Project KX : DIY calibrator / reference source/meter

[TiN]

That was just stress test check.
As of spikes, thats home appliances turned on/off during time.

I got redone connections, now vrefs psu and scanning keithley 2002 are powered from standalone power breaker.

Here attached results with NPLC 50 and three refs. Temperature readings: max6610 on pcb.

Power supply is usual linear with 7815 with some 2200uF caps and 7805 (for MAX6610s)

http://dev.xdevs.com/attachments/1090/log_nplc50.png

Raw data excel


http://dev.xdevs.com/attachments/download/1091/data_log_nplc50.xlsx

[TiN]

Revive, revive! Bring back some life into this thread again.

Went local electronics market today, and got little neat chassis for KX, just for little over $40 USD.
It's 19" rackable, 178mm height, 480mm length, 365mm depth metal enclosure with aluminum front and back panels.
I put one of my 2001's on it for size comparison.





Nothing fancy inside, just panels mounted together and U-shaped cover.



Initial idea for placement.



Alternative way is to make big backplane board, on which separate modules would be plugged in vertically. Also pretty sure it will be two level design, digital stuff on top and front panel, analog stuff on bottom in separate thermally shielded area. LEM current sensor would likely go outside chassis into separate block.

I do like idea of status LEDs around input/output ports, like on Fluke 5730A calibrator. Sometimes can help to spot that trying to source voltage into two wires, while unit configured in 4-wire mode.

Now thinking about how to make low-noise high voltage power supply for output stage (+1200/-1200 VDC with at least 200mA current rating).

[Andreas]

Now thinking about how to make low-noise high voltage power supply for output stage (+1200/-1200 VDC with at least 200mA current rating).

Mhm, 500VA transformer (minimum) and 300W / 1500V power stage.
Does not look like a weekend project.

With best regards

Andreas

[TiN]

I meant 1200V internal voltage, to be able source 1000V. Would like to be able use that for high resistance measurement function. On other corner - high current low voltage converter. But that is easier.

[ManateeMafia]

Do you plan on winding your own transformer or do you have some HV transformers waiting on a new project?

The front panel of the enclosure has a lot of room. Are you tempted to add a touch screen along with the VFD?

[TiN]

No, I don't see much in idea of touchscreen for DMM. This project already overdone way more than it practically should be.

[TiN]

Anybody know source for good copper banana jacks/binding posts for input/output terminals? Best if have different colors as well, but even red+black would do the job. Digikey only have Pomona 3770 , and it's binding post, while I'd prefer banana jacks. Ebay have nothing as well, if not counting audiophoolery stuff.

For low-current stuff I have triax already, but need have copper connectors for general stuff as well.
Not-so high price preferred as well (3-6$ USD okay, 30-60$ not so much).

[ManateeMafia]

TiN,

My replacement panel from Keithley is supposed to have the press fit jacks. Perhaps they are available as a separate part? I have found some of the listed parts are not available like the bare front panel.

I have seen the black/red binding posts from Fluke on Newark or Mouser. I can look later if you need the number.

[TiN]

Still not much options there, just found AIE 125061/125065, but no material spec in datasheet.

[quarks]

Anybody know source for good copper banana jacks/binding posts for input/output terminals? Best if have different colors as well, but even red+black would do the job. Digikey only have Pomona 3770 , and it's binding post, while I'd prefer banana jacks. Ebay have nothing as well, if not counting audiophoolery stuff.

For low-current stuff I have triax already, but need have copper connectors for general stuff as well.
Not-so high price preferred as well (3-6$ USD okay, 30-60$ not so much).

for your calibrator I would use binding post and go for good low EMF ones.
In my own search for low EMF stuff, I found the Pomona 3770 the easiest to find/buy and also the cheapest.
Others very good (but more expensive) ones are from Multi-Contact PK4-TS (around 12$ each), Lowthermal Model 2758 Binding Posts (around 25$ each) and IET BP1000 (around 60$ each)

[lukier]

Anybody know source for good copper banana jacks/binding posts for input/output terminals? Best if have different colors as well, but even red+black would do the job. Digikey only have Pomona 3770 , and it's binding post, while I'd prefer banana jacks. Ebay have nothing as well, if not counting audiophoolery stuff.

Maybe (if this is for the calibrator) just buy Agilent part 34401-62121 (input block assembly), I see it's < 15 GBP on their website. It seems quite economical and hopefully has high quality copper and is goes straight to the PCB.

