Project KX : DIY calibrator / reference source/meter

[TiN]

Hi all, 

Since I start playing with Keithley meters, I could not sleep peacefully anymore.
This is a project to build very own custom precision source / meter / calibration reference. 
It's a public project targeted primarily for engineers self-education and learning, to show how are things evolve and what problems are there for a precision
instrumentation like this. No commercial profit expected.

So far thinking of next spec items:

* 3458A footprint, 2U 19" rackmount chassis with front-panel. Aluminum or steel frame, aluminum front panel.
* Noritake 256x64 dot-matrix graphics VFD
* LTZ1000A isothermal voltage reference (maybe two), target - 1ppm/year max
* Integrating multislope ADC in CPLD + LTC2442 ADC for residual charge measurement.
* 8 1/2 resolution (will see about actual accuracy)
* LTC2756 DAC
* Front and rear therminals (6 wire)
* DC Voltage measurement: 2mV, 20mV, 200mV, 2V, 20V, 200VDC ranges.
* DC Voltage source: 2mV, 20mV, 200mV, 2V, 20V ranges.
* DC Current measurement: 20uA, 200uA, 2mA, 20mA, 200mA, 2A, 20A ranges.
* DC Current source: 20uA, 200uA, 2mA, 20mA, 200mA, 2A, 20A ranges.
* 2W ohms, 2, 20, 200, 2k, 20k, 200k, 2M, 20M, 200M, 2G ranges
* 4W ohms, 2, 20, 200, 2k, 20k, 200k, 2M, 20M, 200M ranges
* 6W ohms, 2, 20, 200, 2k, 20k, 200k, 2M, 20M ranges
* Temperature measurements: RTD, -200°C to +1000°C, Thermocouples T,K -200°C to +1000°C
* Simple AC measurement features, up to 400VAC (prolly base on AD637JR as TRMS converter)
* PASS/FAIL Function
* Configurable buzzer
* Digital inputs/outputs (8bit + 8bit)
* Digital side: NXP Cortex MCU, ALTERA FPGA.
* USB, RS232, Bluetooth interfaces. Maybe Ethernet, but that's for sure last-last option.
* 100W linear power supply (25W display, 30W analog stuff, 25-30W heaters, 15W for anything else)
* Fanless
* 4-layer PCBs
* Extensive internal monitoring (temperature sensors, voltage sensors)
* Maybe battery option for voltage references (to keep them hot at least 24-48 hours during transportation for example)
* Software support - LabView on host PC.

So far I plan maintain this project as open-hardware, closed-firmware. I'm not good on firmware side, and working on this project alone, so unless there will be someone who
willing to handle software/firmware side (with dedicated 4-8 hours a week for this, for example), i see no point to share my spaghetti code with mistakes to public

I started buying some parts already,

Already have:



* LTZ1000ACH x 1
* some FPGAs, MCUs
* VFD
* LM399, LM199 from K2001
* some LTC1043 switched cap front-ends

On it's way few LTC2756A, LTC1150, LTC2442.

As a initial reference I will use Keithley 2400 which got calibrated this january by official Tek/Kei service center.
Later should get one of my Keithley 2001 calibrated as well for extra reference. That's most accurate gear I can get here,
no access to anything like 3458A yet...

First step i decided to make reference board, so will test couple samples of complete reference first.

Maybe somebody here can help characterize with calibrated 3458A (I could send reference boards and cover shipping both ways)?

Ideas, suggestions?
This is crazy project, so take it with grain of salt.

[TiN]

Got VFD today, it's huuuuge..



I placed Keithley 2001 VFD near for size comparsion.



Back size, made by Noritake in 1992. Parallel interface and single power input (+5VDC, 5Amp pk)

Test jig:



MCU Board powered from USB, VFD powered from Meanwell NET-35C (5V 2.5A)

Some junk from SRAM contents, in character mode:



Graphics mode



Could not resist to hook it to random LPCXpresso with LPC1769 and try control it.

[TiN]

Voltage reference module

Parts



Board - FR4 4-layer 40x80mm

+15VDC input
LTZ1000A schematics from datasheet
LT1112A Opamp
PZT3904 NPN in SOT223
Three MAX6610 temperature sensors.
LT1761ES5 5V LDO for temp sensors
Planed resistors: Vishay Z202

Inner layers right now pretty much solid power ground (opamp GND and heater- GND).

