Quality multimeters using sigma-delta?

[pitagoras]

The now discontinued Agilent 34405A 5 1/2 digits DMM used sigma-delta converter.
Are there other examples?
Not cheap handheld generic DMMs, I'm asking for reliable instruments.

EDIT:
Apparently the original post may lead to think that I need a multimeter and that I think that only sigma-delta is good.
Quite on the contrary, I try to understand the industry evolutions in high quality ADC, I don't want to build a multimeter, and I don't have any specific preference on ADCs. 
My understanding plan includes experiments with homemade converters, I did a tiny experiment with a sigma delta based on a microchip application note, next in the row is a dual slope, a multi slope, even a single slope (based on a TI AN).
The Art of Electronic suggests that sigma-delta is like a "natural evolution" of multi-slope, but I had read that they have INL limitations... so I wondered if an application demanding high quality A/D conversion where using sigma-delta.
Now that you agree with my wife that I need to say more words, you can add more replies aligned with my declared interests
Thanks

[voltsandjolts]

Which 5 1/2 digit or less meters don't use sigma-delta?
Aren't most DMM chipsets using sigma-delta internally?
e.g.  http://hycontek.com/attachments/DMM/DS-HY3131_TC.pdf  (U1282A)

[pitagoras]

So most 5.5 digits uses sigma-delta? Even the top players? Do they use ADC chips?
At 6.5 and more they continue using multi slope?

[tooki]

I think most handheld meters use dual-slope ADCs, too. The ICL706/7107 used in tons of 3.5 digit meters (including the $2 830B meters) is dual-slope, so in terms of sheer numbers, they probably win.

If anything, I think using non-dual-slope ADCs is probably much rarer.

[Zbig]

Why do you think that matters as long as the meter performs up to specs?

[voltsandjolts]

Hard to tell what is being used, not part of published specs. Needs a teardown to tell.

Keithley 2110 5-1/2 digit is using TI SD ADC, see this vid at 7:33

[macboy]

Fluke 8505A (and 8506A with thermal RMS) 6.5/7.5 digit meters and older 8502A 5.5/6.5 digit meters use an interesting technique somewhat similar to sigma-delta. Fluke refers to it as a "recirculating remainder A/D".  The operation of the converter is fully described in the manual. Essentially, the input is converted through a 5 bit Flash ADC, then the remainder (difference between the 5-bit value and the actual ADC input) is captured, amplified and fed back through for conversion. This is repeated 5 times, and the result is a 24-bit sample. 

Dual-slope or multi-slope is much more common than other techniques. It is simple and has inherently high linearity.

[crazyguy]

dual-slope/multi-slope ADC sampling rate is too slow for high counts meter

I think modern high resolution meters already use Sigma-Delta ADC for a long time

Understanding Integrating ADCs
https://www.maximintegrated.com/en/app-notes/index.mvp/id/1041

even so many cheap multimeter use SD-ADC now, such as Uni-T UT139C, Uni-T UT210E, HoldPeak HP-890CN (Hycon HY12P66)
http://www.hycontek.com/wp-content/uploads/DS-HY12P65-EN.pdf

[pitagoras]

dual-slope/multi-slope ADC sampling rate is too slow for high counts meter

Isn't 3485A ADC's multi-slope?

[Dr. Frank]

Practically all higher grade DMMs ( 6.5 ... 8.5 digits) use multi-slope conversion, so do all HP DMMs.
This technique  might be called differently at other companies,
And it might also be, that the last remainder of charge on the integrator will be converted by a flash-ADC, like in the 34410/411.

Only multislope ADCs feature high resolution, linearity and low SNR, due to its integrating characteristics.
This can't be done by sigma-delta.

The multi-slope conversion is definitely slower than sigma-delta for the same amount of bits.
8.5 digit DMMs usually need 1 ..20sec for a single conversion.

The 3458A has a special ADC, by using binary weighted slopes.
This increases its speed, i.e. 1.4µs for 4.5 digits, 10µs for 5.5. digits, and 166/200ms for full 8.5 digits.
That makes it the fastest of the 8.5 digit DMMs, but that's still far slower than equivalent sigma-delta.

Btw.: the conversion method usually is specified in the manuals.

