16 Bit Multimeter "DIY"

[Ole]

Since I couldn´t finde the specific topic here I´m going to ask bluntly:
I am currently planning a 4 3/4 Digit multimeter (I know sounds kinda boring compared to others but its a start), and am on the search for a Reference Voltage source that can supply 5.0V, as thats the intended maximum input voltage for the intended ADC, a MAX 195BEPE.
Everything from the Input to the ADC input has been planned and trimmmed for up to 5V input to the ADC.

Thank you in Advance
Ole

[Kleinstein]

There are plenty of 5 V reference - which are actually available is another point. The usualy sources often have a search function that includes the in stock point.  For the 4.5 digit range this would likely be a better bandgap type refrence like the max6070, ref5050 or similar.

The ADC is a bit unusual for a DMM, though still possibly. The mormal case would likely include quite some averaging / digital filtering to suppress mains hum.

[Ole]

Well the MAX 195 is just one of the options. I took this one onto my priority list because it its socketed, so i can switch it out. Another spot on the list is the MCP 3427.

[Gyro]

If you want the added stability of a buried Zener reference, and don't mind expending more supply current, then you could try an AD586.

EDIT: (You didn't specify handheld or Bench DMM).

[Ole]

Well my plans for the Multimeter are "household" applications i.e. being able to measure 0 to 50V with the 2^16 Steps that a 16Bit ADC can get you.

I don´t know exactly how other DMM´s are constructed but my plan is to have a Input Amp which gives me a 0 to +5V signal (an a polarization signal) for the ADC.
The ADC could then be either a 16bit or something else, depending how accurat i could get the entire ensemble to work.
The Current measurement is planned to be done using a up to 500mV Drop (which compared to the other multimeters I have access to is quite fine) and a .5, 5 or 50 Ohm shunt.
Comparing it to the Multimeter I´m using at work which gets up to 2,5V drop in the 500mA range, i find this pretty harmless.
This is then amplified to the desired 0 to 5V for the ADC via the "depolarizer".
Resistance is measured via a controlled current source (have to work on that one).
The entire device isn´t meant for speed or accuracy, its meant to be functional.

(You didn't specify handheld or Bench DMM)

I am planning for it to be a stationary i.e. bench DMM

[Gyro]

500mV drop sounds rather high by modern standards, 100 - 200mV is a more typical burden for even modest handheld DMMs (0.1R lowest shunt). Imagine that you are trying to measure the current draw of a 5V (or even worse, 3V3) logic circuit.

For a Bench DMM, it is often better to do away with the current ranges completely and use external 4 terminal (kelvin connected) shunt resistors instead, using the mV range.

[Kleinstein]

The max195 is quite fast and could allow to do digital RMS for AC measurments with a reasonable BW. It still needs a good (flat frequency response) front end for that. For more DC operation one would need to do averaging in SW and maybe a little AA filtering too if lowest noise is wanted. So it is a bit more to do for the µC, but still not a bad choice. Having a +-5 V range is also nice. With using a pseudo differential reading of 2 input signals (low side and high side) one could even get a +-10 V range without a divider.

The max195 is however a pretty expensive chip. Replacing it is still not fun even if in a socket.

The MCP3427 is not really that attractive. It has an integrated reference, but the integration time is not that suited for 50 Hz operation. One can get a bit around this by averaging 3 (or multiples of that) conversions in SW. Still the integration time is not very accurate and thus more mediocre 50 Hz suppresion is expected. That said I had an earlier nearly finish DMM project (only volt and amps) using the related MCP3421. It does work reasonable - though never got really finished and still mediocre shunts and divider.

I would consider the ADS1219 or MCP3562 worth a look. They are 24 bit ADCs, but one is OK to through away a few bits. With a modern design gain is adjuted by a numerical factor and a little more resolution (even if not noise free) can help to avoid rounding errors.

Usually it is better to let the ADC handle both polarities and not do a manual / separate change in the polarity. It is already a weak point in many dual slope ADCs to handle positive and negative signals separate.

[Ole]

Well, considering that this is my first DMM i dont think I´ll integrate a AC measurement just yet.
The voltage drop that i used as a measurement for he current measurement comes from a very very old multimeter that uses a 1kOhm shut for the µA range.
Still this is in the quite early design phase, so I can still play around with the measurement ranges and specifications freely.