Temperature sensor locations:

Sensor 1: Near Opamp with copper plane under it.
Sensor 2: On wires, glued to top of LTZ1000A in dead-bug position.
Sensor 3: On bottom near LTZ1000A

Initial PCB idea is shown below

Top layer:



Inner 2 layer:



Inner 1 layer:



Bottom layer:

[fcb]

Wasn't there a monster LTZ1000 thread on the forum recently?

[branadic]

Maybe somebody here can help characterize with calibrated 3458A (I could send reference boards and cover shipping both ways)?

Can offer to make measurements on a calibrated Keithley 2002, if you like.

[Andreas]

LTZ1000A schematics from datasheet

Hello,

I am missing the zener sense (+/-) lines on the layout.

Why do you use the LT1112A instead of the LT1013 from the data sheet?
Is there a reason?

How do you plan to do the input voltage divider? Relays? Which resistors?

With best regards

Andreas

[TiN]

fcb,

Yes, there is nice thread about LTZ, which I'm reading thru, just want keep all related stuff here in dedicated thread, as reference is only 5% of all planned.
I don't want clog that thread with ADC/DAC discussions, etc.

branadic,

Cool, mind if I add you to a list, whom to send reference module?

Andreas,

It's ZA and ZC nets.
LT1112A seem better in specs compared to LT1013, from what I understood in datasheet.

Input divider will be using reed relays, and Vishay resistor networks (probably, if they are not astronomical cost and obtainable in small qty's).
Anybody have experience ordering resistors from Vishay Precision group? I'm gonna try it later.

I think will have few voltage reference PCBs (no parts likely) available, so I can donate those for volt-nuts, who may want that.

[Andreas]

It's ZA and ZC nets.
LT1112A seem better in specs compared to LT1013, from what I understood in datasheet.

Input divider will be using reed relays, and Vishay resistor networks (probably, if they are not astronomical cost and obtainable in small qty's).

Hello,

sorry but I still do not see going the sense lines directly from the reference (without carrying current) to any connector.

Which spec do you mean?
The LT1112 has factor 3-4 less amplification (RL=2K which corresponds to the 5mA of the current regulation output).
According to my experiences with current regulation loops this will be the key spec.

The LT1112 output current is only specified to about 9 mA.

Resistor networks (high ohmic): The CN471 is available from RS-components. DigiKey has some Caddoc networks.

Some Z201 resistors are available at catalog distributors in small quantities. (RS, Farnell, DigiKey).

At DigiKey you can order VSMP resistors which are trimmed to the value that you want.

With best regards

Andreas

[TiN]

Pink and blue traces going from zener to bottom right 2.54mm pitch connector.
Layout is not final yet, just initial placement and routing.

As for opamp, ill study more , thanks for highlights. Its pin-pin compatible with LT1013 as well.
Likely best will be try various amps to see which one works best.

[chickenHeadKnob]

TiN writes:

Since I start playing with Keithley meters, I could not sleep peacefully anymore. Volt-nutting demanded more, so this is a project to build
 
* Integrating multislope ADC in CPLD + LTC2442 ADC for residual charge measurement.
* 8 1/2 resolution (will see about accuracy, lol)

The true tragedy  of a volt nut makes itself known by stealing the quality thinking time I use to have staring at the ceiling  waiting to fall to sleep. I used to accomplish sucessfull resolution of real every day problems during this period. sigh.

As to using a 24 bit delta sigma to process the residue/zero cross you may find it too slow for the task. I would be delighted to be proven wrong though. Will be watching if you figure this out.

[quarks]

Maybe somebody here can help characterize with calibrated 3458A (I could send reference boards and cover shipping both ways)?

Hello TiN,

that will be a monster DIY project, but for sure very interesting and you will learn a lot even if you fail.

About your question, maybe I can help, because I have all neccessary cal lab gear at hand.

But which country you are from? I am asking because German custom can be a real pain, if you send from outside Europe and you have to go there and pick things up.

Bye and good luck for you great project
quarks

[TiN]

I sent stuff before to EU, including Germany. I live in Taiwan atm, and use EMS. They deliver to door. Usually small packages arrive safely. Anyway I don't have hardware yet, thanks for offer.

[jeremy]

looking forward to seeing this! Love those noritake displays, did you manage to find one without burn in?

[TiN]

It have spots on pixels, but at given price it is nothing to complain.
Even will dim pixels its way better than LCD. I'm trying to save some money on non-performance related parts first.

Already expect EMI issues from HV SMPS on VFD. 

[branadic]

Cool, mind if I add you to a list, whom to send reference module?

I'm fine with that.

