In message 1320939951.98248.YahooMailNeo@web37902.mail.mud.yahoo.com, Randy E
vans writes:

Do you mean 8.5 digits single-shot, or 8.5 digits with averaging ?

Considering that you can get off-the shelf ADC's where 24 bit is routine
and 26 bits available, I'm not sure I'd bother...

http://ohh.de/5610.htm

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

8.5 digits with averaging.
 
I did consider that but the ADCs I am familiar with only have a 5V input range so range resistors would be required, not desirable for me.  Also, they won't directly take the 7.2V ref input from a LTZ1000.  Plus the AN-260 circuitry is not too complex.  Otherwise, I agree with you. 
 
Randy

From: Poul-Henning Kamp phk@phk.freebsd.dk
To: Randy Evans randallgrayevans@yahoo.com; Discussion of precise voltage measurement volt-nuts@febo.com
Sent: Thursday, November 10, 2011 7:51 AM
Subject: Re: [volt-nuts] Homebrew DVM

In message 1320939951.98248.YahooMailNeo@web37902.mail.mud.yahoo.com, Randy E
vans writes:

Do you mean 8.5 digits single-shot, or 8.5 digits with averaging ?

Considering that you can get off-the shelf ADC's where 24 bit is routine
and 26 bits available, I'm not sure I'd bother...

http://ohh.de/5610.htm

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe   
Never attribute to malice what can adequately be explained by incompetence.

Op 10-11-2011 16:51, Poul-Henning Kamp schreef:

Oooh, gorgeous!

Where do I buy one and what does it cost?

• Rob.

Hi Rob,

Try the LTC2441, 8.5 hz updates at about 26 bits with pretty good linearity.

Most of the "24 bit" DAC are only really good for about 20-22 bits due to noise and non-linearity.

MK

Op 10-11-2011 20:39, m k schreef:

Ehm, a search on LTC's website for "LTC2441" returns this:

Your search - ltc2441 - did not match any documents.
No pages were found containing "ltc2441".

No 8.5Hz, 26-bit converter shows up in the parametric search, either.

• Rob.

Hi Rob,
It was the LTC 2440 and 2442.

Hello MK,

did you already build a real cirquit with the LTC2440?
It has a differential input 2.5+/-2.5V for bridge measurement cirquits like
weight scales.
So I think it will be difficult to get a unipolar 0..5V measurement out of
this device without resistive divider.

Which buffer cirquits do you use to feed the low-impedant (capacitive)
inputs?

With best regards

Andreas

----- Original Message -----
From: "m k" m1k3k1@hotmail.com
To: volt-nuts@febo.com
Sent: Thursday, November 10, 2011 10:50 PM
Subject: Re: [volt-nuts] Homebrew DVM

Hello Randy,

with a capacitive divider 2:1 or 3:1 based on a LTC1043 + LT1051
you can extend the 5V range to 10 or 15 V. The capacitive divider
is very stable over temperature and time.

With my 2:1 dividers I have less than 1 ppm drift over a
10-40 degrees (celsius) temperature range.

Absolute accuracy of the divider factor depends mainly on
charge injection of the switch and capacitor losses.
I measured values between 2-10ppm depending on different
pin configurations. Im not shure wether the examples in the
datasheet are using the correct pin numbers for minimum charge injection.

Another Question: how do you get a unipolar 0..5V range with a ADS1282?
The absolute input of the PGA-Amplifier is specified from 0.7 .. 3.75V.

With best regards

Andreas

----- Original Message -----
From: "Randy Evans" randallgrayevans@yahoo.com
To: "Михаил" timka2k@yandex.ru; "Discussion of precise voltage
measurement" volt-nuts@febo.com
Sent: Friday, November 11, 2011 5:01 AM
Subject: Re: [volt-nuts] Homebrew DVM

Hi,

I have not built one yet, but it seems to be one of a very few that goes beyond 24 bits.

MK

Randy Evans randallgrayevans@yahoo.com writes:

Hi Randy,

I have an ambition to do this too. But I think there are better circuit
arrangements that are not really any more complicated. Or at least it
seems that way right now, before I have tried to make them work!

