Interesting precision voltage divider with schematic and calibration PDF

[Overspeed]

Hello

In my quest to built a home made voltage divider . I have located this article , good source of data

Regards
OS

[Gyro]

The paper seems to describe a standard decade dividing resistor chain with construction notes for minimizing thermal emfs [Edit: and using 4 terminal configuration]. Importantly they describe it as a short term transfer standard which is calibrated by a 'Top class calibrator' and 'Precision DMM'. In the context of home use, if you had those, you wouldn't need the divider.

For home use, if you want something that gives you decent accuracy decade divides, and is effectively self-calibrating, then try a forum search for 'Hamon divider'. One of these, in combination with a fixed voltage reference will probably be closer to what you want.

[Conrad Hoffman]

Very standard sequence used in test equipment, but still a useful article!

[jwet]

Not quite as academic but a very interesting line of products was put out by Maxim in the 90's, Charlie Allen is the application definer and inventor.

The MAX4578/79 data sheets explain the idea better than I can.  The idea is that the ratio of a long divider and analog switch can exceed 15 bits in a commodity process because of how the tolerances and errors stack up.  You could have a 4096 divider and a 15 to 4096 and a 4081/4096 divider with extremely tight tracking and accuracy based on process physics.  It takes advance of a somewhat unusual spec in long dividers that negates a lot of errors.  These are used in industry but the technique is not well known.  I directed analog switch product definition (as well as all "standard products devices") when these parts introduced.  A few of our customers "got it" and use these extensively but most didn't (the secret dies with me).  They let you do some pretty accurate stuff for little money.  The referenced paper is describing similar things.  Its just an observation of how cleverly the math works out, something Charlie Allen was famous for...

The 15 bit spec above was driven by customer requirements and cost- industrial automation/PLC's.  The ideas however can be extended to any arbitrary level of precision for low relative cost.

[mawyatt]

Not quite as academic but a very interesting line of products was put out by Maxim in the 90's, Charlie Allen is the application definer and inventor.

The MAX4578/79 data sheets explain the idea better than I can.  The idea is that the ratio of a long divider and analog switch can exceed 15 bits in a commodity process because of how the tolerances and errors stack up.  You could have a 4096 divider and a 15 to 4096 and a 4081/4096 divider with extremely tight tracking and accuracy based on process physics.  It takes advance of a somewhat unusual spec in long dividers that negates a lot of errors.  These are used in industry but the technique is not well known.  I directed analog switch product definition (as well as all "standard products devices") when these parts introduced.  A few of our customers "got it" and use these extensively but most didn't (the secret dies with me).  They let you do some pretty accurate stuff for little money.  The referenced paper is describing similar things.  Its just an observation of how cleverly the math works out, something Charlie Allen was famous for...

The 15 bit spec above was driven by customer requirements and cost- industrial automation/PLC's.  The ideas however can be extended to any arbitrary level of precision for low relative cost.

Interesting devices!!

Don't some of the new high resolution DACs use a very long series string of the same resistor in a voltage divider arrangement to the Vref, then simply use a switch to each junction for the output selection, sort of a massive tap Digital Pot configuration. Recall TI and/or ADI doing this on some of their newer high resolution DACs.

BTW we ran one of the first SiGe chips in the GST2 process at Maxim way back, fun story.

Best,

[Kleinstein]

The max4578 has some neat features to check an ADC. There is still a big downside: it is a rather special chip with linkely no 2nd source and given that it was a Maxim chip, there is a risk with the supply chain. It is much more convenient to have a more normal (that is standard pinout) MUX chip and the check part separate at 1 or more of the 8 input paths.

Switching the reference current can also be an issuse.

[jwet]

>>BTW we ran one of the first SiGe chips in the GST2 process at Maxim way back, fun story.

Maxim was a lot of fun in its heyday with Jack Gifford at the wheel.  I came in 1993 from industry to build up their apps and product definition for standard analog products.  I served as a part time Strategic FAE in the Southwest and RTP, NC as an excuse to live remotely from the bay area.  Retired as the managing director of standard products about 5 years back- can't remember.  Were you stationed in Oregon?  Those SiGe processes were way ahead of their time- Tektronix.