[Kleinstein]

The burden voltage drop at the shunts can be chosen later. It is a balance between the need to measure small votlages (good amplification) and the heat release. In some cases a lower burden can save on the power rating of the shunts. So some 200 mV for the shunt is usually a reasonable compromise. More drop can come from the fuse. The shunts many still need to withstand up to some 1.5 to 2 V worst case, as that is about the clamping voltage from 2 diodes.

A point that can make quite some difference is the question of manual ranging and auto ranging / computer controlled range setting.

For the design the voltage range of the ADC makes a difference. I would not care too much how this translates to ADC resolution, as the ADC LSB steps and the display steps don't need to be the same. There is possibly added resolution from oversampling and in some ADCs noise of more than 1 LSB. So the ADC nominal resolution does not directly related to the counts and ranges. Quite often the ADC could give more theoretical resolution and the number of digits is more set by the stability and noise level. With an output to the computer it does not really matter if there are more digits send than the noise or stability would for a display suggest.

[Karel]

Some time ago I built a DVM based on & inspired by this video:



I filled in the missing gaps and made some improvements but basically it's the same design and
it works surprisingly well and it's stable.

[mycroft]

Could you share those improvements?

Thanks.

[Kleinstein]

The old LTC2400 milivoltmeter in the video is a bit limited in that it is one polarity only. There were a few iterations for the input buffer and protection, some with not so ideal properties (leakage from Schottkey / zener diodes). The OP777 in the early version also has a bit drift and bias, that is not good. The later used AD8628 is a good choice with acceptable bias and low noise. One may also get away with lower bias and slightly high noise versions ifo ne does not need resolution down to the 100 nV range. The large cap at the OPs output deserves a few    .

A big limitation at least in the early version is that the buffer has no negative supply and thus does not work correct around zero - that LTC2400 should at least work a few mV to the negative side too and could do a proper zero adjust, but this does not help if the OP is at it's limit.

[Ole]

If anyone is wondering, I currently am working on the exact circuit of the input board.
Turns out, for the functions that I want i could get the three needed resistor networks in a DIP-16, if they are isolated enough.
The current Boards/ Sections are:
A0 Power Supply (concept)
A1 ADC Module (concept)
A2 Switching Logic driving (concept)
A3 Input and Range switching (Currently WIP)
A4 Resistance measurement Current Source (concept)

[Kleinstein]

The reistors in the networks are usually reasonably good isolated, but the voltage may be limited. So directly at the input this may not be good enough. For later uses like a reference divider or gain the isolation should be OK. The demands can still be different depending on the part of the circuit and ranges (could be high if you want a very low current range (to the pA resolution). Some standard dividers are also available in a SO8 and SOT23 case.
For a high voltage divider at the input there are special resistor networks available with a SIL pinout.

In some areasy a chain of equal, but seprate resistors could be an alternative to a special network. With parts from the same roll, chances are the matching can be relatively good - no garanteed, but chances are good.

[Ole]

So, while this post was quite silent for some time I have modified the input board massively.
And not just that, Ive worked on the PSU and the Resistance Measurement Board.
The boards/ sectors are:
IA1: PSU
IA2: ADC Module & Internal Protection System
IA3: Input Board
IA4: Resistance Measurement Source
IA5: Voltage Reference
IA6: Internal Calibration/ Sanity Check Module
I think I`m on Revision 6(a) of the input board. Which is just a fancy voltage divider that divides everything down to +-5V (or 10V in some cases)
Then there is the ReMeSo, ofr which I now have a good idea and a concept.
The same goes for the PSU and I even have some Concepts on how to connect the ADC in an elegant way.

[Karel]

Could you share those improvements?

Thanks.

Sorry for the delay, here is a part of the schematic.
The input opamp is still the same as in the original project (which I copied) but I would like to replace it with an OPA140 (but it's out of stock now).

[Andreas]

Hello,

according to data sheet the 1uF on the input of the ADC will give a relative large offset and full scale error.
I am usually using 680-820 pF for C210 and 825R for R204 to get lowest noise and best INL. (together with LTC1050/LTC2057/ADA4522)

with best regards

Andreas

[Karel]

Hello,

according to data sheet the 1uF on the input of the ADC will give a relative large offset and full scale error.
I am usually using 680-820 pF for C210 and 825R for R204 to get lowest noise and best INL. (together with LTC1050/LTC2057/ADA4522)

with best regards

Andreas

Interesting, I'll try that.

