Best version of the ICL 7135 ADC

[Melaa]

Everyone knows that ICL7135 has been produced since 80s. For building a 4.5-digit voltmeter without microcontrollers this is the best choice. Which manufacturer makes the most accurate and stable 7135 today?

[Kleinstein]

With the ICL7135 I would not worry so much about the ADC chip itself. The tricky part is more the choice of a good integration cap, the adjustment for the corrections around zero and the reference. For stabilit the extra amplifier / divider is more critical than the chip.

The offers today may be to a large part old stock and hard to tell the actual manufacturer.  The chip is from times of true second sources, so that the chips are to a large part identical. The 2 sources I know (Intersil and Maxim) should be OK.

There are also alternative modern DMM chips in the 4.5 digit range, that may be a better choice. However they may be hard to get and even getting the DS could be tricky. Some need a µC, other may directly drive a LCD.

[Andreas]

True words regarding the surrounding components.

The 2 sources I know (Intersil and Maxim) should be OK.

Intersil is now part of Renesas. (if you want to buy "original").
The other alternative is TI. (ICL/TLC7135).

with best regards

Andreas

[Melaa]

The tricky part is more the choice of a good integration cap, the adjustment for the corrections around zero and the reference

I have integration capacitor with polystyrene dielectric and good stable zener diodes, no problem with that components.

...however, its better to use some kind of 16...20 bit delta-sigma ADC, with a microcontroller like STM32. I have old desktop multimeter with discrete dual-slope ADC, and I want connect it to PC. But this multimeter have some problems, eg 10M input resistance on all ranges and bad switches on signal circuits. Its impossible to fix it, easier make voltmeter from scratch.

[Kleinstein]

A polystyrene cap should be good, but with the zener reference it is not so clear. No all zeners are good references - the usual cheap ones are more for a supply. The other problem is that the ICL7135 wants a relatively low reference, like 1 V or even 0.1 V depending on the way the input is build. So one would need a divider and trimmer that can also add to the drift. A better start would be more like a 1.2 V bandgap reference.

If one plans to send the data to a PC, it may be easier to use a SD ADC and µC.  There is not much demand on the µC, so even a 8 bit version should be good enough.
It still needs some work for the front end, especially if one want's more than 2 V with a high input impedance.

Chances are fixing the old meter may still be easier than building new - as long as it does not use special unobtanium parts. Of cause this would not change things.

[Dr. Frank]

Hello,
the 7135 has 3 active and venerable sources, like 35 years ago, when I messed around with that chip.
It's Intersil, now owned by Renesas, Maxim who always made clones, and T. I., who might have acquired the chip design via National Semiconductor.
These are all fully compatible, and have no quality difference.

I always used a polypropylene foil capacitor MKP for the integrating 470nF.
MKS is not good enough concerning linearity and Rollover error.
Then I used an ICL8069A, 10ppm/K bandgap reference  which is sufficient for 1digit change over a +/- 5°C temperature range.

The 7135 is intended for direct and easy driving of an LED display, and I don't know any new A/D which provides this possibility.
Therefore you might use this old chip with a precision  divider as a good  4 1/2 DCV meter.

Frank

P. S. Here's a picture of my old modules. The 3rd source at that time was Harris Semiconductor, as mine, and that could be the one which T. I. acquired
My module could be equipped with an additional chopper amplifier, 7652 at that time.
Another specialty is a 100kHz tuning fork quartz for precise 50 Hz mains suppression, and a µC interface

[Melaa]

No all zeners are good references - the usual cheap ones are more for a supply

Of course! I use special precision zeners on 6.4 and 9.0 V. I think that good trimmer resistor don`t give perceptible drift on 4.5 digits.

Chances are fixing the old meter may still be easier than building new - as long as it does not use special unobtanium parts. Of cause this would not change things.