Kleinstein- I joined Maxim in the thick of those delivery problems around 93- it really sucked being face to face with irate customers.  We grew at 30+% through the 90's, amazing for the stock price but we couldn't grow enough capacity to really keep up until 2002 or so.  Maxim, after just its first few years did all its own Fabrication (98%)- it was a  quality thing and before the rise of big foundries.  The problem with rapid development and internal Fabs is if you've got new parts that aren't up to stable yields, have very high deman, your output sucks.  We were doing a bunch of really nice state of the art Dual Synchronous bucks with yields of 15% for notebooks on a workhorse in-house 6", 1.2 micron process and these were big dies.  Demand was infinite.  We did a bunch of things to fix it, we bought the Tektronix fab in Beaverton to make their exotic stuff and our standard sexy analog BiCMOS with thin film, poly caps etc.  We bought another big Fab in Sunnyvale.  We kept investing in boutique analog and move most  of the successful 1.2u parts down to .8u and finally to .4 micron process that we ran 300 mm wafers on by 2005 or so. The first analog company to run 300 mm.  This really fixed the capacity issues, huge wafers, tiny high yielding die and great design experience.   We drew the internal process line at about 180 nm and went outside for small CMOS.  We learned how to do sexy analog on 45 nm processes without all the bells and whistles.  Precision analog is tough because you need stuff in the process and none of the foundries especially back then had any.  This delivery mess really cost Maxim a ton of reputation loss- bad management for growth- I don't think our CEO really believed the growth was real until it was too late and ran us over.  Its easy to get fooled by forecasts, etc and overbuild, run yourself out of cash and die.  The valley is littered with them.  I can tell you on the design side of the house, it was a great place to work- some absolutely amazing designers who really produced.  Engineers of a certain age have this bad taste from the 90's that will never go away- I know Dave Jones has made comments.  I had a great career there that spanned 25 years.  It all belongs to Analog Devices now with all their great stuff and  LTC's stuff.  They have retained a large portion of the design staff which is a positive.
Take care.

[mawyatt]

>>BTW we ran one of the first SiGe chips in the GST2 process at Maxim way back, fun story.

Maxim was a lot of fun in its heyday with Jack Gifford at the wheel.  I came in 1993 from industry to build up their apps and product definition for standard analog products.  I served as a part time Strategic FAE in the Southwest and RTP, NC as an excuse to live remotely from the bay area.  Retired as the managing director of standard products about 5 years back- can't remember.  Were you stationed in Oregon?  Those SiGe processes were way ahead of their time- Tektronix.

We are/were in Florida, developing the first RF/MW/MMW chips and SoCs in silicon at the time. Lots of folks said can't do what we wanted and did in silicon, they all said we needed less complexity and in 3/5 technology. 

Anyway, we had just developed the 1st fully integrated (everything on-chip) MW receiver in a research Bell Labs complementary silicon process, later know as CBIC-V2, and trying to get access to IBM's early SiGe BiCMOS technology. We had also developed a wafer level active Laser Trimming process with Cascade Microtek utilizing their new Flexible Membrane probes. Someone at Maxim heard about us (probably Join Roberts) and invited us out, we worked deal to exchange the Laser Trim technology for early access to the GST-2 SiGe process (lots of details and fun storied but don't want the clutter things up unless folks are interested). Later we acquired early access to IBM (another fun story, the seedling for the Trusted Foundry) and created the 1st RF/MW fully integrated Transceiver SoC.

Lots of activity back then 

Best,

[jwet]

Very cool, I wasn't that close to RF stuff except late in my tennure, I started an Automotive group and we were defining and designing advanced receivers for XM and Sirius and other digital terrestrial standards in Japan and elsewhere.  A lot of talent and a lot of hustle.  I do miss parts of it.

Take Care.

[The Soulman]

(lots of details and fun storied but don't want the clutter things up unless folks are interested)

Interested. 

[jwet]

mawyatt:
I'm interested too.  We could start a new thread in some out of the way place- we don't have to hijack something.  I have some war stories of my 25 years at Maxim too.  If you start it, I'll contribute, just let me know where.  I think part of retirement is to reflect on past situations.  The one thing that I univerally appreciate is the quality of most of the people around me and the exceptional quality of a few.  I didn't appreciate or cherish them as much as I should have.

[RoGeorge]

(lots of details and fun storied but don't want the clutter things up unless folks are interested)

Yes, interested, and thank you both for the peek.
Always a delight to read such inside stories. 

[mawyatt]

Ok, seems we should start another thread. Probably best in general discussions area.

Best,

[mawyatt]

Semiconductor related topic started here:

https://www.eevblog.com/forum/chat/a-little-semiconductor-related-history/

Best,