Author Topic: OSHW - 24bit ADC measurement system for voltage references (Read 36992 times)

liquibyte · « **Reply #25 on:** April 28, 2014, 01:20:13 am »

Damn what a clusterf***. Here. The next person that posts Eagle schematics without realizing there's a image export under File, I'm going to spam them with PM's until they beg the mods for me to stop.

branadic · « **Reply #26 on:** April 28, 2014, 07:50:31 am »

Quote

Why have these cut-outs of the PCB around the IC on the bottom side ?

I followed the suggestion given in this video for the reference package http://www.linear.com/solutions/1129

Quote

Branadic, thanks for sharing the board. Did you have a chance to build it yet?

Not yet finished, cause I'm currently on vacation.

CaptnYellowShirt · « **Reply #27 on:** April 28, 2014, 05:30:25 pm »

Subscribed!

bingo600 · « **Reply #28 on:** April 30, 2014, 06:51:14 pm »

Thanx Branadic

Looks interesting.

I do hope you are going to continue this thread , despite the IMHO quite unreasonable people attacking.

The first post about a .pdf request was ok , but the others .....

Keep on the good work , and you did get me interested immediately. Wo.any further seed.

/Bingo

EEVblog · « **Reply #29 on:** May 01, 2014, 07:59:46 am »

Quote from: bingo600 on April 30, 2014, 06:51:14 pm

I do hope you are going to continue this thread , despite the IMHO quite unreasonable people attacking.
The first post about a .pdf request was ok , but the others .....

I agree. A first polite request to post in a common format like PNG or PDF is fine (and fully expected, Eagle isn't the way you'd normally show of your project in a forum).
But yes, the rest was definitely not cool

Play nice.

branadic · « **Reply #30 on:** May 30, 2014, 06:14:36 pm »

I've finished (soldered the missing parts and cleaned them with ultrasonic and IPA) the boards today. Next step is to flash the microcontroller, to do some test and write a pc software (I guess that will be done in Matlab/Octave). Not to forget to mention that the boards will be packaged, I've choosen Hammond 1455C801.

I made some minor changes in the board files and will update them (first post).

After that the next board will be layouted, the LTC1043 divider.

Feel free to duplicate...

Neganur · « **Reply #31 on:** May 30, 2014, 08:30:30 pm »

sehr geil, danke!

branadic · « **Reply #32 on:** June 15, 2014, 02:05:44 pm »

Updated the files in the first post, because the value of C5 in 24 Bit ADC.sch was changed. I also modified the dimension layers in both files somewhat, so that the board better fits into the Hammond cases.

Current status:
I calibrated the NTC of my both ADC boards using a SHT75 reference sensor, but will verify my calibration soon with an environmental chamber again.
After a first 300h preaging periode with extended load on the LT1236LS8 I will measure the current reference voltage and the TC on a Keithley 2002 with an oven. Andreas suggested to use a cross-adapter and to measure the TC with one board heated while the other stays at room temperature. Thats for sure a good possibility if you don't have access to calibrated gear.
I'll than check linearity using the resistor divider board and the voltage reference of both ADC boards, because of the lag for the 2:1 LTC1043 board to do that with my LM399 reference, that is now running for more than one year.

I also started to write a Matlab program, that requests the data via serial interface and display them in a diagram. That works fine up to now. Next step is to build a gui around that code and make it a standalone application.

branadic · « **Reply #33 on:** June 19, 2014, 01:44:17 pm »

While the voltage reference is still preaged with extended load for another few days until 300h have passed and in most places of germany christian people celebrate "happy cadaver" and the none christian people have a free day too, I decided to do something 'bit off topic to this theme. I started to build myself a small temperature chamber, because our climate chambers at work are in use for the next weeks. Well, no problem at all.
I found all the required parts for that in my home lab, a Styropor box with inner dimensions of 270x165x160mm³ and 30mm wall thickness, a heatsink with a size of 150x150mm², a CPU cooler with fan, a Peltier element with size 40x40mm² 12V/30W, a heater foil and a fitting sheet aluminium, two laser cut D-Sub panels and several D-Sub connectors.
All that is still missing are a few mechanical components, that will be finished by the end of the week.
A solution to read out two SHT25, less then half the price of a SHT75 and somewhat more accurate, via VCOM with additional 3x16 LCD display is currently in progress and will complete the setup.
One SHT25 is to track the chambers temperature, the second one is for the device under test, e.g. the NTC of the ADC board inside its case to calibrate it.
It would have been deluxe if the temperature is controlled via a µC or something similar, but as such controllers are very expensive I will start using one of my lab power supply and set the temperature by manuelly adjust the current on the PSU.

