MV-106 repair and subsequent modification

[Insatman]

Greetings,
Some weeks ago I bought an MV-106 voltage standard.   The EDC( now Krohn-Hite) unit was apparently manufactured in 1992 based on the chip date codes.    When I first turned the unit on it was spot-on according to my HP34401A.  But within a few minutes it began to drift up.   Overnight it was 130uV high.  The next morning another 100uV high.  Over the next 10 days the unit drifted by almost 800uV.   Standard deviation as measured by my HP3457A was 100-200E-6 with total spread by as much as 600uV per overnight run.   Something was clearly wrong.

I opened up the unit and saw nothing obvious.  A bit of dust but no swelled caps or other tell-tale signs.
Looking at the schematics in the manual I noticed that the chopper op-amp shown is actually two ICs on a daughter card.  No schematic is supplied for that so I reversed engineered a schematic.  The two ICs are still available for Digikey so I ordered a set.   Cost about $6 for the pair.

When the chips came in I installed them and while I was at it, checked every capacitor on the daughter-board and main-board for ESR.   All caps were good.   I was able to measure value and leakage of the caps on the daughter board once I removed the ICs and they all checked out.   I installed the new ICs and ran the unit.   I quick check of the Zener diode voltage showed is well above it’s marked voltage (~20mV or so…I lost my notes on that).   I left it unadjusted for now, just wanting to compare unit stability without multiple changes. 
Overnight stability improved.  The first night only slightly from ~200uV total variance to about ~160uV.  The following night however the total variance was only ~60uV.  This is likely due to the fact that insufficient warmup before starting the first night’s measurement run.   Standard Deviation (SDEV) was 8.5E-6.
It occurred to me that I could likely replace both IC’s with modern replacements with improved specifications.  So after some research these were ordered.   Cost about $16 for the pair from Digikey.
The unit was tested for a further 3 days with the replacement original part number chips and overnight variance was between 60-110uV.   This is reasonable considering the 8 hr stability spec of 0.001% of range +2uV or 102uV per 8 hours.   Note I only have 10uV resolution except for math functions on the HP3457A.
The amplifier daughter-board was modified as per attached schematic. 
The TC7652 chopper amp was replaced with an LT1052CN part.
The NE5534 was replaced with an OP27EPZ. 
Two resistors were replace with low temperature coefficient types and a damping resistor was added to the capacitor on U2.   All three 100nF film capacitors were replaced with 100nF polypro types. 
Bypass caps were added to both ICs and a larger bypass cap was placed across the +/- 15V supply rail.
No circuit modifications were made to the main board at this time.
The only other modification to the MV-106 was the installation of a foam box around the sensitive portions of the main board.  See pictures.   This should thermally isolate the voltage reference and amplifier boards as well as stop any air currents.   The foam block is hollow with a roughly 1x1" hole around the heatsink of the main transistor (TIP48) to allow for heat to escape.  A simple 1/2" foam pad is also applied to the bottom of the board, thereby completely enclosing the area except for the hole previously mentioned.  The block is held in place by a single nylon tiewrap.  A 5/16" hold was drilled in the metal frame of the MV-106 to accommodate this tiewrap.

Total cost of the parts involved in only about $20 and everything is available from Digikey.

After turn-on the unit drifted about 80uV and settled down.   After an hour of stability, I adjusted the Zener voltage of the unit to it’s marked specification and then proceeded to follow the MV-106 calibration procedure for scale and offset adjustment.  I did not complete the portion of the procedure for linearity of the 1st decade switch resistors as they seemed quite well adjusted already.
An initial 4 hour stability test yielded promising results.  SDEV was only 5.96E-6 and total variance was measured at 20uV on the 34401A and 32.7uV on the 3457A.
A subsequent overnight test (~16 hours) yielded the following results.  Total overnight variance (spread) was 40uV as measured by both meters.  SDEV was 6.17E-6 indicating that the vast majority of the time the unit was within 20uV of nominal.    Overall I am quite pleased with this result.
Long-term drift is still unknown but hopefully it will also be greatly improved.

Attached are the original reversed engineered daughter-board schematic.
The Modified version Schematic with notes.
Pictures of the modified daughter-board and foam enclosure installed.

Insatman

[USMC_Spike]

Magandang hapon, Instaman,

Nice work, and simple. 

Wondering do you have the GPIB for the Hp3457A?
or did you use functions from the Hp34401A?

Or combine them with a GPIB device?

The MV-106 is kind of large, or it looks large,
heave or not bad.

Did you do  the usual cleaning of the pots, etc?

Aalis na ako,

Spike

[Insatman]

Hi USMC_Spike,
I'll answer your questions as best I can.

Yes I have GPIB for the HP3457A but I don't currently use it.  I have as yet been unable to make it work with the hardware I currently have.  It is an ongoing thing on my to do list.  The values I used for the analysis were obtained using the front panel.  You can access the 7.5 digit capability of the 3478A when reading registers such as MEAN or HI, LOW, SDEV...etc. 

I did use the MEAN, HI and LOW functions of the HP34401A accessed through the front panel. 

Note the 3478A and HP34401A agree with an average offset between them of about 140uV.   This varies by a few 10's of uV but generally they track each other pretty well.  I keep records in a hand written log book and transfer the data to Excel occasionally.

the MV-106 is a 5 1/4" high rack chassis about as deep as the HP3457A.  It's pretty light as the inside is mostly empty space.  I bought the unit from a very reputable used instrument dealer in the US.  I did not do any extensive cleaning, simply blew out some dust and used some tuner cleaner on the switches, rotating them  back and forth several times after spraying.  Getting the unit through Philippine customs was a bit of a nightmare.

