EEVblog Electronics Community Forum

Products => Test Equipment => Topic started by: quarks on August 04, 2013, 01:40:38 pm

Title: Geller Labs SVR and SVR-T
Post by: quarks on August 04, 2013, 01:40:38 pm
Hello all,

I have already bought and tested several SVR from Joe Geller and reported that they are great and accurate (<5 ppm agreement to my 6.5 to 8.5 digit gear).

Yesterday I received the SVR-T which is an improved version that uses a secondary temp compensation (developed by Lars Walenius).

For more details see:
http://www.gellerlabs.com/SVR%20Series.htm (http://www.gellerlabs.com/SVR%20Series.htm)

I meassured/compared it with my very best calibration gear and meters.

Here is a short setup info:
>4h DMM meter warmup
man. 10VDC range with 8.5 digit and averaging
using high quality PFTE low emf cable with crimped spade lugs and crimped pure copper alligator clips
checking and making sure DC offset is well below 1µV before meassurement started
using good/stable 15.00 V linear power supply 
ambient 25.1 to 25.4 °C (stable over the meassurement time)

  9.999 978 7 V (seconds after turn on)
  9.999 990 3 V (after 2 min.)
  9.999 997 7 V (after 5 min.)
10.000 000 6 V (after 10 min.)
10.000 001 3 V (after 15 min.)
10.000 001 4 V (after 30 min.)
10.000 001 3 V (after 60 min.)
10.000 001 3 V (after 90 min.)
10.000 001 3 V (after 120 min.)
10.000 001 2 V (after about 3 h)

Whatever my real absolut meassurement accuracy might be, after only 5 min. warmup time, this little device (compared to my gear) stays easily within <0.5ppm of the stated

"Calibration:    Jumper in:   10.000 00 V  at 24 °C ambient" in Joe Gellers "SVR-T Calibration Paper" that came with my SVR-T.

For me that really is a great improvement to the already very good SVR!

I would be interested to hear from other SVR useres what they found out and like to see some results.

bye
quarks   
Title: Re: Geller Labs SVR and SVR-T
Post by: Circuitous on August 04, 2013, 03:41:10 pm
I agree, the SVR-T is a neat little device.  I got one a year ago, and just fired it up again last week.  It showed 10.00000 on my Agilent 34410A. 
My two year old SVR (not the T model) was calibrated about 18 months ago, and it showed 10.00004
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 05, 2013, 07:25:55 pm
The Geller has been a great way to obtain a 10V check via mail.  However, I've experienced that its sensitive to ambient temperature and humidity, my unit is now over 1.5 years old and reads about 6uV high.  It had tended to swing 10-100uV positive with higher room temps or low humidity and read low with high humidity or cooler room temps.  I've been able to control temp exposure but not humidity, such as when it rains heavily and the atmosphere is saturated with water vapor, so when I do voltage checks I insure ambient is as close to similar to when Geller did the original calibration.

Could you do a simple test?  Raise the ambient temperature of the reference to higher  room temperature, say 30oC?  Can you retest the voltage? 


Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on August 05, 2013, 08:03:25 pm

Post edited to reflect the correct readings (i consequently left out a zero ... ie. 10.0004 should be 10.00004)

My Geller  a SVR (Calc'ed @ 23C) , are typically 10.00004 v @26 Deg , and was 10.00008 @28.5C.
I have never seen it drift above 10.00009 that was around 29.3C.
It doesn't seem to be affected noticably by "humidity" ... Meaning raining outside ...
Measured @ Keithley 2015  (Cal  expired May-2012)

My new SVR-T also Calc'ed @ 23C , is to new to talk behaviour about right now ...
But it also seems unaffected by rain outside.

Right now the SVR is 10.00008 @28.6 (Keithley) , and the SVR-T is also 10.00008  , but measured by my 3457 (Cal unk. , and it's a few ppm high)

Well .... I just swapped the SVR-T & the SVR

SVR-T is 10.00000 on Keithley @28.6C    :-+
SVR is 10.00015 @3457 , so i guess my 3457 is 7..8 ppm high  :-\

Well the 3457 is an old fella i got used this spring for $260 (incl 8 chan scanner) ....