Frank

[Kleinstein]

The recirculating remainder technique used in some Fluke / Phillips meters is more similar to a successive approximation ADC. It is different from a SD in that it is sampling and not integrating. The usual SD converters are more or less continuously integrating, which is good for low noise.
The Prema DMMs (6.5 .. 8.5 Digits) used sigma delta converters - though different from modern chips (they used external precision resistors instead of internal switched capacitors in modern chips).
AFAIK some of the modern Keythley SMUs also use an SD converter.

Also many modern cheap meters up to about 5.5. digits - as SD converters can be easily integrated to a chip.
The SNR of SD converters is actually quite good - just limited to the usually low supply voltage of the chip.

The modern charge balancing integrating ADCs that modern 6 and more digit meters use are not that much different from an sigma delta ADC. The main advantage of classical integration is the good linearity (INL) and large input range - SNR and resolution of SD converts can be on par or even be better.

[DavidDLC]

Why do you think that matters as long as the meter performs up to specs?

My exact thoughts

David DLC

[tooki]

dual-slope/multi-slope ADC sampling rate is too slow for high counts meter

I guess someone forgot to tell Keysight, Keithley, etc...

[pitagoras]

Why do you think that matters as long as the meter performs up to specs?

My exact thoughts

David DLC

This is a very strange question to me. You're not interested in knowing HOW they achieve a given performance?
That is a really annoying position for an engineering mind. 
Moreover, these are not kitchen scales. In engineering one needs to understand how the instrument works.
But those are just my thoughts.

[voltsandjolts]

Why do you think that matters as long as the meter performs up to specs?

My exact thoughts
David DLC

There are few folks here interested in how stuff works. You know, the don't turn it on, take it apart kinda folks!

[gamalot]

Fluke 8808A use LTC2440

[Zbig]

Oh, of course there's nothing wrong with being curious how things work, quite on the contrary. I apparently just misinterpreted your original post. From its wording I got under impression that you're looking for a meter but somehow feel that it must be sigma-delta to be any good

[DavidDLC]

Oh, of course there's nothing wrong with being curious how things work, quite on the contrary. I apparently just misinterpreted your original post. From its wording I got under impression that you're looking for a meter but somehow feel that it must be sigma-delta to be any good

Yes, I also thought that you wanted to buy a multimeter and it needs to have a Sigma-Delta converter.

You should put in your original post that you want to know how the instruments achieve the performance, and that would make more sense.

David DLC

[David Hess]

Slope converters are just one type of integrating converter; delta sigma and voltage to frequency are also.  They all integrate the input signal for a specific amount of time producing a sin(x)/x frequency response.  With the proper integration time, nulls can be placed at the 50 and/or 60 Hz power line frequencies and their harmonics which is important for low noise.

You can build an integrating converter out of a sampling converter but its linearity depends on the sampling converter which explains the darth of multibit delta sigma converters in high linearity applications.  I am not sure if anybody makes them anymore; maybe for audio?

The higher resolution and linearity slope converters are not just dual slope or multislope.  The highest performance ones are some combination of run-up and run-down which I suspect helps with dielectric absorption issues in the integration capacitor.  Siliconix made chipsets for these in competition with dual slope designs from Intersil and others (Texas Instruments?) but for some reason the dual slope designs won out and the run-up and run-down designs are all custom now; maybe they gave too many people headaches trying to understand them.

I know of one example of a modern handheld meter which does not use slope conversion.  Tektronix almost got into the handheld multimeter business competing with the Fluke 97 series resulting in a lawsuit and eventually Fluke buying their handhand multimeter products and an agreement that Tektronix would stay out of that market.  What was interesting is that Tektronix was not using slope integration and their RMS meters also lacked the ubiquitous Analog Devices RMS converter.  I suspect they used some type of sampling converter and did the integration digitally so RMS values could be computed.  They were pretty nice compared to the Fluke 97 series at the time.

[pitagoras]

The higher resolution and linearity slope converters are not just dual slope or multislope.  The highest performance ones are some combination of run-up and run-down which I suspect helps with dielectric absorption issues in the integration capacitor.  Siliconix made chipsets for these in competition with dual slope designs from Intersil and others (Texas Instruments?) but for some reason the dual slope designs won out and the run-up and run-down designs are all custom now; maybe they gave too many people headaches trying to understand them.

