Signal level shifting and rescaling

Thread Starter

Strike-the-root

Joined Apr 10, 2019
28
I'm working on a battery-powered piece of instrumentation and ran into a problem that so far has resisted my efforts.

We're using an AD736 to measure the RMS voltage of a signal. The 736 needs a symmetric supply, so it's powered from the 0V and 12V battery terminals and the "ground" at midpoint (6V) is established with a voltage divider that has ample overhead to meet the needs of this circuit. The RMS output signal from the 736 ranges from 0-1Vdc and is referenced to this midpoint 6V "ground". Therefore, with respect to the negative battery terminal. the RMS signal ranges between 6 and 7Vdc.

I need to feed this signal into a 0-5V ADC which is powered from a separate supply but must have its 0V reference tied to the negative battery terminal.

Usually, I just use op-amp sum or subtraction circuits, but this time I'm running into problems because I'm trying to shift the level by 6V--from the midpoint "ground" level down to the bottom rail. I wouldn't be surprised if I have to give up some resolution on my ADC, but it'd be handy to hear if there are some well-known solutions for this.
 

KeithWalker

Joined Jul 10, 2017
1,911
Is it absolutely necessary to tie the two supply negatives together? It would make things much simpler if you could tie the positive rails together an use the 5V- as ground for both. The whole system would benefit from having a good, solid ground reference. There is nothing in the spec of the AD736 that states that the dual supply has to be exactly symmetrical, in fact, to the contrary because they specify +2.8 V, −3.2 V as the minimum dual supply voltages.
 

AnalogKid

Joined Aug 1, 2013
9,351
To save both of us 20 questions, please post a basic wiring diagram showing the various power rails, grounds, signal references, and the interconnection between the RMS and ADC systems.

"Engineers don't take a dump, son, without a schematic."

ak
 

Irving

Joined Jan 30, 2016
2,058
While @KeithWalker solution works it could potentially become a maintenance/support nightmare if your system ground is actually 6v above the -ve battery terminal. You'd have to ensure that anyone working on it with 'grounded' test gear such as a 'scope ensured the system under test was floating or the scope was battery powered or fed from an isolation transformer.

It would have been better to have generated a -ve rail for the AD736 with, say, a 7660 voltage inverter, then you could have maintained a solid ground reference throughout.

That would still be my preferred option, but if you can't retrofit then a couple of other options are:
An isolated opamp
An analog sample & hold level shifter.

However with all these options, unless you provide a small -ve rail you're going to have to offset the ADC 'zero' and lose some dynamic range. I'd advise against this as its another calibration issue due to not having a solid ground reference.
 

AnalogKid

Joined Aug 1, 2013
9,351
It would have been better to have generated a -ve rail for the AD736 with, say, a 7660 voltage inverter, then you could have maintained a solid ground reference throughout.
Agree. And a much lower cost than a xx-bit-accurate voltage reference plus a level-shift circuit. And it solves the where-do-I-clip-the-scope-ground question.

And it wouldn't have to be a power supply chip. The part draws less than 1 mA, and almost anything can drive a charge pump for that - opamp, logic gate, <choke> 555 ...

ak
 
Last edited:

Thread Starter

Strike-the-root

Joined Apr 10, 2019
28
Thanks for all the replies. It looks like I made a few fundamental mistakes by breadboarding this circuit without a coherent design plan. While it will involve some backtracking, I can rework the circuits operating on 0-12Vdc to operate on 0-5Vdc which will solve the problem. Lesson learned here.
 

crutschow

Joined Mar 14, 2008
27,697
Yes, the minimum is 6V with a pseudo ground.
My battery has been faulty for a number of years now.
 
Last edited:

KeithWalker

Joined Jul 10, 2017
1,911
While @KeithWalker solution works it could potentially become a maintenance/support nightmare if your system ground is actually 6v above the -ve battery terminal. You'd have to ensure that anyone working on it with 'grounded' test gear such as a 'scope ensured the system under test was floating or the scope was battery powered or fed from an isolation transformer.
Why do you consider it necessary to use the 12V negative as ground? It is of absolutely no consequence where the ground reference is made in a circuit powered by batteries. Test gear should use the same ground as the circuit for reference. There is no need for isolation transformers.
I think your mind is stuck on popular convention rather than practicality.
This is what I suggested as the simplest solution:

ad736.jpg
 
Last edited:

Irving

Joined Jan 30, 2016
2,058
@KeithWalker On the grounds of simplicity I'd agree with you - If, and only if, the TS's use case is that battery operation throughout is mandated at all times.

However the TS said " I need to feed this signal into a 0-5V ADC which is powered from a separate supply but must have its 0V reference tied to the negative battery terminal "

My experience tells me that "separate supply" leaves the possibility that things can go badly wrong when simple 'obvious' assumptions are made about what constitutes 'ground'.

At the end of the day its the TS's choice.
 

KeithWalker

Joined Jul 10, 2017
1,911
In that case, TS is trying to solve the problem fro the wrong end. The circuit would be much simpler and give more predictable and accurate results if he replaces his 5V supply rather than mess around with the added complexity of level shifters.
A negative, adjustable, three terminal regulator chip set to -5V using couple of resistors could replace his 5V supply and solve the whole problem.
 

Thread Starter

Strike-the-root

Joined Apr 10, 2019
28
I reconfigured things today using +5V and -5V rails with the 12V battery negative terminal connected to the 0V midpoint. Now everything works great. I should have seen the problems coming, but I appreciate you all pointing out how kluged my approach was. I was just digging myself into a deeper hole without the -5V. I'm using a bench supply to deliver the +/-5Vdc right now but ordered some 7660 and LT1054 charge pumps from Digikey to deliver the negative rail voltage in the actual device. Look forward to trying those out next week.
 
Top