I've also drawn the panels for the Hammond cases and will post them here soon in DXF format. I used the free front panel designer by Schaeffer AG for that purpose. Pictures will follow when finished.

branadic · « **Reply #34 on:** June 22, 2014, 10:10:49 am »

It's not yet finished, but here is a picture of my temperature box with heater foil and TEC. A few D-Sub9 connectors as feedthrough the wall.
I will order some pcbs next week and put a small ADC extension board to the batch with PT1000 + AE / Vishay 10K 0.01% 5ppm

eBay auction: #380191504233

Couldn't find PWW available from stock at lower cost.

branadic · « **Reply #35 on:** July 05, 2014, 10:39:23 pm »

Found an intersting link today and I can't remember that it was linked somewhere on eevblog before:

Long term characterization of voltage references

rs20 · « **Reply #36 on:** July 06, 2014, 03:34:26 am »

Just a thought, and feel free to point me to another thread if this has already been discussed, but the cut-outs around the voltage reference seem a bit sub-optimal? There are little PCB bridges on all four sides of the voltage reference "island", which means that that island can't freely thermally expand or contract. I don't see that thermal stresses will be reduced much at all. The suggestions in the relevant datasheet, to have a cut-out running continuously around all three sides of the reference so that it's freely cantilevered in space seem more effective and more space-efficient.

branadic · « **Reply #37 on:** July 06, 2014, 06:34:36 am »

Please refer to the video available by Linear Technology (Brendan Whelan) that suggest this 4-sided cutout geometry to give similar or better performance compared to the 3-sided pcb tab cutout:

http://www.engineeringtv.com/video/Linear-Technology-LS8-Package-I (watch minute 6:34)

Furthermore you have the advantage of running the traces over the corners instead of having all traces at one side. So, I preferd using the 4-sided cutouts and while the device runs within +10°C ... +40°C I'm not affraid of any form of cutouts.

rs20 · « **Reply #38 on:** July 06, 2014, 07:08:30 am »

Point taken, although I can't reconcile this with my intuition. Interesting that the context was FR4 swelling due to humidity. I'd be curious if anyone's come across some actual studies on this -- all the video tautologically says is "A can give similar or better performance than B", which doesn't rule out A being better than B at all.

</ot>

Andreas · « **Reply #39 on:** July 06, 2014, 08:06:29 am »

Quote from: rs20 on July 06, 2014, 07:08:30 am

Point taken, although I can't reconcile this with my intuition. Interesting that the context was FR4 swelling due to humidity. I'd be curious if anyone's come across some actual studies on this -- all the video tautologically says is "A can give similar or better performance than B", which doesn't rule out A being better than B at all.

</ot>

You are invited to do some studies on this.
I for my part prefer the "dead bug in hole mounting" method for LS8-devices. (do not try this in production).
https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/msg278825/?topicseen#msg278825

A other method would be mounting the reference on a ceramic substrate with the same CTE than the housing of the reference.
Or using a flexible part of a PCB (kapton foil) for the mounting of the reference.

With best regards

Andreas

branadic · « **Reply #40 on:** July 06, 2014, 08:23:23 am »

My board is designed with the opportunity to cut the reference part of the pcb and replace it with a flexible substrate, such as 100µm FR4. I had that in mind when I designed the board, but my pcb manufactor doesn't offer 100µm substrates and I don't have such a thin substrate for home-processing. I wouldn't use capton, it's expensive and there is no need for it.
The method of Andreas using "dead bug" is a way, but I wanted to avoid such solutions.

A complete board made of ceramic would be very expensive I guess and you'll never now what ceramic the package is made of to match the CTEs as close as possible.

Edit: In addition to the linked publication, I wonder why they used AD587 (10V) instead off AD586 (5V) and why they have used LTC6655BHM (MSOP8) instead of LTC6655BHLS8. I would expect somewhat better performance of that devices.

Andreas · « **Reply #41 on:** July 06, 2014, 08:34:26 am »

Quote from: branadic on July 06, 2014, 08:23:23 am

The method of Andreas using "dead bug" is a way, but I wanted to avoid such solutions.

A complete board made of ceramic would be very expensive I guess and you'll never now what ceramic the package is made of to match the CTEs as close as possible.

I can understand this: this mounting method needs much time in assembling.

