uDR1399 - ADR1399-Based 10V Reference Module

[TylerPeppy]

Crosspost from a TE thread from earlier...

I Finally got around to building my new uDR1399 reference boards. They use bootstrapping and an LT5400 to set the amplifier gain. Both of those are things I would have wanted to see in the already available standards like the TEN-399 board and VREF10. I may be incorrect on what the VREF10 uses, so feel free to correct me. They have only burned in for about 72hrs so far, but seem to be humming along stabily. The results are from unit 1 of 3 that I have built.

I will edit and reiterate what is mentionned on the GitHub docs in spite of issues. As of current, do NOT connect a battery or other live supply to the unit while making connections. I tend to use a linear power supply and use the ON / OFF control of the channel to turn the uDR1399 unit on after connecting. Please add a 47uF electrolytic across the input posts if you wish to do so. There is a known issue where transients resulting from the hotplugging of the power cables can cause transients exceeding the LT3045's maximum inputs.

Unit 2 is reserved for Josh, but unit 3 can be spoken for if someone wishes to test a reference like this.

I measured using my DMM6500 on 10NPLC, 10 sample averaging DCV filter, AZ on, and 10GOhm InZ. Sensing is done through Banana-BNC adapters, and the internal wiring in my refs are twisted pair to the Pomona 3770 plugs. The STD-Dev running is using 1799 samples (or what 10NPLC for 30s would work out to, not including averaging filter) as suggested by iMO. The noise is high compared to the 34465A, but I have gotten ~4.15uV STD-Dev when disabling AZ, which is contrary to results we've seen so far in this thread.

More info can be found on my GitHub where I documented everything about the project: https://github.com/T-622/uDR1399-10V-Reference-Standard/tree/main

Stability test results can be found here: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/

Josh has one of the reference boards, and is testing it with a Keithley 2010, 34465A, and DAQ6510. He has reported numerous sets of findings, highlighting the temperature stability of the boards, and the overall noise within 30min windows on a Keithley 2010 to be <1uV std-dev. I am also attempting to build a thermal chamber to test the temperature coeffecient of the voltage output. This is proving quite difficult.

Tests Conducted:

1.) Josh's Initial Tests: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6332874/#msg6332874
Results Using 34465A For Best Stability:

Moving Filter:
2min - 10 Sample Moving Filter: 582nV std-dev
50min - 10 Sample Moving Filter: 727nV std-dev
75min - 10 Sample Moving Filter: 753nV std-dev
100min - 10 Sample Moving Filter: 751nV std-dev

No Moving Filter?:
018min - 2.92445uV std-dev
030min - 3.02086uV std-dev
040min - 2.90277uV std-dev
075min - 2.63610uV std-dev
100min - 2.46730uV std-dev

2.) Josh's Continued Test of 10 Sample Moving Filter: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333148/#msg6333148
Using 34465A:
* Note: These sets of tests featured a lot of thermal variance in the uDR1399's environment temperature, and the thermal influence on the reference was relatively minimal.

075min - 10 Sample Moving Filter: 785nV std-dev
128min - 10 Sample Moving Filter: 817nV std-dev
267min - 10 Sample Moving Filter: 1.61831uV std-dev
360min - 10 Sample Moving Filter: 2.04978uV std-dev
500min - 10 Sample Moving Filter: 2.92553uV std-dev

3.) Battery Bank + Keithley 2750: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333938/#msg6333938

No Filter:
40min warmup: 2.56582uV std-dev. I wouldn't count this one, as the reference was still warming up. However, it appeared to reach relative stability within a 16.67min window.

Test 4: 2min Filter ON: 811uV std-dev

Test 5: 2min Filter ON: 866nV std-dev

Test 6: Filter is 10 sample moving average (?) 1799 sample std-dev moving used in TestController.
2min - Filter ON: 347nV std-dev
10min - Filter ON: 677nV std-dev
30min - Filter ON: 848nV std-dev
100min - Filter ON: 668nV std-dev

4.) Tyler - DMM6500. 10 Sample moving avg filter, AZ ON, 10V range, 10PLC: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335186/#msg6335186

* Note: My reference featured here has been burning in for 432hrs at this point of measurement.

140min - Filter On: 2.906uV std-dev. Very little temperature variation

5.) Josh - Keithley 2010 Test: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335736/#msg6335736

Test 1: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
2min - 612nV std-dev
5min - 654nV std-dev
10min - 677nV std-dev
638min - 875nV std-dev. 2010 Std-dev as measured by TC is 3.1577428uV.

