16 Bit Multimeter "DIY"

[Kleinstein]

A string of resistors is a quite common solution, e.g. found in the HP34401, HP3456, The SDM30xx and many other meters. The resistors are sometimes lower power and may not withstand the full 1000 V for an extended time.  The main downside it the large size (to get rid if the heat) and when going beyond 7 digits the extra noise from the resistors. The next step up is to turn of the input path with overload for an extended time.

The resistor chain is cheap and reliable, so not a bad solution up to some 100 nV resoluition. It gets even more practical if one limits the maxium voltage for an extend time to a lower value, like 500 V.

[Ole]

I do intend to utilize other means of protection as well, with the hardwired protection being a last line of defense.
Additionally I will utilize a Guard Terminal (I do have to purchase some triax cable for it anyways so I could utilize it as my test lead as well)
The plan, as of now, is to utilize a set of DC preamplifiers (and maybe later some AC ones too, but thats another story), one for Voltages and one for current.
The preamps should utilize the minimum possible bias current and would consist of the actual signal OpAmps and the Bootstrapping circuit.

[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

I tried it and there's a big improvement of the INL 
(my "reference" is just a 34461A)

Next step is to try the LTC1050 but I need to order it.

Thanks!

[Ole]

Hiho,
I haven´t given anyone an update in quite some time.
So here we go:
IA1 Powersupply:
-Inguard Transformator Output 1:
   Bipolar +/- 20VAC for +/-24V Rails with low Current and +/-15V Rails and +/-5V Rails.
   Supply for all OPAmps and analog switches as well as the Vref module
-Inguard Transformator output 2:
   Bipolar +/- 15VAC for +/-12V Rails and 5V Rail
   Supply for Relais and High Current Source and the Inguard Processor

IA2 ADC
-AD7177 based ADC Sector (very similar too the HPM7177)

IA3 Analog Frontend
-DCV LN Amp via OPAx140 with bootstrapping with near Zero Current Auto-Zero
-DCV LI Amp via ADA4530 for very high impedance sources
-DCI via 100mR and 10R Shunts and (this is for the ac part aswell) 5mR for very high currents
-DCI Low Current via Inverting Amplifier feat. ADA4530 and 1k and 100k (may change R value)

-Signal Amplifiers with x1, x10, x100, and x1000 amplification
-Analog I/O (at the joint connecting the Amplifiers to the ADC Sector) for +/-10V differential.

IA4 Voltage Reference and Buffer
Vref Module in the HP3458 A9 style and a HPM7177-style divider with trim to get the Vref to 4.9xxV with the 5V rails being at 5.1xxV

IA5 Resistance Current Source/ Variable Voltage Source
Module based on two Caddock 1776C6815 Decadal Divider Networks
-Produces Constant Currents and Voltages in decadal steps.

[free_electron]

before you start doing all that goodness : ask yourself if the chosen A/D is the right way to go ... there is a reaon all multimeters use a dual or multislope converter : they can be calibrated very simply and all errors are self erasing.

[Kleinstein]

For the origial 16 bit aim a ready made SD ADC is not such a bad idea. It just is made to work with a 3 or 5 V reference and not with a 7 V reference.
The main advantage of a multislope ADC is that they can easy work with a higher reference voltage like 7 V or 10 V and have an input range of some 10 V, which may be convenient to get high input impedance also for up to 10 V or even a little more. With an SD ADC this needs extra effort.

For the beginning it is OK to start with a simpler/cheaper solution - the 1st version has a chance to fail or at least have quite some weak points. The learning part is not so much about using expensive parts, but to get reasonable performance even from cheap parts.
The ohms part is a separate topic. A simple Ohms version should be OK with more normal resistors, at least for the start. It would be after one has a somewhat working solution to think about what parts need to be better and plan for a final, stable version.

With the usually relatively low power (e.g. some 2 - 5 W) for a DMM it does not make much sense to use 2 transformers. This only adds to the coupling capacitance and leakage and the extra tranformer loss is likely more than heat from linear regulators droping a little more voltage.  For a higher supply with low power one could consider a mains frequency charge pump to double the votlage.

Expecially with an ADC like AD7177 one usually will not need a gain of 1000, except for a special nV input amplifier. One can even question the need for a gain of 100.  Some DMMs (e.g. SDM3055) seem to have no analog gain before the ADC (in this case AD7190), at least not for the voltage ranges. The smaller input range of the ADC chips means less gain is needed.

[iMo]

Anecdotal note: I did many times use ADS1100 (2 diff inps, 6pin chip w/ I2C, 8smps/s @16bit, internal 2.048V ref w/ 5ppm/K) and it worked nice. Put a higher voltage floating opamp stage (discussed here several times) with a divider in front of it and you will get a nice voltmeter (as the first step in your effort).

[macaba]

-AD7177 based ADC Sector (very similar too the HPM7177)

Among the bigger challenges of the HPM7177 were:
1. Fully differential input amplifier CMRR [for both DC+AC] (solved with high performance resistors, CMRR trimpots, TEC temperature regulation and very careful amplifier selection/layout)
[Note: ADC CMRR specification does not help here - front end CMRR error shows up as differential error on ADC inputs]
2. SAR input charge injection (solved with high performance/speed amplifiers)

With those 2 tricky design points in mind, you might be able to make things a lot easier with the AD4630-24 instead of AD7177. It has a similar noise specification, with the addition of a few very useful features...
It's a differential ADC that also fully supports single ended input [see Figure 36 of datasheet] (with In- fixed at REF/2 or 0V, and In+ between 0V and REF) and has defined specifications in the datasheet to support this thus allowing single-ended input stage (with output shifted to fit the ADC CM range) which is easier to get right. Challenge 1 solved.
It also has easy-drive inputs which use pre-charge tricks to significantly reduce input charge injection, making the front end design easier again by allowing a wider range of amplifier choices. Challenge 2 solved.