This multimeter called V7-38, made in USSR in 1988. Most other soviet universal voltmeters ("unversal voltmeter" is the name of multimeter on soviet standard) had metal cases, good relays with gold contact pads, very good precision resistors, military-version of ICs and transistors, some universal voltmeters had interfaces (mostly soviet version of GPIB) for connect to computers, etc. Together with that this voltmeters had great weight (metal cases!) and dimensions, and very expensive cost. V7-38 have plastic case like Gainta G7xx series, electronic components are trivial plastic DIP. By reason of simple circuit this voltmeter have some problems:
- 10M input resistance on all ranges. 0.2V and 2V ranges does not need input divider;
- Type of measurement (VDC, VAC, resistance) selected by manual switches (this switches have doubtful quality), thus no possibility change type of measurement remotely;
- Doubtful PCB tracing;
- Used a lot p-channel MOSFET for autorange switch circuits. This transistors are rary today and if they burn out, there is nothing to replace them;
- Sadly, obsolescence. Similar voltmeter can be makes on a 7135+7447, some RMS-to-DC IC converter and current generator for a resistance meter. Doubly sad that in USSR was made analog of 7135.

Fix all this bugs very difficult and quite expensive. Easier buy another multimeter... or made it 

These are all fully compatible, and have no quality difference.

This was what interested me most. Thank you!

The 7135 is intended for direct and easy driving of an LED display, and I don't know any new A/D which provides this possibility.

On the other hand, microcontrollers are used everywhere today. Simple and fast autorange, connect to computer, mathprocessing, digital calibration - 7135 is not able to do any of this.

Another specialty is a 100kHz tuning fork quartz for precise 50 Hz mains suppression

When I designed my version of 7135 voltmeter, I use 2 MHz quartz with logic CMOS IC divider on 10. 200 kHz frequency should give faster a/d conversion. Is there any reason for using 100 kHz specifically?

[Kleinstein]

One of the weak points of the 7135 is the range around 0 due to the dealy from the comparator. It gets more critical with a higher clock.  While the DNL is vey good at most points, it can be quite bad in the +-2 LSB range, especially with a fast clock.
The positive point with a faster clock is that one can get away with a smaller integration cap. As a downside one really needs the trim for the DNL around zero.

Rare P-FETs and aging switched are indeed a problem.

The ICL7135 can be used with 200 mV or 2 V range, but it can be quite tricky to switch between the 2 options.

[Melaa]

While the DNL is vey good at most points, it can be quite bad in the +-2 LSB range, especially with a fast clock.

Measurement error like 0.02-0.05% without digital correction is a good result. V7-38 have error 0.04% (0.2 and 2 VDC ranges) and 0.07% (20, 200 and 2kV VDC). Likely, rise of the measrement error caused by nonlinearity of input divider.

The ICL7135 can be used with 200 mV or 2 V range, but it can be quite tricky to switch between the 2 options.

I was planning use reed relay DPST type which switched 1V and 0.1V from wipers of 40-turn trimmer resistors. Resistors connected between cathode of precision 9V zener and ground.

[kleiner Rainer]

Recommended reading about the error sources of the ICL7135:

https://www.renesas.com/us/en/document/apn/an017-integrating-ad-converter-icl7135

Here most of the old Intersil application notes:

https://www.renesas.com/us/en/products/analog-products/data-converters/analog-digital-converters-adc-precision/integrating-display-output-ad-converters/icl7135-4-12-digit-bcd-output-ad-converter#documents

Greetings,

Rainer

[Kleinstein]

Changing the range of the ADC by only changing the reference votlage has a negative effect on the noise and linearity (near zero).
The diode / resistor combination at pin 4 would depend on the reference current - so the turn over compensation would likely be wrong when just reducing the reference.
Also the noise would be a little higher with only a reduced reference.
Ideally one should also change the integrator resistor if the reference is changed a lot.

For switching the reference voltage CMOS switches (e.g. HC4053) should be better than a reed relais.  A reed relais may be an option for adding a 2nd parallel integration resistor for the 100 mV range. However I am not sure about the extra stray capacitance.

For adjusting the reference voltage settings a combination of fixed resistors and trimmer is likely better than just a trimmer.

The accuracy rating for the DMM include more than just the ADC. There is especially reference drift and as already noted the input divider.