Firstly you have to make a pretty good circuit before you can beat a
cheap single-chip sigma-delta ADC, as others have pointed out. That
application note predates these I think.

You could look at some of the patents on multislope converters (as used
in the HP3458A, and all the top meters as far as I know). They are
surprisingly readable. Also many of them have expired by now, although
that does not matter for educational use.

The "basic" HP patent is 4357600:
http://www.google.com/patents?id=vCUxAAAAEBAJ
I think this is the HP3458A.

There is a later HP one 4951053 which refines it with a greatly improved
switching technique http://www.google.com/patents?id=INImAAAAEBAJ. But
the 3458A does not use it AFAIK.

Then fluke get in the game with 5321403 and 5565869, describing a
switching sequence with imnproved elimination of charge injection
errors. (These have not expired).

There are lots of others if you search on "multislope" but I found these
the most interesting.

--

John Devereux

John,
 
Thanks for the info.  I will read them tonight and assess wether I want to proceed with my original plan or revisit it.  I have no particular preference, just what is easiest and cheapest but high resolution and accurate.
 
Randy

From: John Devereux john@devereux.me.uk
To: volt-nuts@febo.com
Cc: Discussion of precise voltage measurement volt-nuts@febo.com
Sent: Sunday, November 13, 2011 11:42 PM
Subject: Re: [volt-nuts] Homebrew DVM

Randy Evans randallgrayevans@yahoo.com writes:

Hi Randy,

I have an ambition to do this too. But I think there are better circuit
arrangements that are not really any more complicated. Or at least it
seems that way right now, before I have tried to make them work!

Firstly you have to make a pretty good circuit before you can beat a
cheap single-chip sigma-delta ADC, as others have pointed out. That
application note predates these I think.

You could look at some of the patents on multislope converters (as used
in the HP3458A, and all the top meters as far as I know). They are
surprisingly readable. Also many of them have expired by now, although
that does not matter for educational use.

The "basic" HP patent is 4357600:
http://www.google.com/patents?id=vCUxAAAAEBAJ
I think this is the HP3458A.

There is a later HP one 4951053 which refines it with a greatly improved
switching technique http://www.google.com/patents?id=INImAAAAEBAJ. But
the 3458A does not use it AFAIK.

Then fluke get in the game with 5321403 and 5565869, describing a
switching sequence with imnproved elimination of charge injection
errors. (These have not expired).

There are lots of others if you search on "multislope" but I found these
the most interesting.

--

John Devereux

volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.

The Wikipedia page on the integrating ADC also has some more
description on the basic dual-slope ADC as well as some of the
enhancements that have been made to it over the years.  Some
of the material from the patents is mentioned as well.

http://en.wikipedia.org/wiki/Integrating_ADC

Many of the designs end up being quite simple, but the devil

is in the details, obviously.

What I find most interesting about the design is that, in many
cases, components with absolute precision are not required.
Stability and low TC are what is needed and sometimes the
ratio of certain critical values are important.  Everything
else can be calibrated out at the end (within reason, of course).

Randy.

I did some dual slope ADC experiments. Just on a breadboard and using an Arduino for the math and display. No fancy stuff. The VRef was just a resistive divider from the 5 V. It was just an experiment. Used a 4066, LF356 and a LM339 but the thing worked wonderfull. Got it stable up to 1 mV in a 1 V range. Not shocking but with a nice Vref, better switches, a real good capacitor it can be made better. While figuring out how that worked I misunderstood how it should work and by accident re-invented a ( self nulling) VF converter, using my counter as display. That was fun too.

Now I have a 26 bit ( 24 bit usable) ADC waiting to be used interfaced to the arduino, I am making/experimenting with some Vrefs and have parts for a compound chopper based on a LT1052. But the realisation will be up in the future, first gathering more knowledge about percision and micro volt techniques.