[Ole]

I am currently working on the input portection and bootstrapping circuit for the input amplifiers.
As I have heard, the Datron 1061 (iirc, its definitely a Datron DMM) can get a full 1kV applied to its input even when measuring on the lowest range.
Some Multimeters use Zener Diodes for the input protection and I was wondering if one could use multiple 1N4007 as protetion diodes in combination with
a set of resistors both in front and behind the diodes as well as a GDT between the first resistors.
The primary OpAmps for the input are intended to be LMP7721 and the bootstrapping Amps are intended to be OPA106 (or other comparable sub pA 30V OpAmps)

[Kleinstein]

The protection has 2 parts: a clampling part, that normally used 2 low leakage diodes and than something like 2 zener diodes or a TVS or similar. The slightly tricky part here is not keep the leakage low and make sure it also works when the supply is off. The 1N4007 is of limited use here as the leakage is more on the high side and the current is normally limited to a few mAs even with 1 kV at the input. This is kind of needed to avoid excessive heat.

The 2nd part is the current limiting part - e.g. series resistors with high withstand voltage or PTCs or high voltage FETs (especially depletion ones) or a combination of those. A GDT is mainly used for short spikes of very high voltage, like ESD - especially in the combination with FETs. With resistors for the main protection one may get away without a GDT.

[Ole]

I know that the 1N4007 are happily leaking up to a few µA, thats why I chose them.
The 4007 should easily conduct the few pA that I would need without any kind of voltage drop (except ohms-style)
I do intend to place multiple 4007 in series, so that even the max. rated voltage (1kV) wont make them sweat.

[Ole]

I hope the schematic is at least somewhat understandable.
The usual Voltage difference at the OpAmps (those are Precision Amps and their respecitve bootstrapping circuits) should not exceed about +-12V
The input current is as such:
dU <30V: <2.5pA
dU 30-1200V: <5µA
dU >1200V: Whatever the GDT and the resistors in front of it decide to conduct.

[Kleinstein]

The 1N4007 should not leak to the µA range, more like low nA at low voltage. The leakage may reach the µA range when at 1000 V, but this may as well also be only the test limit.

The diodes as series element are not a good idea. With low current / low voltage (e.g. < 10 mV) they still act a little like a high value (e.g. 100 M range) resistor. So a little bias can cause quite some offset and they also add the noise like such a large resistor. In addition the diode action can cause rectification of superimosed AC and this may not cancel out perfectly for the back to back diodes, as parasitc capacitance and leakage plays a role. So even a very leaky diode that acts like a 1 M resistor would not be good.
The easier way is a resistor (e.g. 100 K - 300 K range) or a chain of resistors in sereies so they can withstand a high voltage and dissipate enough power.

With the usual DVM ciruits 1 input is connceted to ground or a low impedance connection. So the 2 input terminals are not symmetric, but one is low impedance and the other is high impedance. The low impedance link is needed for possible common mode current.

[b_force]

When making these kind of things, I would start differently.

1 - What is the expected accuracy?
A lot of people don't realize that their fancy new multimeter has something like a 0.06%+5d (or something) accuracy on the voltage.
Practically meaning that the last digit won't do much.

2 - Are you able to calibrate eventually the DIY multimeter yourself?
Resistor dividers, other range switches etc, it all adds up to the error.
When you're able to calibrate a meter by another device, you have to put a lot less effort in the design itself.

[Gyro]

I am currently working on the input portection and bootstrapping circuit for the input amplifiers.
As I have heard, the Datron 1061 (iirc, its definitely a Datron DMM) can get a full 1kV applied to its input even when measuring on the lowest range.
Some Multimeters use Zener Diodes for the input protection and I was wondering if one could use multiple 1N4007 as protetion diodes in combination with
a set of resistors both in front and behind the diodes as well as a GDT between the first resistors.
The primary OpAmps for the input are intended to be LMP7721 and the bootstrapping Amps are intended to be OPA106 (or other comparable sub pA 30V OpAmps)

The Datrons do it the 'brute force way'. The actual clamps are back to back zeners in series with inverse parallel low leakage jfets. The heavy lifting is done by a series chain of four 22k 3W resistors. This chain can dissipate 12W continuously (as you say, 1000V protection on all ranges). EDIT: That's for the high impedance ranges from 10V down to 10mV, the higher ranges go through the normal 10Meg attenuator so are protected anyway.

You can find all of the relevant service manuals on K04BB... http://www.ko4bb.com/getsimple/index.php?id=manuals&dir=06_Misc_Test_Equipment/Datron