Test 2: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
5min - 1.2195uV std-dev

Test 3:
5min - 803nV std-dev
10min - 671nV std-dev
15min - 711nV std-dev

Tyler

[KungFuJosh]

You should link to specific posts by using the post title instead, for example: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335736/#msg6335736

The /200 doesn't show the same for everybody depending on their display settings.

Thanks,
Josh

[TylerPeppy]

You should link to specific posts by using the post title instead, for example: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335736/#msg6335736

The /200 doesn't show the same for everybody depending on their display settings.

Thanks,
Josh

I have made some edits, included all the test until now that have been taken... And fixed links including specific posts instead.

Tyler

[KungFuJosh]

Here's the last screenshot from Test 3 today before I shut it off.

Thanks,
Josh

[mawyatt]

Crosspost from a TE thread from earlier...

I Finally got around to building my new uDR1399 reference boards. They use bootstrapping and an LT5400 to set the amplifier gain. Both of those are things I would have wanted to see in the already available standards like the TEN-399 board and VREF10. I may be incorrect on what the VREF10 uses, so feel free to correct me. They have only burned in for about 72hrs so far, but seem to be humming along stabily. The results are from unit 1 of 3 that I have built.

Unit 2 is reserved for Josh, but unit 3 can be spoken for if someone wishes to test a reference like this.

I measured using my DMM6500 on 10NPLC, 10 sample averaging DCV filter, AZ on, and 10GOhm InZ. Sensing is done through Banana-BNC adapters, and the internal wiring in my refs are twisted pair to the Pomona 3770 plugs. The STD-Dev running is using 1799 samples (or what 10NPLC for 30s would work out to, not including averaging filter) as suggested by iMO. The noise is high compared to the 34465A, but I have gotten ~4.15uV STD-Dev when disabling AZ, which is contrary to results we've seen so far in this thread.

More info can be found on my GitHub where I documented everything about the project: https://github.com/T-622/uDR1399-10V-Reference-Standard/tree/main

Stability test results can be found here: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/

Josh has one of the reference boards, and is testing it with a Keithley 2010, 34465A, and DAQ6510. He has reported numerous sets of findings, highlighting the temperature stability of the boards, and the overall noise within 30min windows on a Keithley 2010 to be <1uV std-dev. I am also attempting to build a thermal chamber to test the temperature coeffecient of the voltage output. This is proving quite difficult.

Tests Conducted:

1.) Josh's Initial Tests: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6332874/#msg6332874
Results Using 34465A For Best Stability:

Moving Filter:
2min - 10 Sample Moving Filter: 582nV std-dev
50min - 10 Sample Moving Filter: 727nV std-dev
75min - 10 Sample Moving Filter: 753nV std-dev
100min - 10 Sample Moving Filter: 751nV std-dev

No Moving Filter?:
018min - 2.92445uV std-dev
030min - 3.02086uV std-dev
040min - 2.90277uV std-dev
075min - 2.63610uV std-dev
100min - 2.46730uV std-dev

2.) Josh's Continued Test of 10 Sample Moving Filter: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333148/#msg6333148
Using 34465A:
* Note: These sets of tests featured a lot of thermal variance in the uDR1399's environment temperature, and the thermal influence on the reference was relatively minimal.

075min - 10 Sample Moving Filter: 785nV std-dev
128min - 10 Sample Moving Filter: 817nV std-dev
267min - 10 Sample Moving Filter: 1.61831uV std-dev
360min - 10 Sample Moving Filter: 2.04978uV std-dev
500min - 10 Sample Moving Filter: 2.92553uV std-dev

3.) Battery Bank + Keithley 2750: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333938/#msg6333938

No Filter:
40min warmup: 2.56582uV std-dev. I wouldn't count this one, as the reference was still warming up. However, it appeared to reach relative stability within a 16.67min window.

Test 4: 2min Filter ON: 811uV std-dev

Test 5: 2min Filter ON: 866nV std-dev

Test 6: Filter is 10 sample moving average (?) 1799 sample std-dev moving used in TestController.
2min - Filter ON: 347nV std-dev
10min - Filter ON: 677nV std-dev
30min - Filter ON: 848nV std-dev
100min - Filter ON: 668nV std-dev

4.) Tyler - DMM6500. 10 Sample moving avg filter, AZ ON, 10V range, 10PLC: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335186/#msg6335186

* Note: My reference featured here has been burning in for 432hrs at this point of measurement.