Fred PA4TIM

Op 15 nov. 2011 om 00:10 heeft Randy Evans randallgrayevans@yahoo.com het volgende geschreven:

In message 1321314126.66316.YahooMailNeo@web161704.mail.bf1.yahoo.com, Randy
Scott writes:

Well, that's the fundamental trick of the integrating A/D:  You
move the precision requirement to your timebase, and we are
much better at those, than at precision resistors/capacitors etc.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hello Fred,

Did you check only noise and temperature or also linearity?

My experience is that with a simple dual slope integrator the capacitor
soaking creates big linearity errors.
I built my first dual slope with a 2,49V LM336 reference a 4051 and a TLC271
and a simple 1uF mylar capacitor.
Due to single supply I used the 2,5V of the reference as vitual ground. So
the input range of 0..5V was
divided into two ranges from 0..2.5 (negative with respect to virtual
ground) and from 2.5 to 5V. (positive).
With a simple mylar capacitor the non-linear gap between negative and
positive range was around 1 mV.

So linearity is only around 12 bits with this design. To reach this
linearity I had to do some software
trick to deal with the soakage. After each switching of the input
multiplexer to another channel I did
some waste A/D conversions with the new channel to let the capacitor
stabilize.

Years later I read the article of Bob Pease about capacitor soaking and
exchanged the mylar against
polypropylene. The non-linear gap is now around 100uV wich is around 15 bits
linearity.

At the moment I'm experimenting with the 24 bit LTC2400.
He has a unipolar input range of 0..5V (+ some mV overrange) with a 5V
reference.
The best results I get with a LT1027CCN8-5 reference. From the many
references
I tried up to now the LT1027 and MAX6250A in DIL-housing seem to be the best
when regarding hysteresis.

The effort to get around 10 ppm stability is enormous:
Galvanic isolated battery powered design.
Temperature of reference is measured and compensated by a NTC (3rd order
polynominal compensation)
Linearity of A/D-Converter is compensated. (see AN86 of LT)

The main issue is noise of the converter of around 20uVpp.
So to get less than 1 uV I have to integrate measurements over 10 minutes.
Reference drift is around 10 ppm over 1 year.

I am looking now for a converter with less noise.
So what ADC are You using?

With best regards

Andreas

----- Original Message -----
From: "Fred Schneider" pa4tim@gmail.com
To: "Randy Evans" randallgrayevans@yahoo.com; "Discussion of
precisevoltage measurement" volt-nuts@febo.com
Sent: Tuesday, November 15, 2011 12:56 AM
Subject: Re: [volt-nuts] Homebrew DVM

Andreas Jahn wrote:

The HP3458A and the HP34401A ensure that the integrator output voltage
only represents a small fraction of the integral of the signal input.
This significantly reduces the effect of dielectric absorption in the
integrator feedback capacitor.
The integrator and control logic act somewhat like a sigma-delta
integrator during the runup phase.
The effect of dielectric absorption can be further reduced by suitable
choice of runup algorithm.
The on resistance modulation of analog switches in series with the
resistor connected between the signal and the integrator summing
junction can be a significant source of nonlinearity, however it is, in
principle, possible to correct for this.

Bruce

No, did not check all. Just hooked up a powersupply and looked what it did. It was at first absolute not lineair. Before it was finished my ardiuno went up in smoke because of the breadboard ( +5 V jumped out and I did not think and just pushed it back in the +15V rail) but It was not very lineair until I made a sort of autozero and used a correction factor. After that it was sort of lineair, I had made it so that it did a measurement of the test-voltage then It switched the input to ground and made a second measurement. That value I used to correct the other.

It was just for fun, to learn about the principle of ADCs. But it infected me with the volt nut virus.
It was not intended to become a real meter.

http://www.pa4tim.nl/?p=2482 there you see the schematic and some pictures ( the top one is from the circuit that, i was told, turned out to be a VF converter)

Fred PA4TIM

Op 15 nov. 2011 om 06:38 heeft "Andreas Jahn" Andreas_-_Jahn@t-online.de het volgende geschreven:

Hello Fred,

I asked for the 26 bit (24 bit usable) ADC.
Do you really get this (noiseless) resolution with a simple dual slope ADC?

With best regards

Andreas

The ADC I am gonna use is the LTC2400. A friend who is a digit-nut made
the pcb and will make the software for me.

