Advice on isolated DVM design

[linux-dude]

Hi,

I've been tinkering with an idea for a long time; a PC based DVM with isolated test nodes; i.e. each channel is floating. Each channel should be able to do all the usual measurements like A/V AC/DC, Ohm etc. I can't count the number of times i wanted something like this in my lab...

My idea is something like
- Battery (or perhaps small transformers) to power each channel.
- Opto-isolated SPI communication to the main board.
- Isolation voltage good enough for 240V mains testing.
- Main board connects via USB to the PC.
- All settings controlled from the PC, i.e, no buttons on the nodes.
- Several channels, maybe like four to six, should possible.
- Precision should be 1% or better.
- Simplicity wins over saving a few bucks, but it should be reasonable.
- Common parts, easy to source.
- Open source - DIY's should be able to build it.

I have implemented all the digital and firmware parts above in other projects, but the DVM front-end part is new to me, and this is where i want some advices.

I've spent a few hours searching for DVM front-ends, and came up quite empty, except for designs using ICL7135 and the like (**broken link removed**). Though that is a possibility, i would like to explore other paths.

If anyone have some ideas, i would appreciate it much.

Thanks.

 

[chuckey]

One way to go would be to use a galvanic isolated instrumentation amp. This is a self contained chip that transfers an analogue voltage across a galvanic barrier rated up to 7.5 KV. One per channel , then a multiplexor to "switch" your analogue inputs from channel to channel which is synchronised with the display and your single ICL7135 A to D.
For decent accuracy you will duplication of the analogue range switching resistors, 1V, 5V, 10 V. . . A lot of expensive components.
Frank

 

[linux-dude]

One way to go would be to use a galvanic isolated instrumentation amp. This is a self contained chip that transfers an analogue voltage across a galvanic barrier rated up to 7.5 KV. One per channel , then a multiplexor to "switch" your analogue inputs from channel to channel which is synchronised with the display and your single ICL7135 A to D.

Interesting idea. I looked at some possible devices, and they were not that expensive. Not sure how accurate it would be, typically they claim 1% accuracy, but i suppose that it might not be quite that simple.

For decent accuracy you will duplication of the analogue range switching resistors, 1V, 5V, 10 V. . . A lot of expensive components.
Frank

Yes, and with additional relays/analog switches, it will cost a bit. Still, considering the gain with a floating design, i think it's worth it.

But this brings me back to square one; alternative front-ends besides the ICL7135? Any ideas?

 

[chuckey]

Which A to D, sorry can't help. I have been rethinking about your original suggestion just to see where the design would end. Your A to D needs +- 5V = two 9V batteries + linear regulators/channel or eight single cell batteries + two linear regulators - you could use re-chargeables. Or a SMPS with a toroid type core with N secondary windings + rect bridges + linear regulators. Special attention must be given to the insulation between the secondaries to get you galvanic isolation. I think this is the way to go, best bang for a buck, more so as more channels are required and relatively low power per channel.
So now you have to provide opto isolation for the digital output from your analogue PCBs. 5 data bits/channel +6 (?) control lines/channels = 11 bits per channel = 66 optos for 6 channels - loads of spaghetti here!!!
Frank

 

[linux-dude]

Which A to D, sorry can't help. I have been rethinking about your original suggestion just to see where the design would end. Your A to D needs +- 5V = two 9V batteries + linear regulators/channel or eight single cell batteries + two linear regulators - you could use re-chargeables. Or a SMPS with a toroid type core with N secondary windings + rect bridges + linear regulators. Special attention must be given to the insulation between the secondaries to get you galvanic isolation. I think this is the way to go, best bang for a buck, more so as more channels are required and relatively low power per channel.

Agreed, the SMPS for local voltage is easy and is cheap. The power transfer not that hard either, basicly a MOSFET driving 20KHz-100Khz through a small transformer. Isolation voltage is the only issue there.

So now you have to provide opto isolation for the digital output from your analogue PCBs. 5 data bits/channel +6 (?) control lines/channels = 11 bits per channel = 66 optos for 6 channels - loads of spaghetti here!!!
Frank

I intend to use SPI, with only 4 wires (CS, CLK, IN, OUT), via optoisolators. SPI is easy to use. I probably go with Microchip in both ends; they have integrated SPI - and i know them inside out =) On the other hand; I've done bitbanging before and it's simple. Might need to do that to get down in frequency as to lower optoisolator requirements/price. Update frequency could be like 5Hz, so 1000bps would be more than enough. That even the cheapest isolator can handle.

My idea was to match resolution from the A/D, whatever that will be. A dual slope A/D can easily provide 15-18 bits.

With that covered, I'm really lost with the front end. I spent more time searching for design ideas, but it's not much out there - or i'm looking at the wrong place? I found a couple of older schematics for analog instruments, but it feels quite outdated. Looks like I have to roll my own =)