I now have data on the drift of the MV-106 unit from mid August until now.  I will write a post on this thread in a few days.   As a spoiler to that post, I am generally happy with the performance of the unit after the mods.  I do suspect that the drift I have observed is probably due to the unit acclimating to the higher humidity here in the Philippines.    The MV-106 was originally from New Mexico in the US where the average humidity is very low.  More on that when I post the data.

Insatman

[texaspyro]

I just finished restoring the MV-100 I got off a recycle pile.   It is a little bit drifty, but stays well within the 0.003% of range spec and is improving with time.  The room temp is changing over around 4 degrees C.  The 10V range seems to drift around +/- 50 uV.

Have you checked the voltage error between the odd and even first decade settings?   When I cal the first decade to 10.000000V the 9V range is around 9.000050V.  The linearity cal procedure says to set the even voltage then go back and tweak the pot for the odd voltage.  I split the difference between the two settings to 9.999975 and 9.000025V, etc.

Mine has the potted chopper amp, so I would need to build a complete new board.

I wonder how well replacing the 1N829's with LM199's would work?  I also have a bunch of the 2DW232/233 ultra low noise zeners from China.

The specified reference zener voltage setting is the value where the zener has it's lowest tempco.  I suspect that value has drifts over time.  There is probably a new setting that will give better performnace.  Tweaking the 825 ohm resistors in the chopper amp to optimize the zener tempcos might also be useful.

I wonder what it would take to replace the reference zener with something like an LM199.  The substate diode between the heater ground and zener ground might be an issue.

[texaspyro]

I did not do any extensive cleaning, simply blew out some dust and used some tuner cleaner on the switches, rotating them  back and forth several times after spraying.

The MV-106 manual warns you to re-lubricate the switches if you clean them or else they can wear out after a month of use.   They say to use DeOxit Shield to lube the switches.  I cleaned my switches (they were nasty) with the recommended Deoxit-5 and used a switch lube with polyphenol ether.

[Insatman]

@ Texaspyro....interesting  screen handle

I did lubricate the switches with a bit of silicone lube.  I didn't have any deOxit at the time...it's hard to find here in Philippines.  But I do have some now.  Probably should have a go at lubricating the set of switches.  I doubt they would wear out so fast for me as I seldom adjust the switches, but your observation is noted.

Insatman

[Insatman]

@texaspyro.

I've thought about replacing the main zener reference in the MV-106 with a LM399 or LTZ1000.  A LM199 if you can get one (from Ebay?) would also be a good choice.  I haven't looking into it much however because I wanted to see the performance of this mod over time first. 

Also I continue to build various fixed voltage references and evaluating their performance over time.  Awhile back I posted some data on this effort : <Metrology / Re: Best out-of-the-box 10V reference?> .

I felt it would be pointless to replace the existing reference in the MV-106 until I built a fixed reference with better performance and tested it over several months in my lab.  The effort is ongoing and has become a hobby (or addiction?) in of it'self.   

Insatman

[texaspyro]

@ Texaspyro....interesting  screen handle

I used to build very high temperature monitoring devices (inquiring minds want to know "just how hot is that volcano").  Also used to shoot fireworks shows (until government regulations and paperwork got out of hand).   

I do fly rather large rockets (up to "P" class motors).  I've held several world altitude records.  Someday I will post about my rocket launch controllers.

[Insatman]

3 month performance update after MV-106 repair/modification.    After gathering data for about three months, I am really pleased with the performance of my MV-106.   Refer to attached chart in PDF form.  After the repair/mod the unit drifted up about 11ppm over a period of 500 hours.  After that I monitored the unit less frequently.  Over the next ~1700 hours the unit varied by only +/-3ppm or so.  Total spread over the ~2300 hours was ~14ppm.   Given these tiny numbers I can't be sure the variance is the MV-106 or the HP34401A or likely some of both.  I strongly suspect that the initial rise during first 500 hours was the MV-106 as I didn't notice any significant change in the HP34401A measurements of other voltage references during that period.   A spot check of the HP34401A against my VoltageStandards.com VREF10-003 unit showed a change of +10uV over the three months period.   

Note that the accuracy ratings of the MV-106, HP34401A and VREF10-003 are roughly the same, all allowing for around +/- 300uV or more or about 30ppm.  So the total variance of ~14ppm is really outstanding.   As for absolute accuracy, the only recently calibrated unit was the VREF10-003.  Which when it first arrived agreed with the HP34401A within 10uV (1 count).   Now, 6 months later the HP34401A reads 10.00090 while the unit was calibrated at 10.00000.   Note that the VREF10-003 was calibrated on a HP3458 with NIST traceable cal by VoltageReferences.com.

I wish I had a nicer standard such as a 3458 or Kiethley 2002 to measure against but alas they are just too expensive.
To help muddy the waters further I have recently bought a second HP34401A, a Fluke 8505A and a HP3456A.   I got all three at bargain prices.  The new (to me) HP34401A is in spec but the 8505 will need some cal work.  Although the specs for the old fluke unit are pretty good really.   The HP3456A has yet to arrive.   XDev did some really nice work on one which is documented on his website.

Insatman

[Insatman]

Attached are new charts for the MV-106 performance with Temperature and Humidity data overlayed.

Insatman

[Insatman]

Further update on long-term performance of my MV-106 after repair and modification

[Insatman]

Further update on long-term performance of my MV-106 after repair and modification.

It appears the long-term drift may be humidity related.  Overall however the performance is quite good.