About the dual slope experiment. I could look better as it is. The
arduino died before I could do more tests but first results amazed me.
Remember it was just breadboard, no Vref (just resistive divider from
Vcc and, I do not dare to say, but the measurement pictures were taken
after I blew up my last 4066, so I used optocouplers to switch and
strange enough that went very well.

The drift you see in the measurements is more because the powersupply I
measured the voltage from, is not realy stable. Noise I do not know.

The experiment was after a discussion on a Dutch forum. There was some
talk about multimeters and ADCs and the a member who works at ST as an
EE told the top meters used dual slope about how easy it is to make a
dual slope ADC. He told us to just try it on a breadboard and gave some
tips. A few of us including me started experimenting with it and I was
trilled it was so easy to get it working like it did. Was fun too. I'm
now reading a lot of literature and your interesting discussions to
learn more before I build my meter using the LTC2400. So thanks for all
the valuable information. As a beginner I appreciate it very much.

Fred PA4TIM

Andreas Jahn schreef op di 15-11-2011 om 23:19 [+0100]:

I would think that Ron Tipton's 6 1/2 digit diy DVM, based on LTC2400 and a
statistical pool of Thaler(now Cirrus) VRE305A's would be the baseline
design to beat in this thread's quest for excellence. Yes?
www.tdl-tech.com/hires-vm.pdf

Stan

On Tue, Nov 15, 2011 at 7:20 PM, Fred pa4tim@gmail.com wrote:

Ok Fred,

thats the ADC im currently using too.
Although he has internally a readout of 28 Bits due to noise only 18-19 Bits
are usable.
To further reduce noise you will have to average many measurement values.
On the other side against many other sigma delta chips the LT2400
has very low offset and gain drift over temperature. And the main advantage
is the Overrange feature of +/-12% which makes offset and full scale
adjustment easy.

Im still hoping that someone knows a ADC chip with less noise
and true 0..5V (or even 0..10V) input with the overrange feature.

With best regards

Andreas

----- Original Message -----
From: "Fred" pa4tim@gmail.com
To: "Discussion of precise voltage measurement" volt-nuts@febo.com
Sent: Wednesday, November 16, 2011 1:20 AM
Subject: Re: [volt-nuts] Homebrew DVM

Hello Stan,

thanks for the article
he shows some of the key issues with precision A/D converters.
Although with the noise he mixes Vpp with rms values.

On the other side I have bougth recently two VRE3050A-References from
Cirrus.
(speced nearly the same than VRE305A but only usable with lower minimum
power supply voltage)
For the 3uVpp noise I found something around 9uVpp. (0.1-10 Hz)
Whereas on a LT1027C I measure around 1.9uVpp with the same
filter-amplifier.
And even with temperature hysteresis the VRE3050A that I have measured
is much worse than a LT1027 in a 10-40 degree (Celsius) range.

So I would be interested if the VRE305A is much better than the VRE3050A or
if the specs have
changed when renaming the references from Thaler to Cirrus.
Do you have any measurement values of some recently bought Cirrus VRE305?

With best regards

Andreas

----- Original Message -----
From: "Stan Katz" stan.katz.hk@gmail.com
To: pa4tim@gmail.com; "Discussion of precise voltage measurement"
volt-nuts@febo.com
Sent: Wednesday, November 16, 2011 2:47 AM
Subject: Re: [volt-nuts] Homebrew DVM

Hi Andreas,

I own no VRE305A references. However, if you take a look at a co-marketing
article between TI and Thaler in the Nov. 1999 TI Applications Journal, the
typical noise spec for the VRE3050 back then was listed as 3uv between
.1-10hz, just as good as the VRE305A  references Tipton used. However, when
one looks at the noise specs in the datasheets for VRE3050 on the current
Cirrus site, and at the original Thaler datasheets for VRE305/VRE3050 the
maximum/minimum noise specs bounding the 3uv typical are unspecified. I
find this unusual for such a pricey unit.

Stan

On Wed, Nov 16, 2011 at 3:10 AM, Andreas Jahn Andreas_-_Jahn@t-online.dewrote: