Isolated GND'S idea

[Wizpic]

I came across this project and it looks good and like the idea how he measures the voltage with isolated ground's, It's a 2 channel PSU and he as used 2 GLD'S with one controller. This got me thinking about using the same idea using High-Linearity Analog Optocoupler(HCNR201) to measure upto 8 voltage's from different units and rather than use one unit for each unit I thought about doing it with one controller as I cannot common the grounds together.
Now I've not used these before and are quite costly so rather jump in I'd thought I would ask to see if anyone else as used them  as not sure how good accurate they would be or is there a better method, See attached
Let me know your thought's or recommendations
Project link
http://www.ianjohnston.com/index.php?option=com_content&view=article&id=92:powersupply&catid=3:hobbies&Itemid=8

[optikon]

that circuit has offsets, drift/degradation concerns, terrible linearity. The part says high linearity and for an opto-coupler it probably is, but for quality measurement, it stinks.


whats wrong with using a quality DMM for a good, highly isolated measurement?


define your requirements & explain what you are trying to do for meaningful suggestions.

[Vineyards]

I haven't tried above part. I had obtained samples of a similar IC from Avago a couple of years ago but I stashed them somewhere and now they are out of sight. The third LED is meant for compensation.

The circuit you mention does serve a purpose: ground-loop elimination especially when interfacing a high z sensor with a microprocessor. Of course you will need an isolated DC/DC converter, and an isolation amp such as ISO122 or 124. You could also do something similar (but probably more prone to drift) using V/F and F/V circuitry. Such a scheme will work for some motor control applications too.

[UncleBog]

Hello Wizpic,

A more accurate / expensive approach is to move the ADC across the isolation boundary so that only digital signals are transferred. Look at components like Analog Device's ADuMxxxx range which will allow you to connect to a serial ADC and provide an isolated supply for it. If a high sampling rate is required then you would have to isolate a parallel data bus plus control signals. I've attached part of a schematic which shows an isolated converter which allows a micro to monitor an isolated high voltage DC supply.

[Vineyards]

UncleBog, I think your approach is actually more cost efficient than the analog approach. I have used the analog method in the past and the parts involved are really very expensive. The linear optocoupler version is relatively new and seems suitable for general applications. Digital isolators (I believe there are also other even cheaper options) are much cheaper than the analog ones. There are varieties giving you different number of ports and data directions to conform with your specific application. The only additional cost item here is the standalone ADC. There are very reasonably priced ADC's on the market so that is not a big problem. So, your solution is also my favorite one.

[Wizpic]

I will have a look at drawing when I get home as I can't really see it on phone

[Old_but_Alive]

I agree about using digital isolators.

I am currently building an isolated JTAG adapter due to problems I have (see my recent posts)

I am using ADuM3473, which has its own internal power switcher, so all you need on the other side is a transformer, a SOT23 dual diodes, and a few filter components

check out the Wurth transformer 2-1 ratio p/n 760390015 its tiny, the diodes are a single SDM40E20LC-7 dual schottky 0.4A 20V SOT-23

[Sideshow Bob]

I remember from using the HCNR201 that the signal quality at best was about the same as a 12/13 bit ADC would have been. So I also second that a digital approach in our days is the best. A lot have changed since the HCNR201 was introduced. The HCNR201 is a quite old circuit design

[UncleBog]

The ADuM6401 in my schematic at #3 integrates the whole DC-DC switching regulator so that only smoothing capacitors are required externally. The downside is that the power available on the isolated side is limited to a few hundred mW. Old_but_Alive's ADuM3473 requires more external components but consequently can supply more isolated power. I designed that circuit 4 years ago and I can see that Analog Devices have added more digital isolators to their range, plenty to choose from.

[Wizpic]

whats wrong with using a quality DMM for a good, highly isolated measurement?

I could do but I want to be able log the data to sd card/computer, thought about 8 data loggers or 8 meters with usb output but not found any thing to suit really so I think it's better to design your own plus you get the buss of  designing it and a good  challenge.  

Unclebog thank I will have a look into that way.
 The idea is I want to measure 8 different voltage's from 8 different units to a main control room to monitor and set of alarms if threshold is reached either way (high/low). rather than having 8 separate units to look at just have one screen later on they might consider Bluetooth modules again having single unit make it easier and cheaper in the long run. Not sure if we will be using PIC24/32 or even the LPC1768 as this would be an on going development and more features could be added.
hope this make it a bit clearer in what I'm trying to do.

The ADuM6401 in my schematic at #3 integrates the whole DC-DC switching regulator so that only smoothing capacitors are required externally. The downside is that the power available on the isolated side is limited to a few hundred mW. Old_but_Alive's ADuM3473 requires more external components but consequently can supply more isolated power. I designed that circuit 4 years ago and I can see that Analog Devices have added more digital isolators to their range, plenty to choose from.

There will be no current required really as I only want to measure voltage up to 35 volts max it would be so easy IF we could join all the grounds together but this would be impossible as they would effect each other (already tried it  )

 

Posted on: January 08, 2014, 09:26:18 21:26 - Automerged
I've had a quick look at your way unclebog and the down fall I can see with that is I want to use 1 micro to measure the 8 channels at the same time not sure if it is possible to use 8 A/D converters with 8 SDA/SCLK to one micro to transfer the data.  Let's say the measuring board needs to convert the 0-35V signal to 0-5V one for each unit required and these will have there own 5V supply. then I  need to send that 0-5V some how (isolated) to one master unit using the micro ADin inputs. So would the HCNR201 chap be better if it give 12/13bit's as the micro A/D is either 10bit or 12bit. accuracy can be 50-100milli volts either way. I've had a quick look at the digital isolators but my understanding is they only transfer logic signal am I correct ?
This is looking like to be a bigger challenge than I first thought 

Thanks to everyone for all the good ideas and they look very interesting and my suit other applications

[UncleBog]

Wizpic,

If you choose an ADC with a chip select (like the one in my schematic) then you need one SDA / SCL pair plus eight chip select signals. You could drive these directly from microcontroller pins, or use a 74xx138 decoder function to save some pins. If you find an ADC with an I2C interface then the addressing is embedded in the serial data and you only need SDA and SCL.

However all this may be irrelevant if you need to have widely separated sensors. Bluetooth is pretty good isolation.

[Vineyards]

You could also build a single input circuit and use a mux with which to switch between input signals. You could use a port expander to control the mux inputs going through and digital isolator from your microprocessor.

When done like that you could add some fancy extras like multipole active filters, surge protection etc to your input circuitry. After all you don't have to repeat the same circuit multiplied by the number of inputs.

[h0nk]

Hello Wizpic,

You should also consider the Analog Devices Isolated Sigma-Delta AD's (e.g. AD7400).
They include a transformer on-chip for the Sigma-Delta data signals but not for the supply.
The AD7400 requires 12 mA @ 5 V on the input side.

Since they output the raw Sigma-Delta data a fast processor or FPGA is needed to compute
the results. One decimation filter of one channel needs approx. 8 % of the smallest Altera Cyclone1
EP1C3, so 8 channels will fit for sure.

Best Regards