As a side perk - the loss of dynamic range from not using differential input can be alleviated by running both ADC channels of the AD4630-24 in parallel from a single input.

If this ADC was released at the design time of the HPM7177, it probably would have been the HPM4630!

[Ole]

So concerning the Choice of ADC,
my first choice was the AD7177 because the HPM7177 was based on it so i wouldnt have to do most of the hard work.

I´ve looked around for some other comparable ADC´s and here is my choice so far:
AD7177:
DS, max 10kSPS, 32bit
0.07µVrms noise @5SPS (24.6 bits)
AD4630-24
SAR, max 2MSPS, 24bit
17.6 (105.7dB SNR) Noise Free Bits @2MSPS
LTC2500-32
SAR, max 1MSPS, 32bit
17.5 (104dB SNR) Noise Free Bits @ 1MSPS
148dB Dynamic Range @61SPS
ADS1263
DS, max 38kSPS, 32bit+24bit
21.9 ENOB @2.4kSPS (SINC4) ADC1

I might have to rework the noise specifications and make a better comparison.

[iMo]

The sad state of AD7177
all the ppms are weeping
"buy me! buy me!"

Digikey
AD7177-2BRUZ-RL7 = 9/4/2024 (do they mean Sep or Apr ?)
AD7177-2BRUZ = 5/1/2023  (is this May or Jan ?)..

In the meantime you may create the schematics, design a quality pcb and write the software.. 

PS: I think that from 85% the shortage is caused by the "business practices of the distributors" thus many P1/P2/P3 customers order directly these times..

[maat]

AD7177-2BRUZ = 5/1/2023  (is this May or Jan ?)

That is a MAYbe 

[Ole]

So, I´ve been brainstorming over the last few weeks and have found a slight problem,
I need a set of analog switches with very low Ileak and preferably a low Ron

These Switches would be used to connect the various functions to the amplifiers and the ADC(s)
Therefore I could allow myself to use OP´s like the AD797 or LT6018 for minimized noise.
My ideas so far:
1. Utilize the MAX326/7 and use 1.5kR Ib canceling resistors

2. Try to build a JFET multiplexer with a low resistance/ low leakage FET (e.g. 2N4858)

[Kleinstein]

Getting good CMOS swtiches is currently a bit tricky. The lower R_on ones (e.g. DG411) have usually more leakage. So one kind of has to compromise. Quite often one does not need low R_on.
The leakage current depends on the voltage (may be nonlinear and temperature dependent) and cancelation is usually hard.
The AD797 is likely not the right choce in a DMM. One should choose OPs that don't need extra low On_resistance.
Dending on the circuit one can tolerate some leakage / offest and just correct numerically.

[Ole]

I just checked with Maximintegrated.com, Digikey has just shy of 300 327 in a DIP-Package in stock.
Those switches would be used for most of the instument, i.e. Function connetcing, amplifier gain selection (that might be done via JFETs or Relais) and the input AZ-Cycling as well as the shunt selection on the sense side.
But for the ADC driving the current plan is to use the schematic from the HPM7177 and maybe add AD797 instead or auxilliary to the AD8065.
Then I´d be able to use JFET´s more easily, since my signal is not +/-15V max. and I can just use -5V Gate Voltage to lock the JFETs.

[sahko123]

I believe adi may be phasing out the ad7177 be it for a successor to have fresher ppms or because of too low a demand to merit spinning up a few wafers so even if they make more to restock chances are it'll either be sooner than 2023 or never again.

[iMo]

Look at Octopart
https://octopart.com/search?q=ad7177&currency=USD&specs=0
there is a seller with large stock (be careful, of course)..

[Kleinstein]

The max327 are very low leakage, but also quite high in resistance. Except for very low bias inputs there is not much need for such low leakage. Another popular choice is DG211B. Ti also has a few new CMOS chips with low leakage (e.g. TMUX611x and a few more). Directly at the ADC input one may get away with lower voltage rated parts, that may be better available. There the high resistance of the max327 is likely a problem.
JFET switching is possible, but it needs an extra negative voltage some 5 V below the range. This makes it a bit tricky for a +-12 V signal range. One also need an auxiliary signal to drive the gate when on. Driving from a buffer behind the switch can cause quite some switching spike.  A positive point is that there is essentially no leakage when on. The availability may be a bit better than with special specific CMOS switches.  The JFETs of choice are MMBF4117 , MMBF4393 and maybe MMBFJ202 or 2Sk208.

[iMo]

FYI - femtoAmps leakage with 1N4148 protection diodes..
Will work with bipolar the same way..
PS: 52fA with 1N4007 ( ͡° ͜ʖ ͡°)

[Ole]

FYI - femtoAmps leakage with 1N4148 protection diodes..
Will work with bipolar the same way..

If I had gone for ultra low leakage with ADA4530 for example I would have chosen something similar
But for my current design iteration I´m going with DPAD1 low leakage diodes and 30k Resistance up front.
The DPAD are connected to the power rails of the MAX327 (i.e. +/-15V)