I did not think about a complete board: but what about a small ceramic with 4-5 pins (mechanical decoupled from PCB) on one side (upright mounting) together with a small temperature controller for the reference on the back side.

Note that I have also a influence of the PCB on the reference on ceramic references like VRE3050AS when I solder all PINs around the ceramic case to the PCB.

Quote from: branadic on July 06, 2014, 08:23:23 am

Edit: In addition to the linked publication, I wonder why they used AD587 (10V) instead off AD586 (5V) and why they have used LTC6655BHM (MSOP8) instead of LTC6655BHLS8. I would expect somewhat better performance of that devices.

Edit:
AD587 and AD586 should not differ much. (in LQ or UQ option).
For the LTC6655BHM I think that they simply started the study before the LS8-Option was available.
On the other side I have large hysteresis with the BHM package. I do not think that the same chip with similar die attach will behave much better in LS8 housing. (would be a interesting test).

With best regards.

Andreas

SeanB · « **Reply #42 on:** July 06, 2014, 09:16:44 am »

You can get flexible boards made cheaply now, so probably you can get a sheet made with multiple small adaptors, screen the solder paste on them and mount multiple devices then reflow them, then afterwards simply cut them out of the sheet when needed for use. Easy to have the connections on one side with the space for a thermistor and decoupling caps as well, then simply have some larger pads with a large solder ring on them ( probably on a 0.2in pitch so you can either solder the board to the board direct or use a pin header) so you can mount them. You can even include a guard trace around the unit and a copper trace for a cutting guide. Assemble with ordinary kitchen scissors.

branadic · « **Reply #43 on:** July 06, 2014, 09:27:38 am »

Quote

For the LTC6655BHM I think that they simply started the study before the LS8-Option was available.

Could be the reason, yes.

Quote

On the other side I have large hysteresis with the BHM package. I do not think that the same chip with similar die attach will behave much better in LS8 housing. (would be a interesting test).

I would expect better results even in the lower 0.1Hz region as the die is glued with only the bottom side to the package, while in the BHM package the die is surrounded by the thermoset and glued to the lead frame. But this is all speculation.

Quote

You can get flexible boards made cheaply now

This is true only for manufactors outside of germany

While complete devboards are offered by Asian you wouldn't even get the board fabricated nor the parts from a distributor for a comparable price here. And 100µm substrate is called to be "special" not standard and you have to pay for something "special".

branadic · « **Reply #44 on:** September 16, 2015, 07:33:10 pm »

Added the files for the 2:1 Resistor Divider with LT5400-1 in the first post. Finished the assembly for this board today. Still to do: layout the 2:1 divider with LTC1043 for the Hammond case.

branadic · « **Reply #45 on:** June 24, 2016, 07:30:56 pm »

Since LT1027DxLS8 is available and seems to be a good choice as a reference for this system here's a version of the ADC board with this reference replacing the former LT1236LS8. Hope you like it.
I will build up this version as well, as it is interesting to see what it's about.

branadic · « **Reply #46 on:** June 28, 2016, 07:28:34 pm »

I plan to order pcb's. If someone is interested in pcb's, version with LT1236LS or version with LT1027LS8 and/or the resistive divider board, please write me an obliging email and let me know how much boards of which version. The more we get together, the cheaper the boards.

Boards are made for Hammond 1455 cases.

Deadline is the 10.07.2016.

kutte · « **Reply #47 on:** June 29, 2016, 08:51:58 am »

Quote from: branadic on June 24, 2016, 07:30:56 pm

Since LT1027DxLS8 is available and seems to be a good choice as a reference for this system here's a version of the ADC board with this reference replacing the former LT1236LS8. Hope you like it.
I will build up this version as well, as it is interesting to see what it's about.

thank you, but once again: is it really that difficult to include a pdf-version of the schematics?
Kutte

branadic · « **Reply #48 on:** June 29, 2016, 07:40:09 pm »

Quote

is it really that difficult to include a pdf-version of the schematics?

Give me a good reason what you want to do with a pdf? I don't see any advantage. The only argument I can agree are Gerber production files.

kutte · « **Reply #49 on:** June 30, 2016, 10:08:58 am »

Quote from: branadic on June 29, 2016, 07:40:09 pm

Give me a good reason what you want to do with a pdf? I don't see any advantage. The only argument I can agree are Gerber production files.

how can I have a look at the schematic ( filename.sch ) without installing Eagle ?
Kutte


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: OSHW - 24bit ADC measurement system for voltage references (Read 36992 times)

Share me