8 channel current source solution for RTD sensors

[engr_joni_ee]

Hi, I am looking for 8 channel current source 1 mA for RTD temperature sensors. One option is to use 8 current source chips on the board but eight chips make the board messy. I am wondering if there is a current source chip with eight channels.

 

[KlausST]

Hi,

8 BJTs with 8 emitter resistors ...and a constant voltage...

Or one 1mA source and a MUX, if you use a MUXed ADC..

Klaus

 

[danadakk]

This is an example project for 1 channel. The muxes can be expanded to ~ 16 - 24 channels
if needed to support more sensors. The muxes are right clicked, # channels entered, and then
pins are dragged out of component catalog and wired up to muxes using wiring tool wizard
in PSOC Creator. This is a single chip solution. Note resources used/left in right hand window
show a lot more onchip functions can be used in the design, like display, digital, LUT, logic,
counters, timers, pwm, mux/demux, opamps, comparators, COM, USB, Serial...... Of course code
would be modified to handle the additional channels. This is just one of 100 example projects in
Creator.

Note dashed lines are simply showing offchip components needed, eg. the RTDs, the rest is
on one chip. Also the SAR as shown is 12 bits, also onchip is a 20 bit DelSig if you need more
accuracy/resolution. And a .1% Vref is onchip as well. You can also add DMA to design so that
all measurements done in HW are sent out to USB or serial (SPI, USART, I2C) w/o code inter-
vention. The DMA also onchip. And using the onchip DSP filter the samples to eliminate most
noise if needed.

A catalog of onchip components attached. In PSOC language a component is an onchip resource.

IDE (PSOC Creator) and compiler free, board to do this is ~ $ 10. For 8 channels, for more a different
board would be needed.

https://www.cypress.com/products/psoc-creator-integrated-design-environment-ide

https://www.cypress.com/documentation/development-kitsboards/cy8ckit-059-psoc-5lp-prototyping-kit-onboard-programmer-and

Note this example was done with low end part, PSOC 4, I recommend doing it in
PSOC 5LP, much more capability for other features.


Regards, Dana.

 

[BradtheRad]

One current mirror can drive many duplicates. Just add another transistor for each additional channel.

 

[FvM]

You should urgently mention the intended temperature and respective current accuracy. E.g. 0.3 K requires about 0.1% current accuracy. Some solutions proposed in this thread are not suited for accurate temperature measurements.

 

[engr_joni_ee]

Hi,

The requirement is to read temperature between 0-100 C with 0.2C precision. The data should be read every 30 sec from all 8 RTD temperature sensors. I will use PIC microcontroller with 10 bits ADC.

I am wondering which of the two solutions would be the best.

1- 8 BJTs with 8 emitter resistors having 0.1% tolerance

2- One 1mA current source 0.1 % accuracy with a MUX/DeMUX that select/switch every two second to another temperature sensor

 

[FvM]

Look at the expectable Vbe and B variation of BJT and realize that solution 1 is effectively impossible, even with massive degeneration.

As for solution 2, 0.1% current accuracy isn't enough, but you'll most likely perform adjustment and rather look for drift than absolute accuracy. 0.2 K precision may be a problem with regular RTD.

 

[engr_joni_ee]

Hi,

If I use emitter resistor with 1 % tolerance in solution 1, and in solution 2 if I use 0.1 % current accuracy source then how much accuracy in temperature measurement should I expect in each solution. I am also planing to calibrate them to adjust the linear drift if it happen. How about precision of 0.3 or 0.4 C ? Is it possible to get from the solutions mentioned ?

 

[danadakk]

You could use the DelSig in the PSOC and a 16 bit PWM in PSOC to gen a 16 bit
V DAC, use a .01% R in a closed loop to calibrate a precision V to I current source
built out of onchip PSOC OpAmp and external discrete like JFET or PNP to gen up
the I source. Your 10 bit ADC on PIC not accurate enough to do this. Just a thought.
Whole solution still single chip.

The one part that may be needed externals is a << .1% Vref once you add up all
the T and V and G and INL and PSRR, mux linearity, and noise errors. Note the
onboard DelSig is 20 bits. The onboard Vref can be caled to .1%, this may not be
enough.;

Note the DelSig is fast enough at 20 bits to do all your channels at the rate you
want, do the Isource cal as well, and has advantage its CM range is outside the
rails by 100 mV.

This discusses some of that - https://www.cypress.com/file/123806/download
It also refers to caling onchip Vref for better accuracy.

https://www.cypress.com/CE219929

A useful classic paper on high precision, attached.

An alternative approach, used in very high accuracy instrumentation is DUT cal in
manufacture test. Basic approach looks like -

This was a Raytheon patent back in 1960's if my memory serves me which led to all the high res HP and others
instrumentation.


Regards, Dana.