140min - Filter On: 2.906uV std-dev. Very little temperature variation

5.) Josh - Keithley 2010 Test: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335736/#msg6335736

Test 1: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
2min - 612nV std-dev
5min - 654nV std-dev
10min - 677nV std-dev
638min - 875nV std-dev. 2010 Std-dev as measured by TC is 3.1577428uV.

Test 2: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
5min - 1.2195uV std-dev

Test 3:
5min - 803nV std-dev
10min - 671nV std-dev
15min - 711nV std-dev

Tyler

Nice work!!

Did you consider installing a 3D printed cover that surrounds the ADR1399 top and bottom. This helps keep any thermals away from the 4 Kovar leads. Recall someone mentioning that slow thermals can build up and move around even in a "closed case". Having a regulator near by doesn't help either, we know we have linear regulator on our dual 10V reference (LM399 and LTZ1000) and another single 10V LM399 reference. We added the the 3D print cover but can't say for sure it's helped, but surely doesn't hurt!!

These new LM399 replacements seem worthwhile, and results seem to indicate such! Think we'll be getting a couple to checkout and compare to the other references we have.

Anyway, nice work and keep posting updates/results

Best

[iMo]

..you may see nice "jump(s)" in the graph above, whether they come from the Keithley 2010 or the 1399 module cannot be distinguished in this setup..

[TylerPeppy]

..you may see nice "jump(s)" in the graph above, whether they come from the Keithley 2010 or the 1399 module cannot be distinguished in this setup..

Like I mentioned on another thread, I haven't seen the jumps in my boards I have left, as I cannot discern them from the DMM6500's noise. It will be interesting to see what jumps (if any) are seen on Martin's DMM7510.

Crosspost from a TE thread from earlier...

I Finally got around to building my new uDR1399 reference boards. They use bootstrapping and an LT5400 to set the amplifier gain. Both of those are things I would have wanted to see in the already available standards like the TEN-399 board and VREF10. I may be incorrect on what the VREF10 uses, so feel free to correct me. They have only burned in for about 72hrs so far, but seem to be humming along stabily. The results are from unit 1 of 3 that I have built.

Unit 2 is reserved for Josh, but unit 3 can be spoken for if someone wishes to test a reference like this.

I measured using my DMM6500 on 10NPLC, 10 sample averaging DCV filter, AZ on, and 10GOhm InZ. Sensing is done through Banana-BNC adapters, and the internal wiring in my refs are twisted pair to the Pomona 3770 plugs. The STD-Dev running is using 1799 samples (or what 10NPLC for 30s would work out to, not including averaging filter) as suggested by iMO. The noise is high compared to the 34465A, but I have gotten ~4.15uV STD-Dev when disabling AZ, which is contrary to results we've seen so far in this thread.

More info can be found on my GitHub where I documented everything about the project: https://github.com/T-622/uDR1399-10V-Reference-Standard/tree/main

Stability test results can be found here: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/

Josh has one of the reference boards, and is testing it with a Keithley 2010, 34465A, and DAQ6510. He has reported numerous sets of findings, highlighting the temperature stability of the boards, and the overall noise within 30min windows on a Keithley 2010 to be <1uV std-dev. I am also attempting to build a thermal chamber to test the temperature coeffecient of the voltage output. This is proving quite difficult.

Tests Conducted:

1.) Josh's Initial Tests: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6332874/#msg6332874
Results Using 34465A For Best Stability:

Moving Filter:
2min - 10 Sample Moving Filter: 582nV std-dev
50min - 10 Sample Moving Filter: 727nV std-dev
75min - 10 Sample Moving Filter: 753nV std-dev
100min - 10 Sample Moving Filter: 751nV std-dev

No Moving Filter?:
018min - 2.92445uV std-dev
030min - 3.02086uV std-dev
040min - 2.90277uV std-dev
075min - 2.63610uV std-dev
100min - 2.46730uV std-dev

2.) Josh's Continued Test of 10 Sample Moving Filter: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333148/#msg6333148
Using 34465A:
* Note: These sets of tests featured a lot of thermal variance in the uDR1399's environment temperature, and the thermal influence on the reference was relatively minimal.