If i ever find out how i'll cal the 10v range .... And i still trust the Keithley

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 05, 2013, 08:47:09 pm
Thanks bingo600.  From what you report, the Geller has drifted far off calibration. It also tracks positive largely with temperature.

Assuming your SVR-T is correct as it is new, then you report at 28.6oC the SVR-T is reading 10.000 00V on the Keithley.  Lets assume then the Keithley is still in spec.

Using Keithley, SVR reads 10.000 08, that is 800uV or 80 ppm off, at 28.5oC.  Its far different even if the temperature difference is 0.1oC.

Since you measured 10.000 04 or 400uV or 40 ppm off at 26oC, and is shows the Geller rose nearly double ppm in +2.5oC.

How much your Geller SVR has drifted off is above my expectation, my unit has not exceed 100uV positive at worst, and is now about 6uV high measured in the best ambient conditions.  But if the ambient conditions are not near exact as when Geller did the cal, it can be far worse than 6uV off.

If you drop your ambient to 24oC, I would expect the SVR to be only 20 ppm off.


My Geller  a SVR (Calc'ed @ 23C) , are typically 10.0004 v @26 Deg , and was 10.0008 @28.5C.
I have never seen it drift above 10.0009 that was around 29.3C.
It doesn't seem to be affected noticably by "humidity" ... Meaning raining outside ...
Measured @ Keithley 2015  (Cal  expired May-2012)

My new SVR-T also Calc'ed @ 23C , is to new to talk behaviour about right now ...
But it also seems unaffected by rain outside.

Right now the SVR is 10.0008 @28.6 (Keithley) , and the SVR-T is also 10.0008  , but measured by my 3457 (Cal unk. , and it's a few ppm high)

Well .... I just swapped the SVR-T & the SVR

SVR-T is 10.00000 on Keithley @28.6C    :-+
SVR is 10.00015 @3457 , so i guess my 3457 is 7..8 ppm high  :-\

Well the 3457 is an old fella i got used this spring for $260 (incl 8 chan scanner) ....

If i ever find out how i'll cal the 10v range .... And i still trust the Keithley

/Bingo

Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 06, 2013, 12:04:10 am
I got a new SVR-T from his last production batch a week ago. It reads quite steady 10.0001 (28 C) which is well within specs of my only instrument 34401A (last calib 2003). The value change a bit with 24 h temperature change (maybe 5-6 degrees and 10.00009-10.00015) and cables/noise from switching supplies in other instruments, but very little. I am very pleased anyway because I have got my DMM confirmed for 10 DC.

I also have a 5 V ref from a competitor. It has steadily drifted from 5.00022 to 5.00054 in 3 months, and now I know that it is out of spec 0.01%. I also have other 0.01% initial accuracy 10 V ref chips that have been steady for a long time. So I do not stress too much about it now that I know that the DMM is around 10 ppm.

I am making a slow multiplexing relay card w. microprocessor and LT2400 ADC. When this is working with logging and statistics to PC RS232 and double measurement with both 34401A and LT2400, I will send maybe 3-5 references 2.5 - 10 V for measurement elsewhere. I have a very cheap offer from a person with a calibrated 3458A. When I get them back and connected to the "DAQ" I will start to log drift.

But with only 6 1/2 digit DMM 15+4 ppm short time accuracy if calibrated I think the SVR-T is the closest I can come to a matching 10 V ref. I think I will buy one more (if this one turns out steady).

Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on August 06, 2013, 05:13:10 am
Thanks bingo600.  From what you report, the Geller has drifted far off calibration.  It also tracks positive largely with temperature.

Assuming your SVR-T is correct as it is new, then you report at 28.6oC the SVR-T is reading 10.000 00V on the Keithley.  Lets assume then the Keithley is still in spec.

Using Keithley, SVR reads 10.000 8, that is 800uV or 80 ppm off, at 28.5oC.  Its far different even if the temperature difference is 0.1oC.

Since you measured 10.000 4 or 400uV or 40 ppm off at 26oC, and is shows the Geller rose nearly double ppm in +2.5oC.