[TiN]

Hm, that's a good idea. Let me check quotation.

Meanwhile...

Thermal chamber for noise-measurement preamp/reference design :

Inputs/outputs - BNC or TNC jacks
Material : Aluminum
Dimensions: 110 x 62 x 25 mm



Here's prototype:



With Dual LM399+7V/10V/20V ratio board



With LTZ1000 board. Variants of TNC input connectors mount.



BNC and TNC in chassis locations



Overall:



Module will accept TEC on outer side to control temperature inside slowly and nice.

[TiN]

Datalogging to online website is working now.
Since often measuring some effects of LTZ1000-reference require tremendous amounts of time, it's not always easy to visualize history of measurements data.
Yes, Excel works fine for simple stuff, but when it comes to displaying graph with 10^5 points, Excel is no tool to work with.

And sometimes measurement gear is in another room/building/city/country, without physical access to plug cable in to read data. Here's concept which I'd like to try using few tools:

* LabView to get data, process it, and save
* D3.js to draw graph from CSV
* FTP-server and online website to chart measurement data and show it on WWW.
* Data source, in our case it's two DMMs , Keithley 2001 and Keithley 2002 with shorts on inputs, measuring DCV.

Workflow:

1. LabView accesses DMMs via NI USB-GPIB-HS and reads data.
2. Data is formatted into CSV string and written/append to file on local HDD
3. LabView takes CSV-file and sends to FTP. This file is visible publicly on http://xdevs.com/static/data/test.log
4. Page http://xdevs.com/datalog/ runs D3.js javascript library to read CSV file test.log to input data
5. D3.js plots SVG graph online

Screenshot of everything on one screen - 4480x1600 pixels

[JBeale]

Amazing project, I'm enjoying all the photos and writeup. Nice work!

[TiN]

It may look ugly, but just as test able to get 10V within 2ppm. Will it stay that way? Let's see, datalog capture shown realtime by URL.

Source: Old LM399 from K2001 board
Measurement: 10V output = K2002 (not calibrated), VREF 7V = K2001 (calibrated).

Ratio circuit - from bbs.38hot.net forum, with buffer amp (LTC1050) and ratio amp (LTC2057). Resistors are PTF56 and RNC55, 25-100ppm/C.

[Building_5]

There are some nice gold plated banana / spring loaded wire holder connectors on eBay right now at a rational price.   They're Multi-Contact 23.0410-21 and 23.0410.22, twenty of each for USD $36.  http://www.ebay.com/itm/20-red-20-blk-Banana-and-Tip-Panel-Mount-BINDING-POST-ELECTRICAL-CONNECTORS-/291415440013?_trksid=p2054897.l4275
I've bought this type of connector from this seller and have no complaints.

Anybody know source for good copper banana jacks/binding posts for input/output terminals? Best if have different colors as well, but even red+black would do the job. Digikey only have Pomona 3770 , and it's binding post, while I'd prefer banana jacks. Ebay have nothing as well, if not counting audiophoolery stuff.

For low-current stuff I have triax already, but need have copper connectors for general stuff as well.
Not-so high price preferred as well (3-6$ USD okay, 30-60$ not so much).

for your calibrator I would use binding post and go for good low EMF ones.
In my own search for low EMF stuff, I found the Pomona 3770 the easiest to find/buy and also the cheapest.
Others very good (but more expensive) ones are from Multi-Contact PK4-TS (around 12$ each), Lowthermal Model 2758 Binding Posts (around 25$ each) and IET BP1000 (around 60$ each)

[peter.mitchell]

You could ask franky (iloveelectronics on the frum) about the binding posts used on Daves uCurrent Gold. IIRC he sourced them for him.

[Smith]

For generating high voltage consider Emco hv modules (http://www.emcohighvoltage.com/). They are quite small, and low noise, I've used some of them and they work great.

Btw this is a really cool project  O0

[TiN]

Since at least one KX LTZ ref is out in the wild, here's related data for module:

Schematics in PDF
Schematics archive for KX LTZ1000 reference, Rev B01

Reference itself:



Temp sensors:

[TiN]

Alright, time to revive something..

Here's what I was tinkering lately about, repeating infamous Linear AN86, as it looks to me that no one actually built it before.

But since I went with modern parts, hopefully with better end result, my schematics which I'd like to present is not direct copy.