075min - 10 Sample Moving Filter: 785nV std-dev
128min - 10 Sample Moving Filter: 817nV std-dev
267min - 10 Sample Moving Filter: 1.61831uV std-dev
360min - 10 Sample Moving Filter: 2.04978uV std-dev
500min - 10 Sample Moving Filter: 2.92553uV std-dev

3.) Battery Bank + Keithley 2750: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6333938/#msg6333938

No Filter:
40min warmup: 2.56582uV std-dev. I wouldn't count this one, as the reference was still warming up. However, it appeared to reach relative stability within a 16.67min window.

Test 4: 2min Filter ON: 811uV std-dev

Test 5: 2min Filter ON: 866nV std-dev

Test 6: Filter is 10 sample moving average (?) 1799 sample std-dev moving used in TestController.
2min - Filter ON: 347nV std-dev
10min - Filter ON: 677nV std-dev
30min - Filter ON: 848nV std-dev
100min - Filter ON: 668nV std-dev

4.) Tyler - DMM6500. 10 Sample moving avg filter, AZ ON, 10V range, 10PLC: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335186/#msg6335186

* Note: My reference featured here has been burning in for 432hrs at this point of measurement.

140min - Filter On: 2.906uV std-dev. Very little temperature variation

5.) Josh - Keithley 2010 Test: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6335736/#msg6335736

Test 1: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
2min - 612nV std-dev
5min - 654nV std-dev
10min - 677nV std-dev
638min - 875nV std-dev. 2010 Std-dev as measured by TC is 3.1577428uV.

Test 2: 10PLC, 10 sample filter enabled, 1799 sample std-dev-running enabled.
5min - 1.2195uV std-dev

Test 3:
5min - 803nV std-dev
10min - 671nV std-dev
15min - 711nV std-dev

Tyler

Nice work!!

Did you consider installing a 3D printed cover that surrounds the ADR1399 top and bottom. This helps keep any thermals away from the 4 Kovar leads. Recall someone mentioning that slow thermals can build up and move around even in a "closed case". Having a regulator near by doesn't help either, we know we have linear regulator on our dual 10V reference (LM399 and LTZ1000) and another single 10V LM399 reference. We added the the 3D print cover but can't say for sure it's helped, but surely doesn't hurt!!

These new LM399 replacements seem worthwhile, and results seem to indicate such! Think we'll be getting a couple to checkout and compare to the other references we have.

Anyway, nice work and keep posting updates/results

Best

This is a very good point. For the time being, I only have made closed-cell foam plugs that cover the entire ADR1399 package in the boxes. I have the board step files, so making a cover that clips through the thermal relief cutouts wouldn't be all that difficult.

I mentioned this to Josh, and might create one in my spare time with PETG to see how it performs. As for the regulator, I kept the original placement as I had in my Rev01 boards. I figured if I placed it close to the gain setting components, I might create some more unintended drift.

I might do some thermal imaging with a macro lens to see how terribly the thermals from the regulator creep across the board. I was going to originally make the board longer, and more spaced out, but the Hammond 80x54mm enclosure was the target, so I had little room to work with.

Tyler

[KungFuJosh]

..you may see nice "jump(s)" in the graph above, whether they come from the Keithley 2010 or the 1399 module cannot be distinguished in this setup..

True, but for the record, this K2010 also has an ADR1399. So the question is, which ADR1399?

[iMo]

..you may see nice "jump(s)" in the graph above, whether they come from the Keithley 2010 or the 1399 module cannot be distinguished in this setup..

True, but for the record, this K2010 also as an ADR1399. So the question is, which ADR1399?

.. do use a low noise DUT which does not jump with your K2010. Simplest not-jumping-low-noise-DUT would be a battery, around 10-12V one. Let it run for a week/two while continuously sampling/logging and you will see.. 

[KungFuJosh]

I think I should age my K2010 first for a year or two...after I'm done bastardizing it a little more.

[TylerPeppy]

I think I should age my K2010 first for a year or two...after I'm done bastardizing it a little more.

What I would be interested to see rather, is the difference between an organically-aged reference, versus an accelerated-aged reference baked inside an oven. I would be interested to see if accelerated aging for a week is for example, better than organic aging for a month.

[Kleinstein]

The LM399 and ADR1399 references have the internal heater and run relatively hot (supposedly some 90 C) already. This is already close to the glass temperature of some epoxies. So going higher with an external oven may not be a good option. For these 2 references aging is mainly be having the reference powerd.

[iMo]

..might be interesting to accelerate the aging of the entire "module" - there is the LT5400 divider, ie. how its epoxy package would respond to that..
PS: for example 12 days at 70C could be equivalent to something like ~1 year at ambient 25C..
PPS: the "jumping 1399" will not heal up (meaning it gets jumps-free) with aging, afaik..