How much your Geller SVR has drifted off is above my expectation, my unit has not exceed 100uV positive at worst, and is now about 6uV high measured in the best ambient conditions.  But if the ambient conditions are not near exact as when Geller did the cal, it can be far worse than 6uV off.

If you drop your ambient to 24oC, I would expect the SVR to be only 20 ppm off.

@Saturation

My bad .....

I consequently left out a zero  (blush) .....  :palm:
The Keithley is a 2015-THD 6.5 digit meter (mode slow)  , and the 3457 (100 NPLC) is also running in 6.5 digit mode.
The 10.0004 should be 10.00004 , 10.0008 should be 10.00008 and so on ...

I'm sorry about that  :-[

I'll see if i can make my AVR-USB based DS18B20 temp ref finished , in a short time.
Then i can use GPIB to measure volts/temp , and present some serious data.

Btw: both of my SVR's came with a Ceramic chip , and i think they are less humidity sensitive than the "Plastic dips".

I'm actually quite happy with both my SVR's , and if you add my missing digit to your equations above.
They're actually performing quite well.   The SVR-T is quite impressive , even @+5C it's "spot on"


/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 06, 2013, 01:58:42 pm
Thanks bingo600, in that case my original results have to be reduce by 10.  If then, its only 4-8ppm the Geller is within spec including the rise with temperature.  I've corrected the last post.
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 06, 2013, 02:12:41 pm
Thanks quantumvolt.  Those drifts are fairly high for a new reference in a 24 hr period particularly with its built in temperature compensation, but its inline with Geller's estimate of 1ppm/C for the no temperature compensation SVR board, so the drift of about 60uV is close to the worse case.


I got a new SVR-T from his last production batch a week ago. It reads quite steady 10.0001 (28 C) which is well within specs of my only instrument 34401A (last calib 2003). The value change a bit with 24 h temperature change (maybe 5-6 degrees and 10.00009-10.00015) and cables/noise from switching supplies in other instruments, but very little. I am very pleased anyway because I have got my DMM confirmed for 10 DC.

I also have a 5 V ref from a competitor. It has steadily drifted from 5.00022 to 5.00054 in 3 months, and now I know that it is out of spec 0.01%. I also have other 0.01% initial accuracy 10 V ref chips that have been steady for a long time. So I do not stress too much about it now that I know that the DMM is around 10 ppm.

I am making a slow multiplexing relay card w. microprocessor and LT2400 ADC. When this is working with logging and statistics to PC RS232 and double measurement with both 34401A and LT2400, I will send maybe 3-5 references 2.5 - 10 V for measurement elsewhere. I have a very cheap offer from a person with a calibrated 3458A. When I get them back and connected to the "DAQ" I will start to log drift.

But with only 6 1/2 digit DMM 15+4 ppm short time accuracy if calibrated I think the SVR-T is the closest I can come to a matching 10 V ref. I think I will buy one more (if this one turns out steady).


Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 06, 2013, 06:24:16 pm
I think those 6ppm are 60uV . But to be fair, I guess my numbers are "worse" than worst case. Due to space limitations I have my references on a shelf almost 3 feet away from the bench and DMM, and since all is temporary now, they run on an old printer switched PS with an LM317. The cheap fleabay measuring cables have crocodile clips in 2 places and pass all kinds of lighting and PSUs in a "rack". So I cannot really report  reliable numbers before I get a setup rigged in a shielded box close to the DMM and with a decent cable and external PSU.

To be honest, I did not buy the board for the AD587 reference chip (I have LM399, AD588, LT6655, LT1019 and other good chips). I bought the calibrated 3458A measurement value in order to check the 34401A. And since the 34401A has at best 0.0015% + 4LSB (19ppm). the board is still within the DMM's spec even with this temporary sloppy setup. Adding Geller's customer service level as I perceive it, I am very pleased.
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on August 06, 2013, 06:37:00 pm
Could you do a simple test?  Raise the ambient temperature of the reference to higher  room temperature, say 30oC?  Can you retest the voltage?

