Auto-ranging with ICL7135

[Greg Robinson]

Hi there everyone.

I'm hoping to make an auto-ranging voltmeter using the Renesas ICL7135 4 1/2 Digit, BCD Output, A/D Converter.
Now, my experience with logic/digital is very limited, I almost exclusively work with analog, so, I was hoping that there would be an application note which would demonstrate an example of an auto-ranging circuit. Unfortunately, the app note for this chip, AN017 does not contain this information, and neither does the datasheet.

However, I did find an app note for an older/obsolete part, AN028, which does show an autoranging circuit (page 4), and it should be completely compatible - it's just using the "under-range" and "over-range" outputs, which the ICL7135 shares.
However, there appears to be a number of issues with this reference circuit.
There's an AND gate labelled 7403, but a 7403 is a NAND, another AND labelled 7406 which is an inverter, a NOR labelled 7462 which is I don't know what, and another NOR labelled 7482 which is an adder.
Basically, I can't make heads-or-tails of it.

I emailed Renesas asking about the discrepancies I'd found, and was at first brushed off and told that it was an obsolete part. Once I mentioned that I was applying the information to a current part because the information wasn't presented elsewhere (that I could find), I got a bit of information.
Was told the 7482 "NOR" should be 7428. The four numbered AND - one labelled 7403, the other three labelled 7406, should all be 7406 NAND.
The 7462 should be 7402.

Now, that all seems to make sense to me, but as I said, I'm not at all confident with my logic skills. Do these corrections make sense to the people here? Should this get me a functional autoranging circuit?

And one last question. Both the application notes and the datasheet all suggest using a trimpot for trimming the 1V ref voltage, using a zener or 1.2V reference IC. I would think that an accurate 1V reference IC like ADR130 with no trim would be a much better idea, or is there some potential pitfall I'm overlooking?

[Zero999]

It's full of errors. One of the J-FETS is mislabelled as 2N7002 MOSFET.

The 7406 doesn't make any sense. It's an inverter with open collector output.
http://www.ti.com/lit/ds/symlink/sn5406.pdf

Assuming the symbols in your attachment are correct. The NOR gates should be 7402 and NAND gates 7400, but use modern alternatives for these difficult to get parts: i.e. 74HCT02 and 74HCT00, but check the Boolean logic to ensure they should all be NAND, rather than AND. If you need AND, then the 74HCT08 is a suitable modern gate.

The 74196 should be the 74195, again use a modern alternative, the HCT is too weak to drive a relay, so go with 74LS195.

The relays are missing the important freewheeling diodes. Perhaps the ones they used have them built-in? Check the relays you intend to use have them and include them separately if they don't.

[Kleinstein]

The logic chips a re really off with the part numbering. Today it would be more like 74HC.... or a small µC anyway.

For the reference 1 V reference chips are not that common - 1.24 V bandgap references are much more common. In theory a good 1 V reference would be Ok, but it would need to be accurate too, as there would be no other point of adjustment in the 2 V range. So the divider also has its positive sides.

There is still the problem in how to realize a range below 2 V full scale: The ICL7135 works best at 2 V and would need an amplifier before the ADC.  So the ADC internal auto zero would not help here and one would need a kind of manual offset trim. Just reducing the reference is possible, at slightly higher noise and maybe higher errors. Using the ICL7135 in 0.2 V only would be a kind of compromise solution.

[Greg Robinson]

It's full of errors. One of the J-FETS is mislabelled as 2N7002 MOSFET....
...The relays are missing the important freewheeling diodes. Perhaps the ones they used have them built-in? Check the relays you intend to use have them and include them separately if they don't.

Thanks Zero999. I honestly hadn't paid too much attention to the rest of the circuit, as I plan on simplifying it down to just a voltmeter, and the mechanical relays I had planned on replacing with SSR's anyway, though I am very aware of the need for freewheeling diodes.

The 7406 doesn't make any sense. It's an inverter with open collector output.
http://www.ti.com/lit/ds/symlink/sn5406.pdf

Assuming the symbols in your attachment are correct. The NOR gates should be 7402 and NAND gates 7400, but use modern alternatives for these difficult to get parts: i.e. 74HCT02 and 74HCT00, but check the Boolean logic to ensure they should all be NAND, rather than AND. If you need AND, then the 74HCT08 is a suitable modern gate.

The 74196 should be the 74195, again use a modern alternative, the HCT is too weak to drive a relay, so go with 74LS195.

Ok, thank you very much for the sanity check, seems I'm going to need to brush up on my Boolean logic. Any chance you have a resource to recommend? And thanks, I'd planned on using more modern parts like HC, I'd made the assumption that the schematic was just showing reference "functional" part numbers rather than specifying the originals anyway.

The logic chips a re really off with the part numbering. Today it would be more like 74HC.... or a small µC anyway.

For the reference 1 V reference chips are not that common - 1.24 V bandgap references are much more common. In theory a good 1 V reference would be Ok, but it would need to be accurate too, as there would be no other point of adjustment in the 2 V range. So the divider also has its positive sides.

There is still the problem in how to realize a range below 2 V full scale: The ICL7135 works best at 2 V and would need an amplifier before the ADC.  So the ADC internal auto zero would not help here and one would need a kind of manual offset trim. Just reducing the reference is possible, at slightly higher noise and maybe higher errors. Using the ICL7135 in 0.2 V only would be a kind of compromise solution.

Thanks Kleinstein. Ok I'll keep in mind those pitfalls. Cheers.

[iMo]

FYI- https://www.eevblog.com/forum/projects/7-segment-4-5-digit-autoranging-multimeter-project/
https://www.delabs-circuits.com/cirdir/analog/instrument/del20003.png

[Zero999]

It's full of errors. One of the J-FETS is mislabelled as 2N7002 MOSFET....
...The relays are missing the important freewheeling diodes. Perhaps the ones they used have them built-in? Check the relays you intend to use have them and include them separately if they don't.

Thanks Zero999. I honestly hadn't paid too much attention to the rest of the circuit, as I plan on simplifying it down to just a voltmeter, and the mechanical relays I had planned on replacing with SSR's anyway, though I am very aware of the need for freewheeling diodes.

The 7406 doesn't make any sense. It's an inverter with open collector output.
http://www.ti.com/lit/ds/symlink/sn5406.pdf

Assuming the symbols in your attachment are correct. The NOR gates should be 7402 and NAND gates 7400, but use modern alternatives for these difficult to get parts: i.e. 74HCT02 and 74HCT00, but check the Boolean logic to ensure they should all be NAND, rather than AND. If you need AND, then the 74HCT08 is a suitable modern gate.

The 74196 should be the 74195, again use a modern alternative, the HCT is too weak to drive a relay, so go with 74LS195.

Ok, thank you very much for the sanity check, seems I'm going to need to brush up on my Boolean logic. Any chance you have a resource to recommend? And thanks, I'd planned on using more modern parts like HC, I'd made the assumption that the schematic was just showing reference "functional" part numbers rather than specifying the originals anyway.

Beware SSRs often have a high leakage. You'd be better off going with the 'type B' divider shown in the PDF, which will work with analogue switches.

Boolean logic is fairly straightforward, see the following links below. Go through the circuit with writing 0s and 1s on the inputs and determining what the outputs will be.
https://www.lotame.com/what-is-boolean-logic/
https://www.bbc.com/bitesize/guides/zc4bb9q/revision/1
https://en.wikipedia.org/wiki/Boolean_algebra

[David Hess]

There is still the problem in how to realize a range below 2 V full scale: The ICL7135 works best at 2 V and would need an amplifier before the ADC.  So the ADC internal auto zero would not help here and one would need a kind of manual offset trim. Just reducing the reference is possible, at slightly higher noise and maybe higher errors. Using the ICL7135 in 0.2 V only would be a kind of compromise solution.

Older cheap implementations usually include a separate zero and span calibration for when the reference is switched to 0.2 volts for a 200 millivolt full scale input.  A switched chopper stabilized x10 amplifier includes its own unpleasantness.

I suspect a lot of old designs got away with this because accuracy was limited by the poor common mode rejection in the input buffer which would *not* get worse when operating over 0.2 volts instead of 2 volts.  This makes me wonder how linear modern digital multimeters are; I have almost never seen it specified.

[Kleinstein]

Reducing the reference is probably still the easiest option for a 200 mV. This would be a 100 mV reference as the ICL7135, like the 7106/7 has a range of +-2*R_ref.  AFAIK the linearity in the old dual slope converters is not that bad, except at zero, where the ICL7135 usually likes to have an extra adjustment that is described in the data-sheet. I am not sure this adjustment would be different for the 200 mV range.
For good linearity the integration cap essentially needs to be polypropylene - which makes it rather bulky.

Modern DMMs (with more than 3.5 digits) tend to work a little different: more like a sigma delta ADC and numerical scaling. So no more pots but calibration via software.

 

[David Hess]

AFAIK the linearity in the old dual slope converters is not that bad, except at zero, where the ICL7135 usually likes to have an extra adjustment that is described in the data-sheet.

It is not the linearity of the ADC but the linearity of the integrated CMOS voltage follower which buffers the input after the automatic zero switch which I worry about.  Siliconix made a second version of their 4.5 digit converter which left the integrated CMOS voltage follower out so that a higher performance external JFET operational amplifier configured as a follower could be used.