1
\$\begingroup\$

I need to power a 0-5VDC output ratiometric-sensor from 75 meters far away.

But because the sensor output signal depends on the stability and/or the accuracy of the power supply voltage(because of being ratiometric), I'm recommended to set up a local, well-regulated supply closer to the sensor. But I will not have access to AC mains on the sensor side, so I cannot set up a local supply.

How about using an isolated DC-DC power supply at the sensor end as shown in Figure 2 below?(Figure 1 shows powering directly from far away):

enter image description here

Does Figure 2 provide a better regulation and accuracy for the sensor excitation? If not, what could be a adequate way to achieve this? Does anybody have experience with such scenario?

\$\endgroup\$
  • \$\begingroup\$ Figure 2 will work much better for your application, but be careful about the efficiency of the converter, so you don't put excess drain on your battery. But also, where is the reading of your sensor going? All the way back down the 75 meter cable? Is it being sent as an analog voltage or a digital reading? \$\endgroup\$ – Jim May 31 '18 at 18:23
  • \$\begingroup\$ Battery?? I was not planning to use a battery. What do you mean? They want analog transmission(long story) so I will use a voltage-current converter and send the signal as 4-20mA back. Using somthing like this module: docs-emea.rs-online.com/webdocs/15aa/0900766b815aa3c7.pdf \$\endgroup\$ – GNZ May 31 '18 at 18:25
  • \$\begingroup\$ Sorry thought I saw battery somewhere but must have mis-read. \$\endgroup\$ – Jim May 31 '18 at 18:30
  • \$\begingroup\$ There are many DC-DC converters SMPS and linear; Isolated non-isolated; and which parameter indicates the regulation quality? I also am not sure 9-->5V is adequate or excessive. Thanks \$\endgroup\$ – GNZ May 31 '18 at 18:32
  • \$\begingroup\$ Another option to consider might be kelvin contacts. We commonly use that for ratiometric sensors on the end of long cables. Don't underestimate the advantages of decent (i.e. low resistance, shielded and/or twisted pair) cables too. \$\endgroup\$ – Jack B May 31 '18 at 18:56
1
\$\begingroup\$

I might agree with Maple that it could be better to use a digital connection but if you use analog there are some things you must take into account:

In both cases (figure 1 and figure 2) you don't have a ratiometric output because you don't have access to one therm of the ratio , the power supply voltage at the sensor side. In the figure 2 it's obvious why , in the figure 1 due the unknown line voltage drop.

I also don't see the reason to have an insulated DC-DC converter, what's wrong with a non insulated one in this configuration? In fact an insulated one would need to be a switched mode one that adds a lot of noise right in the proximity of the sensor.For the 3mA needed you can use a linear stabilizer.

From this point of view the solution in the figure 1 is better.

To really have a ratiometric output you will need to have two more wires for signal ground and signal Vcc that don't have the supply currents passing through

I would rather use this setup instead , at 3mA the voltage drop can be at most of tens of mV order, and do not alter the signal since the supply current goes through different wires. Also the noise from the DC-DC converter is far form the sensor side.

schematic

simulate this circuit – Schematic created using CircuitLab

You can also put a linear stabilizer at the sensor side if needed

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ 1-) This is a great point, here is an article about it:analog.com/en/analog-dialogue/articles/…. How to make correction?: Lets the signal in your diagram is measured as 2V but the 5V supply is measured 4V(1V supply voltage dropped)..This means if we relate the pressure to 2V it is an error. Should we correct it this way 5*(4/2) = 2.5V. So we should measure the pressure as if it is 2.5V sensor output. Is that correct? \$\endgroup\$ – GNZ Jun 1 '18 at 0:05
  • \$\begingroup\$ 2-) For around 70 meters I can use STP cables. Do you think if I send the signal back as current by a module? Or should I first try the voltage way? \$\endgroup\$ – GNZ Jun 1 '18 at 0:06
  • \$\begingroup\$ 3-) What do you mean by insulated power supply? Galvanically isolated SMPS or a linear 5V supply? And "linear stabilizer at the sensor" do you mean a voltage regulator? \$\endgroup\$ – GNZ Jun 1 '18 at 0:07
  • \$\begingroup\$ @Genzo 1) Yes it's correct. The document explains very good pluses and minuses for various solutions. 2) the current output is much resisttant to noise but you will loose ratiometric output and you will need to use a precision stabiliser at the sensor side. Also look if errors added by U/I converter are avceptable. An advantage is that you can use a shared ADC for many sensord. 3) Either SPMS or linear (transformer rectifier and regulator) . Yes, voltage regulator. \$\endgroup\$ – Dorian Jun 1 '18 at 6:58
0
\$\begingroup\$

