ICL7107 - Is VRef Greater than 1V Possible?

[Trurl]

I have a 20(19.99)VDC panel meter based on the ICL7107, and I'd like to modifiy it for use as a current meter for an HP "Flex Slot" type Gen9 server PSU (HSTNS-PD43). The PSU supposedly has a current sense output (1st pin on underside of PCB) of 25mV/A.

The panel meter has a 10:1 voltage divider on its positive voltage input, decimal point in the center activated (e.g. 19.99), and internal components setup for a 2V full scale arrangement. Hence the "reference voltage" into pin 36(Ref Hi) is set to around 1V. With this Vref set to 1V, I know I could simply modify the 10:1 voltage divider on the input to a 2.5:1 divider and "move" the decimal point on the display one digit to the right (i.e. 199.9), so that the PSU's 25mV/A would show up as desired (i.e. 100.0 for 100amps with 2.5V input)

But on Intersil's ICL7107 datasheet page 10 regarding "Reference Voltage", there is the following suggestion:

"The analog input required to generate full scale output (2000 counts) is: VlN = 2VREF. Thus, for the 200mV and 2V scale, VREF should equal 100mV and 1V, respectively. However, in many applications where the A/D is connected to a transducer, there will exist a scale factor other than unity between the input voltage and the digital reading. For instance, in a weighing system, the designer might like to have a full scale reading when the voltage from the transducer is 0.662V. Instead of dividing the input down to 200mV, the designer should use the input voltage directly and select VREF = 0.341V(sic, should be 0.331V?). ... This makes the system slightly quieter and also avoids a divider network on the input. The ICL7107 with ±5V supplies can accept input signals up to ±4V. ..."

This note above from the datasheet made me think that if a voltage reference of 2.5V(fed directly from the onboard TL431) could be input into pin 36(Ref Hi) instead of 1V(set by voltage dividers currently), then it would be possible to do away with the 2.5:1 voltage divider on the panel meter's positive voltage input.

[ Questions ]
1. What is the possible maximum input voltage for pin 36(Ref Hi) as a voltage reference(i.e. is a Vref greater than 1V, say 2.5V possible)?
2. Would there be any advantage in my case, in setting up the panel meter by adjusting the voltage reference instead of using an input voltage divider?

[macboy]

Yes, 2.5 V Vref should be completely fine. With higher Vref and input voltage, you might get less noise and less zero offset, but neither is typically an issue with these 3.5 digit converters. Mind the integrating capacitor value which depends on Vref.

[Kleinstein]

The reference input range should be similar to the input voltage range. So in theory up to +-4 V. However, as the input seems to be limited to 4 V it makes little sense to go high than 2 V for the reference. This would limit the usefull range for the input. With only 1 polarity one may still be able to go a bit higher with a shifted common mode level / supply, but I have not checked this.
There are also 1.25 V references, that may be more suitable than the 2.5 V from a TL431.

For the adjustment to a different range it is not so much the integration capacitor to adjust, but the resitor at the integrator. AFAIR there is an upper limit for the integrator current and ideal current to choose for best linearity.

With only 3.5 digits the ADC noise / accuracy is not that critical. Even with only the standard 200 mV range the ADCs own noise and offset are small to external errors. So there is not much gained in going higher with the reference.
An adjustment of the input divider it likely easier and good enough.

[Trurl]

macboy, Kleinstein, thank you for your insights.

The reference input range should be similar to the input voltage range. So in theory up to +-4 V. However, as the input seems to be limited to 4 V it makes little sense to go high than 2 V for the reference. This would limit the usefull range for the input. ...

The max current I would need for the display is 100.0(amps) so half of the scale would be unused (i.e. the actual current sense input would not exceed 2.5V for "In Hi" pin 31).

Hence, if a Vref input of 2.5V could be utilized by the ICL7107 chip without issues up to 1000 counts, that would suffice for my purposes. Any further counts upwards would be unnecessary, so if the only issue were that ICL7107 could not process inputs higher than +4V (fully up to +5V with respect to the Vref of 2.5V), that would be fine as long as the ICL7107 functions "normally" up to 2.5V for "In Hi"(pin 31) input.

If Vref of 2.5V were used,
1. Would the ICL7107 function otherwise "normally" up to +4V (i.e. at least up to 2.5V) input?

2. What changes (if any) would be recommended to the present integrating resistor(470k) and capacitor(0.22µF)?

[Kleinstein]

When using only half the range, the 2.5 V ref. should be OK.

For the changes to the integrator it would be a higher integrator resistor, proportional to the ref. voltage, so to use the same current as before.
So if the original was 1 V than the 470 K would change to 1.2 M, but 1 M should also be OK. One could check the datasheet on the suggested values and maybe get a different combination, maybe with a smaller resistor and larger capacitor. For the capacitor the mechanical space my be limited and a larger cap could be tricky - PP caps tend to be relatively large. So keeping the old capacitor could be a good idea.