I'd like those headaches if you have any reference to research more on the subject... I'be been googling without luck.

[David Hess]

The higher resolution and linearity slope converters are not just dual slope or multislope.  The highest performance ones are some combination of run-up and run-down which I suspect helps with dielectric absorption issues in the integration capacitor.  Siliconix made chipsets for these in competition with dual slope designs from Intersil and others (Texas Instruments?) but for some reason the dual slope designs won out and the run-up and run-down designs are all custom now; maybe they gave too many people headaches trying to understand them.

I'd like those headaches if you have any reference to research more on the subject... I'be been googling without luck.

I have a couple of old Tektronix multimeters which use the Siliconix integrating converters.  The Siliconix datasheets show the basics but I wish they had published some application notes:

DM501A - Siliconix LD120 and LD121A
DM502 - Siliconix LD111A and LD110
DM502A - Siliconix LD111A and LD110

HP has a great description in one of their old service manuals but the closest I can find at the moment is in the HP3456A.  I asked about this in the past here and got pointed toward the HP description but the forum search is so broken that I have no way to find the post.

[2N3055]

Take a look at this:
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf

That is HP Journal that quickly explains principle of dual and multi slope A/D converters, and explains how it was implemented in famous HP 3458..

[David Hess]

Take a look at this:
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf

That is HP Journal that quickly explains principle of dual and multi slope A/D converters, and explains how it was implemented in famous HP 3458.

Thanks, that is probably the article I was thinking of.  When I downloaded it, I found my copy filed in the wrong place.

[2N3055]

Thanks, that is probably the article I was thinking of.  When I downloaded it, I found my copy filed in the wrong place.

  Happens to me too..  There is just too much data, too many files...  Windows search helps, but not all files can be fully indexed... And also do you file it under HP, APP notes, under DMM, or T&M, or theory or ...... So yeah sometimes it gets hard to find right thing. Also, i have duplicates of some documents, intentional and unintentional...

Regards,
Sinisa.

[technogeeky]

Some good multimeters use single-slope integration (Keithley 197). I'm not sure if any modern ones do, though.

[nfmax]

Agilent/Keysight U3606A & B 5 1/2 digit DVM's use a sigma-delta ADC, it says in the specification. This type of converter gives a much more frequently updated display than does a multi-slope type converter, although the actual measurement averaging period may be similar. With multi-slope you don't get any result at all until the conversion finishes, with sigma-delta you can get updates as often as you want. This has both benefits and drawbacks, you can see if a voltage is stable or not, but if it isn't stable, the display is difficult to read.

[Kleinstein]

The faster SD converter Chips like LTC2440, MCP3911 and others usually use a kind of multi-bit system internally. This likely also applies to many audio converters (even if they call them single bit) to get the required resolution at high speed. The multi-bit loop is essentially needed to get higher than 2 nd order or to get the required speed with a 2 nd order SD. The trick is using a thermometer style DAC and internal dithering/swapping the sources - this way one can get good INL, not so much limited by the DAC, as the error is largely averaged out.
Some of the Multi-slope ADCs (especially those without a dedicated run down phase) could also be interpreted as a kind of first order muti-bit SD converter with PWM used as an DAC.

With a SD converter one does not have to use a rectangular integration window and thus sin(x)/x response, but many converters use it, because it is simple. The advantage over classical dual slope is that an SD can use continuous integration and thus don't miss large parts of the signal like the old dual slope does. Modern integrating converters are integrating the input most of the time, so not missing much.

Besides classical integrating multi slope and SD there are also a few high resolution successive approximation ADCs that could in principle work very well in a high resolution DMM. The Fluke/Phillips recirculating reminder could be seen as such an approximating converter. The LTC2376 ADC chip and similar have rather impressive specs (e.g. better INL than SD ADC chips - maybe except the prema ones) - though I don't know a DMM that uses them.

[TurboTom]

The Rigol DM3068 (6.5 digits) utilizes an ADS1256 (http://www.ti.com/lit/ds/symlink/ads1256.pdf) which is specified to be a sigma-delta converter. I would consider the DM3068 a "quality multimeter" though opinions may differ... . At least I'm quite satisfied with its performance and accuracy.

Cheers,
Thomas