetrees.com
Reply by Robert Adsett●March 15, 20082008-03-15
In article <gYmdnQWQ6dJkxUHanZ2dnUVZ_uKpnZ2d@comcast.com>, donald
says...
> PS: There are no single chip MCUs with a built in thermocouple input.
> But there is a app note from TI that can help you: slaa216.pdf.
> In message <47DBF464.8040409@yahoo.com>, Freelance Embedded Systems
> Engineer <g9u5dd43@yahoo.com> writes
>> Robert Sneddon wrote:
>>> I'm looking for a small embedded controller module which can interface
>>> directly to a thermocouple (K-type, I think)
>
>>> We figure PIC, 8051 or similar will be smart enough to do the job. Can
>>> anyone recommend something off-the-shelf that would suit us?
>> Regarding the thermocouple input:
>>
>> The MSC1201 is a 8051 MCU and has 16 digital i/o pins
>> http://focus.ti.com/docs/prod/folders/print/msc1201y2.html
>
> The actual model of CPU isn't that important other than it should have
> about 16 I/O pins. There's no serious computational load required for
> this task.
>
>> Regarding the LCD,
>
> I want to use 7-segment LEDs as they will have to be readable from a
> metre or two in variable light conditions.
>
> I was hoping to find:
>
> a) (long shot) a single-chip MCU with integral thermocouple interfacing
> that would require minimal design overhead (i.e. just connect the
> thermocouple wires to two of the MCU's pins).
>
> or b) a pre-built single-board computer (SBC) with on-board circuitry
> for interfacing a K-type thermocouple.
>
> Low cost is important in this project; our current proof-of-concept
> prototype cost us about 300 bucks in parts and is too big physically to
> do the job we want (liquid bath temperature control in the range 35 deg
> C - 85 deg C). It's also too complex to use in the real world. We'd like
> to bring this down below 50 bucks for the controller and sensor (which
> must be submersible) and simplify the front-end controls a lot.
>
> We're not fixated on thermocouples but we do need +/- 0.5 deg C
> accuracy without drift over time. The info on the Dallas DS1620 suggests
> it could do the job as a sensor/digitiser but we'd have to encapsulate
> it to make it submersible in fluid (water, mineral oil etc.).
> Thermocouples are available off-the-shelf already packaged and proven
> for this sort of application.
Oh ya, $50 may be parts cost, but I doubt you'll see a $50 complete unit.
donald
Reply by donald●March 15, 20082008-03-15
Robert Sneddon wrote:
> In message <47DBF464.8040409@yahoo.com>, Freelance Embedded Systems
> Engineer <g9u5dd43@yahoo.com> writes
>> Robert Sneddon wrote:
>>> I'm looking for a small embedded controller module which can interface
>>> directly to a thermocouple (K-type, I think)
>
>>> We figure PIC, 8051 or similar will be smart enough to do the job. Can
>>> anyone recommend something off-the-shelf that would suit us?
>> Regarding the thermocouple input:
>>
>> The MSC1201 is a 8051 MCU and has 16 digital i/o pins
>> http://focus.ti.com/docs/prod/folders/print/msc1201y2.html
>
> The actual model of CPU isn't that important other than it should have
> about 16 I/O pins. There's no serious computational load required for
> this task.
>
>> Regarding the LCD,
>
> I want to use 7-segment LEDs as they will have to be readable from a
> metre or two in variable light conditions.
>
> I was hoping to find:
>
> a) (long shot) a single-chip MCU with integral thermocouple interfacing
> that would require minimal design overhead (i.e. just connect the
> thermocouple wires to two of the MCU's pins).
>
> or b) a pre-built single-board computer (SBC) with on-board circuitry
> for interfacing a K-type thermocouple.
>
> Low cost is important in this project; our current proof-of-concept
> prototype cost us about 300 bucks in parts and is too big physically to
> do the job we want (liquid bath temperature control in the range 35 deg
> C - 85 deg C). It's also too complex to use in the real world. We'd like
> to bring this down below 50 bucks for the controller and sensor (which
> must be submersible) and simplify the front-end controls a lot.
>
> We're not fixated on thermocouples but we do need +/- 0.5 deg C
> accuracy without drift over time. The info on the Dallas DS1620 suggests
> it could do the job as a sensor/digitiser but we'd have to encapsulate
> it to make it submersible in fluid (water, mineral oil etc.).
> Thermocouples are available off-the-shelf already packaged and proven
> for this sort of application.
A few things I would ask in a project like this.
1) How many are we talking about here ?
1, 10, 100 units
2) What is the life for these units ?
Should last 10 years, 50 years ?
3) What are the other 15 port pins used for ?
I am sure one pin controls the power to the heater,
but what are the rest for ?
donald
PS: There are no single chip MCUs with a built in thermocouple input.
But there is a app note from TI that can help you: slaa216.pdf.
Reply by Robert Sneddon●March 15, 20082008-03-15
In message <47DBF464.8040409@yahoo.com>, Freelance Embedded Systems
Engineer <g9u5dd43@yahoo.com> writes
>Robert Sneddon wrote:
>> I'm looking for a small embedded controller module which can interface
>> directly to a thermocouple (K-type, I think)
>> We figure PIC, 8051 or similar will be smart enough to do the job. Can
>> anyone recommend something off-the-shelf that would suit us?
>
>Regarding the thermocouple input:
>
>The MSC1201 is a 8051 MCU and has 16 digital i/o pins
> http://focus.ti.com/docs/prod/folders/print/msc1201y2.html
The actual model of CPU isn't that important other than it should have
about 16 I/O pins. There's no serious computational load required for
this task.
>Regarding the LCD,
I want to use 7-segment LEDs as they will have to be readable from a
metre or two in variable light conditions.
I was hoping to find:
a) (long shot) a single-chip MCU with integral thermocouple interfacing
that would require minimal design overhead (i.e. just connect the
thermocouple wires to two of the MCU's pins).
or b) a pre-built single-board computer (SBC) with on-board circuitry
for interfacing a K-type thermocouple.
Low cost is important in this project; our current proof-of-concept
prototype cost us about 300 bucks in parts and is too big physically to
do the job we want (liquid bath temperature control in the range 35 deg
C - 85 deg C). It's also too complex to use in the real world. We'd like
to bring this down below 50 bucks for the controller and sensor (which
must be submersible) and simplify the front-end controls a lot.
We're not fixated on thermocouples but we do need +/- 0.5 deg C
accuracy without drift over time. The info on the Dallas DS1620 suggests
it could do the job as a sensor/digitiser but we'd have to encapsulate
it to make it submersible in fluid (water, mineral oil etc.).
Thermocouples are available off-the-shelf already packaged and proven
for this sort of application.
--
To reply, my gmail address is nojay1 Robert Sneddon
Reply by ●March 15, 20082008-03-15
Robert Sneddon <fred@nospam.demon.co.uk> writes:
> which can interface directly to a thermocouple (K-type, I think)
Looking only at the thermocouple...
You can use pretty much any of the Maxim "battery monitor" ICs to
monitor a thermocouple. The key is that the battery monitors have an
ADC that's designed for current sensing, and has a very small (like
uV) resolution. The chip also has a temperature sensor for "cold
side" sensing, making it ideal for a cheap thermocouple interface.
Parallax sells a kit using a DS2760, for example, which I'm using to
monitor my woodstove from the furnace computer. If you stick to the
1-wire interface, it only uses up one GPIO on whatever MCU you choose
to use.
DS2760 - $3
R8C/26 - $5
http://www.sensorsmag.com/articles/0102/29/main.shtmlhttp://www.delorie.com/house/furnace/