Ta-da! You've just invented POL supply! Congratulations. :)

I believe this is how chipsets on the motherboards are powered for the past 10 years or so. Providing you can make/buy DC-DC stable enough for your application Figure 2 is good way to go.

Now, the feedback from your sensors makes me worried. I think digital connection will be much more precise and reliable. Ideally with differential protocol. For example, if your sensor chip has I2C interface you can use something like AN11075 to send it over RS-485 more than 1 km away without any loss of precision.

Here is a ready-made product you might find interesting. It extends I2C up to 300m and.. wait for it... it already has DC-DC on board. Therer are other modules over there, some feeding power into the line, others using it to power themselves.

As you can see, they support true bus architecture, with one Boost module providing power and data link to multiple Buck modules. You can put it in the box with micro-controller, add several (again, ready-made) DAC modules and then write simple program that interrogates sensors and puts their output to DACs in a loop. This is actually one evening project, come to think of it. For the cost of under $300 per 10 sensors.

UPDATE: Here is an illustration of possible system configuration:

enter image description here

DACs: you can use individual modules, like MCP4725, or go crazy with 16 channel ADAU1966A monster (cost is not an issue, right? ;) )

MCU: Arduino, Raspberry or whatever you feel more comfortable with

Boost module: The one I linked before or any other ready-made I2C-to-differential converter.

Buck module: The complementary device to above or a combination of DC-DC module + I2C-to-differential converter.

Amplifier: ADC with I2C interface and desired resolution, like ADS1115. Or better yet, the link that you provided mentions an option for I2C output straight from the sensor. That would simplify things a lot.

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ 1-) Thanks a lot. But for some reason they want analog that's why I thought to convert to current. What could be wrong if I convert and send as current? I was thinking to calibrate out the DC offsets. \$\endgroup\$ – GNZ May 31 '18 at 20:15
  • \$\begingroup\$ 2-) But in case if I can convince them for serial communication, I need a module which can convert I2C or SPI to RS485. Because Im not sure I can make the PCB quickly and a mistake will cost me a lot. The converter should be very robust professional and work down to -20C. Do you think I can find such ready converter module(SPI or I2C to RS485)? Cost is not an issue \$\endgroup\$ – GNZ May 31 '18 at 20:16
  • \$\begingroup\$ Well, if cost is not an issue you can go with something like $200 3RS7000, sure. (I never used anything like that, so can't really comment on it). But about 4 of those would probably cover pre-manufacturing cost of AN11075 or PCA9615 based converters. With about $5 per actual unit afterwards. But if your customers want analog... well, they are always right, aren't they? \$\endgroup\$ – Maple May 31 '18 at 20:53
  • \$\begingroup\$ Another way would be Im thinking to convert(if I can find a module) from IC2/SPI to RS485 and at the receiver end convert back to analog voltage by another module. I have no prob with cost since it is part of a huge project. My problem is finding modules. What do you think about converting to serial and back to analog at the daq side? Then they would have analog inputs with less noise? My worry would be delay more than 50ms. \$\endgroup\$ – GNZ May 31 '18 at 20:59
  • \$\begingroup\$ I added link to pre-made product working at 1MHz. Should be enough to not worry about delays ;) Converting that back to analog is trivial, if that's what customer really wants. \$\endgroup\$ – Maple May 31 '18 at 21:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.