[EEVblog]

The problem with having 2/20/200/2000 etc test values is on meters that are 10,000 100,000, 1,000,000 count, like the Agilent 34461A for example. Seems to be more common these days.
So 2 on these meters is only 1/5th full scale. Not ideal for a reference.
But then you can say the same thing about a 1V reference on  60,000 count meter too.

This is a huge project, good luck!

[Andreas]

Hello Dave,

these are range end values and not test values.
Even on a HP34401A you will need some "overrange" to test the linearity in the 10V range up to 12V.

With best regards

Andreas

[TiN]

That's right, it's range end before mux switching function kicks in. Output will be able to set within range to any value, at given resolution for a range.
I think to include also more stable direct 1:1 and 1:ratio outputs from reference with buffer amp, if more stability is required.

Currently plan to study LTC1043 switching capacitor block, seem quite potent for precision voltage diving, inverting, multiplying.

So far decided to go two directions on this project.

From a analog side start :

* Design main voltage reference (LTZ1000A) on it's own isothermal PCB.
* While LTZ1000 will be burning-in it's thousand hours, design output divider/ratio module (either LTC1043 or resistor networks+opamp). Check here will be against my cal'd Keithley 2400.
* Send PCBs and rotate few reference modules for people with calibrated 3458A/2002's to get actual voltage reference vlaues.
* Rotate 2 modules (VREF + DIV/RATIO module + probably battery to keep power during shipping time) for people with calibrated 3458A to get references.
* While above two modules burn-in - design main ADC (multislope, something like Keithley's 2001/2002 ADC + secondary LTC2442 ADC)
* Interconnect analog and digital world thru optical insulation
* Input protection and front-end design
* Develop voltage measurement function
* Precision current shunts/reference (lowest ranges and high ranges will have probably separate modules, first due to noise sensitivity, second due to thermals)
* Develop current measurement function
* Output stage for voltage/current source
* Calibrate and test voltage/current functions, measuring, sourcing.
* Start on resistance functions....
* To be continued...

From digital side:

* Design front panel, main chassis frame (likely aluminum, CNC-made, have a place here to order)
* Prototype linear power supply design (linear XFRMR, low-noise outputs for analog side, separate winding for VFD supply)
* Develop initial prototype digital control board (main CPU/FPGA, interfaces (RS232, USB, maybe LAN)
* Write some firmware for display, keypad, develop UI a little (ARM C)
* Digital input/outputs function
* To be continued...

I think there will be extra PCBs for everything, which I don't mind giving for free to volt-nuts, if anybody can find use of them.

[Andreas]

* While LTZ1000 will be burning-in it's thousand hours,

* Rotate 2 modules (VREF + DIV/RATIO module + probably battery to keep power during shipping time) for people with calibrated 3458A to get references.

Hello,

I think that 1000 hours are not sufficient. (Its only a standardized test time).

At least the difference voltage of my 2 LTZ1000 needed about 300 days until it stabilized.
I began to measure about 100 days after the younger reference was built.

How will you detect any accidents during shipping?
Usually you would need minimum 4 references to detect a faulty one.
3 which are continuously tracked at home.
And one which is shipped and compared to the other 3 before and after shipment.

With best regards

Andreas

[TiN]

I'll assembly probably 5 sets first, so 3 will stay at home during all time

Made a sketch of chassis and board placement. Just trying to see, what can be fit into 3458A-type case size.

Main VREF, ADC board, AC section and front end will have separate thermal boxes to minimize thermal gradients.
Front panel, digital section and mains power separated from analog side by solid chassis frame walls.

[jeremy]

Looks great! Is this in solidworks or inventor?

[TiN]

First two pics - Solidworks with Realview.
Others - KeyShot 4 renderer

[TiN]

VREF module ready to test.
Last checks

Board layers attached, TOP, BOTTOM up side, INT1 and INT2 bottom side.

Direct LTZ output, LTC2057IS8 amp, +15V single-rail power input, Vishay foil resistors.

Total size is 80 x 40 x 40 mm.

VREF module(s) will be soldered on ratio board with dividers/multipliers and whole thing will be enclosed in thermal box.

Whole PCB module will be encapsulated in isolated box as well.

[Zucca]

Keep going! I was going to buy a reference source/meter (it is missing in my lab) but this thing to me is smoking hot, so my money are on hold now.
Very sexy project, Do you have any project  plan/schedules/milestones for the future?

[dr.diesel]

It would sure be nice is somebody here worked/owned a full call lab.  A project like this with proper written calibration instructions could be sent off for for a truly "custom" calibration on DIY hardware.

Gotta be somebody around here with a stack of equipment that wouldn't mind a small side business as "EEVBlog Custom Calibrator"?