[Kleinstein]

A main point with the epoxy packages, but also the PCB is humidity. Just heating for accelerated aging would also come with low relative humidity and this way drive out water. One may than see even more drift after the heated phase. When new the LT5400 like other plastic cases should start rather dry - high humidity in the plastic can cause issues with reflow soldering.

[KungFuJosh]

I hope @TylerPeppy can forgive me...I might have changed the connection types on mine.

Currently running off a 19.6VDC laptop power supply.

Thanks,
Josh

[TylerPeppy]

I hope @TylerPeppy can forgive me...I might have changed the connection types on mine.

Currently running off a 19.6VDC laptop power supply.

Thanks,
Josh

I would NEVER discourage modifications of my things, they *usually* only ever make them better. This is certainly one of those "cases" (ha!). I am eager to try these panels out as well, and potentially the Keystone 4109 versus my Pomona 3770s.

Tyler

[KungFuJosh]

I would NEVER discourage modifications of my things, they *usually* only ever make them better. This is certainly one of those "cases" (ha!). I am eager to try these panels out as well, and potentially the Keystone 4109 versus my Pomona 3770s.

I know you don't mind; I was being like the Greek god Facetious.

I like the 4109s better than the other ones we discussed, but the panels I made for the 4109s should accommodate all 3 and others.

I emailed you the data for 10 hours with the new panels/connections. I installed the gaskets normally. I think they're better as thermal gaskets than worrying about EMI. The panels still have continuity anyway.

[TylerPeppy]

I designed a little cap for the ADR1399 as opposed to using traditional, closed-cell foam. This supports a maximum insertion height of 9.3mm of the ADR1399's case to the topside of the board.

The current version is a little more loose-fitting than I want, so I am going to dial down the tolerance compensation that I used in Fusion, but it seems to work fine. Once I have a tight-fitting part, I will attach and see if the performance changes.

Tyler

[mawyatt]

Don't think you want a "tight fitting" cover, as this could cause some stress on the IC package. Ours is loose fitting (actually doesn't touch the case) and the cover is secured by the PCB slots (glued in place).

Best

[TylerPeppy]

Don't think you want a "tight fitting" cover, as this could cause some stress on the IC package. Ours is loose fitting (actually doesn't touch the case) and the cover is secured by the PCB slots (glued in place).

Best

The original one had 0.2mm inwards offsets for the thermal relief holes so that the plug could mount properly using the cutouts. When I said "Tight fitting", I meant to refer that there is a bunch of room inside the case for the components, but the mounting holes were too loose. I have reprinted with 0.1mm and 0mm offsets to see how well. I typically use 0.2mm offsets on my Prusa MK3S+ and a 0.4mm nozzle, but sometimes these need tweaking.

Edit: The using 0mm of tolerance around the thermal cutout slots made for a snug but not overly tight fit for the cap. There is plenty of room around the component, and the ADR1399 is not impacted by the cap either.

- Tyler

[TylerPeppy]

New thermal isolation caps have been installed around the 1399s, see the previous post for the .3mf file for the caps.

Updated data with these installed can be found here: https://www.eevblog.com/forum/testgear/bench-dmm-stability-comparisons-with-testcontroller/msg6348896/#msg6348896

[TylerPeppy]

The new laser-cut and engraved end panels have been installed, and look great. These are Hammond Mfg. 1455CSALBK-10, and could be engraved on a cheap hobby CNC or laser engraver machine. Josh has the cutout files for multiple post types, and I will leave it up to him whether he wants to share them or not.

Tyler

[KungFuJosh]

I dunno if I'd share that or not. I'll think about it.

A CNC might be able to cut that, but not a hobby laser. You need a decent fiber laser to cut metal. Any fiber or CO2 laser can do the engraving, but it's easier to setup on a galvo (as opposed to a gantry).

[TylerPeppy]

I dunno if I'd share that or not. I'll think about it.

A CNC might be able to cut that, but not a hobby laser. You need a decent fiber laser to cut metal. Any fiber or CO2 laser can do the engraving, but it's easier to setup on a galvo (as opposed to a gantry).

Yes, of course. The cutting for an engraver might be easier to just mark the drill outline, and then drill by hand later if needed. I suspect this project will never garner nearly enough attention for that to be a thought/worry though.

[Hydron]

You could just order a PCB with black solder mask and the right cut outs and silkscreen text on it. Not as nice but much more accessible.