High performance dual slope ADCs

[fonograph]

I am noob when it comes to dual slope ADCs,I have couple questions.

1. What is the name of the 3458A adc?
2. Who is the manufacturer of 3458A adc?
3. Is it true 3458A adc is best dual slope converter ever made?
4. Is the 3458A adc still manufactured today?
5. Are high performance dual slope converters power hungry?
6. What kind lf sample rate can I expect from 32bit dual slope adc?
7. What are best dual slope adcs I can buy today?
8. How many manufacturers make 32 bit dual slope converters today?


Basicly I looking to obtain some really good dual slope adc,I want something with many bits 24+ and great linearity,low drift and noise and all that good stuff.I have read that 3458A had best adc of its kind ever made,is that true? I dont know if I should hunt for 3458A adc board on ebay or buy brand new different adc from somewhere else.

Isnt dual slope adc relativly simple? Cant I make high performance unit from discrete components?

[dl1640]

it is encouraged to read hp journal for detailed 3458A introduction.

but i lost the link...anyone can help?

[chickenHeadKnob]

I am noob when it comes to dual slope ADCs,I have couple questions.

1. What is the name of the 3458A adc?  Gerry
2. Who is the manufacturer of 3458A adc? HP-Key-Agisight
3. Is it true 3458A adc is best dual slope converter ever made? No, it isn't a dual-slope it is multi-slope, but yes is the best
4. Is the 3458A adc still manufactured today? yes
5. Are high performance dual slope converters power hungry? more than some single chip delta-sigma so yes
6. What kind lf sample rate can I expect from 32bit dual slope adc?  0 sample per second, 32 bit is fantasy
7. What are best dual slope adcs I can buy today? Ti still makes a 20-bit one
8. How many manufacturers make 32 bit dual slope converters today? no one


Isnt dual slope adc relativly simple? Cant I make high performance unit from discrete components? yes

[cat87]

Ok, so first thing you should do, as recommended, is get the manual for the 3458 and have a look through it to get an idea how it works and how the ADC is actually implemented.

Here's some helpful links:
https://doc.xdevs.com/doc/HP_Agilent_Keysight/3458A/5965-4971E.pdf

https://doc.xdevs.com/doc/HP_Agilent_Keysight/3458A/doc/3458A%20CLIP.pdf

Now, one think you should know off hand is that  single slope and dual slope ADCs don't come as a packaged IC. They're usually made with  discrete components i.e. a good cap, a comparator, a voltage reference and most likely some sort of MCU / FPGA to control and record everything.

To get a feel for how such an ADC might be implemented, this Application Note might be of some use:

ftp://bitsavers.informatik.uni-stuttgart.de/components/national/_appNotes/AN-0260.pdf

Also, have a read on how the number of bits influences your readings and about ADC converters in general. You can't just have an arbitrarily large amount of bits and still expect to get  an accurate reading. The more bits you have the noisier you conversion will be.

Now, I'm no volt nut, so I can't really  answer you questions, but as far as I remember, the 3458A has something like an equivalent  20-22 bits resoution, but I might be wrong here. Maybe someone else can confirm/infirm this.

[technogeeky]

For reference, this post (along with the surrounding thread) discuss (at a summary/overview level of detail) all of the available hard-metrology-grade (8.5 digit) DMMs, their voltage references, and (most relevant to this post), their ADCs.

[chickenHeadKnob]

Yesterday I was in a bad mood and not about to spoon feed someone who didn't want to make a minimal search before asking here, but today I am more forgiving. 

First when you say you want a 32 bit ADC, and who doesn't! you should do a simple calculation like any engineer would. Ten volts divided into 2 to 32 parts is approximately 2.3 nano volts (check!). Now look up the noise performance of various good op amps and references. The LTZ1000A is around 1.2 microvolt peak to peak - this is very good for a zener. Next calculate that since HP3458 is called 8.5 digit decimal meter you can estimate in binary that it has about 27 bits resolution, as 2 to 27 is 134217728 (check!). See if you detect where some of the difficulty in achieving high resolution lies.

The hp journal article everyone mentions is from April 1989 here:http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-04.pdf

I think the schematic linked above by cat87 is more informative as it will show you how the integrator is implemented  along with the switching hybrid chip and video op-amp + fast comparator. You do not buy this as a single module or chip from a semi-conductor maker, it is one big PCB. Some where on TiN's Xdev site there is the schematic of the keithley 2002 ADC lovingly reverse engineered by TiN, it might be easier to follow.

It is possible to get good results - say 20 bits quite simply with a microcontroller and an integrator + reference + analog switch using only cheap parts. Beyond a certain resolution any ADC method hits a brick wall, for a single chip delta sigma for example you can get to about 21 or 22 bits before noise from the digital switching section limits  results

[Siwastaja]

Oh, you easily get 32 bits.

Just use any 16 bit ADC and a random number generator generating 16 LSBs more. Make yourself / your customers believe you have 32 bits, and everybody is most likely happy, since they actually didn't need 32 bits to begin with

[alanapar]

Here you go ;-)

https://www.ti.com/product/ADS1263#tech-docs

OK, so it's not dual slope, but it has the bits!

[Johnny10]

 

[Kleinstein]

With the SD ADCs the given number of bits is just the number of bits in the inteface and does not refelct the true resolution. The lower bits often contain little information already with the more common 24 Bit SD ADCs. The did the step from 24 to 32 bits in the interface because they got slightly better than 24 bits with these ADCs. This is allready quite some achievement.
For the ADC resolution it is not just the numerical resolution (quantization steps) but more the noise that is relevant. In addition the accuracy as linearity is an important factor.

[Terry Bites]

 Find out how a dual slope ADC works: www.analog.com/media/en/training-seminars/tutorials/MT-027.pdf  Then see how HP inplemented it. They are inherently slow but very accurate.
It takes 2^(n+1) clock cycles for an n-bit converter to settle on a result.

We've been here before https://www.eevblog.com/forum/metrology/3458a-adc-section-innards/