Both DACs: 18-bit Linear LTC2756 (about same topology as LTC1599, as far I can see).
Feedback ADC: 24-bit LTC2442
Reference: LTZ1000 direct input +7.2 for DACs, Divided (resistors) and inverted (LTC1043) to 2.5000/-2.5000V for ADC.
Digital domain: ALTERA MAXII CPLD
Output stage: That part later, currently task is to build stable -14...+14V DAC using this module.



There is also dual DAC available from Linear, LTC2758, but it's price is more than double of 2756's.
I have DACs and ADC already on my table.

Please, take a look on schematics, and let me know where I failed to understand analog magic
Hopefully we can correct mistakes and I could make few PCBs to try it on. Best contributors can count on PCB (no parts, tho) too

[Kleinstein]

The Ref. signal for the DACs does not need to be that stable, but should be very low noise. So I would prefer a buffered and filtered signal, not directly the LTZ1000. Long time drift is well covered by the ADC. Similar the choice of OP's does not look right. At the DAC outputs, I would prefer something like an LT1007 - low noise, not an zero Offset OP. Again an offset and moderate (slow) drift does not really matter, but noise may matter.

There is not real need for two 18 Bit DACs - the  LTC2756 is so good, its hardly needed to do the correction. The INL of the ADC is not that much better. So the correction DAC might be a lower cost version - something like 12 Bit might be enough. The overlap can also be smaller - so R42/R43 could change to something like 100 Ohms and 1 M. Only for doing a internal consistency checks (and possibly corrections) of the ADC might need the 2 nd high quality DAC.

The input of the ADC may need some buffering / capacitor at its input. 8 K looks like rather high impedance to me. Don't know that particular ADC, but some of the sigma deltas are a little tricky if you want the full INL performance.  Also some protection against overshoots during powerup may be helpful. The amplifier may also like some feedback capacitor.

Depending on how clean the digital part is, it may be helpful to have filtering against very high frequency noise. This may be just a stage of "slow" logic like 74HCTxxx. The SPI bus of the 2 DAC can likely be put together, just keep the CS signals separate. I am not sure if an FPGA is the right chip to control the DACs - a µc might be better suite, possibly using a lookup tables and more detailed filtering for the correction values. The ADC is not that fast anyway.  I don't know if something like a 50 Hz / 60 Hz sine output option is planed - this might need a little more filtering.

[TiN]

Whole system is based on reference stability, so target to get absolute accuracy as well after calibration to known voltage standard.
I already have both DACs, so on that follow AN86. In future feedback ADC likely to be changed to discrete multislope integrating ADC (that part in design still), driven by same CPLD.

For op-amp choice, guilty, yes. I am not that familiar with all op-amp variety, definitely need to look into proper ones for low-noise.

There will be micro behind CPLD, so not to worry about math. I'd like use CPLD since it allows fixed delays/clock distribution, as well as parallel loading and triggering data.
I am thinking of having digital interfaces run thru isolation, but that's probably overkill. Whole board will be in enclosed metal box with thermal control, to avoid thermoelectric effects/drift.

[Smith]

Doesn't the Keithley 2001/2002 adc module feature full dual slope adc, wouldn't that be usefull? Guess they did a good job getting 8,5 digit, but im not sure about driving the module.

I'll like to know how you are going to design your psu. Are you fully splitting the analog and digital power lines?

I guess some low pass filtering wouldn't hurt both the adc input and dac output. Not sure if active or passive would be the way to go ib this aplication.

Last thing. Wouldn't it be usefull driving the digital bus lines trough an isolator? There a a lot of good fast and cheap isolators available. This can prevent a lot of noise.

[Andreas]

Hello,

the LT5400 arrangement at the feedback amplifier looks to me as if the LT2442 would have to handle +/-10V at the analog pins.
(at +/- 2.5V maximum range).

I would also try to get the divider out of the same package of the LT5400.
Otherwise it is not likely to get the 0.2ppm/K ratio drift.

I also recommend to do some EMI-Filtering of the feedback amplifier at the input.
Also the multiplexer input of the LTC2442 might be sensitive to external noise.

The LTC2442 has also a 0.2ppm/K typical gain drift.
So it would be a good idea to use a temperature sensor.

With best regards

Andreas

[Theboel]

This project is far away from my budget but may be I can handle a "cheaper" version of the reference unit.
any sugestion for Visahy resistor replacement ?