I can do that probably tomorow and expect to see max. 1ppm/K, but would not be surprised if it is much better.
I will make sure that only the SVR-T will be at higher temp.
If others also want to do this test, please let us know some details about your setup. In my test I tried to have as little as possible influence by the power supply and the cable connection. When I repeat this test, I think/hope this setup is almost exactly identical and therefore only influenced by the ambient changes (temp and humidity).
Title: Re: Geller Labs SVR and SVR-T
Post by: JimmyMz on August 06, 2013, 06:59:44 pm
I think those 6ppm are 60uV . To be honest, the Agilent 34401A has, at best, 0.0015% + 4LSB (19ppm). The Geller board is still within the DMM's spec, even with this temporary sloppy setup.
Would you mind typing the equation you are using to calculate the 'ppm' values? I want to understand the mathematics behind the figures  :)

Thank You
JimmyMz
Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 06, 2013, 07:19:38 pm
I think those 6ppm are 60uV . To be honest, the Agilent 34401A has, at best, 0.0015% + 4LSB (19ppm). The Geller board is still within the DMM's spec, even with this temporary sloppy setup.
Would you mind typing the equation you are using to calculate the 'ppm' values? I want to understand the mathematics behind the figures  :)

Thank You
JimmyMz

Sorry, maybe I should have quoted.

1) The reference drifts from 10.00009 to 10.00015 V = 60uV for lets say 24 to 30 = 6 degrees C temp change. This is 10 uV pr degree. For 10V (10 000 000 uV, i.e. 10 million microvolts) magnitude it is 10 uV / 10 V = 1 part per million. So we have that 60 uV change in 10 V is 6 ppm for 6 degrees change of temp (1 ppm / degree C).

2) The 34401A has basic DC accuracy in this range 0.0015% of reading + 0.0004% of FS. Worst case 0.0019% is then a factor of 0.000019 of value. This is 19ppm.

---
EDIT: Please forget the 4LSB in the quote on top. I have written the Agilent 34401A spec wrong. It is  +- 0.0015 % of displayed value pluss 0.0004 % of range at 23 degrees C.
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on August 06, 2013, 08:12:50 pm
If others also want to do this test, please let us know some details about your setup.

I'm powering the SVR & SVR-T from an Agilent E3610 Linear PSU , to avoid switcher noise.
The Croks i'm using are Hirschmann , and the Banana cables are Hirschmann.

Both the Keithley 2015 and the HP 3457 have been powered up since mid april.
My SVR is around 1 year old , and the SVR-T is from April (Both powered up since mid-april).

I also have an Agilent 34401A (i seem to remember its around 4ppm low @10v)  , but i'm using it as my Lab meter.

Once i get my DS18B20 temp probes ready , i'll start to use the 8-port scanner in the 3457 , and then also connect my LM399 , and my two diy "Gellers"  ... Geller PCB's but Plastic DIP Ref's  & 1% metal resistors ....
I could use 1% SMD resistors if they have better tempco (hints welcome).


/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 06, 2013, 08:38:08 pm
Whoops sorry, my bad I corrected that missing zero  :palm:.  These zeroes seem to be dropping faster than I can write  ::).  Corrected my post.

Yes, overall I surmise the Geller is doing quite well based on a 6.5 digit DMM, its the same reason I got one but I should get a follow up cal from him soon.

I think those 6ppm are 60uV . But to be fair, I guess my numbers are "worse" than worst case..
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on August 07, 2013, 10:04:01 am
meassured again today with same setup as before:
>4h DMM meter warmup
man. 10VDC range with 8.5 digit and averaging
using high quality PFTE low emf cable with crimped spade lugs and crimped pure copper alligator clips
checking and making sure DC offset is well below 1µV before meassurement started
using good/stable 15.00 V linear power supply 
ambient 24.4 to 24.6 °C and 60% - 61% rel. humidity (stable over the meassurement time)

9.999 969 8 V (after turn on)
9.999 981 1 V (after 2 min.)
9.999 995 7 V (after 5 min.)
9.999 998 4 V (after 10 min.)
9.999 998 9 V (after 15 min.)
9.999 998 9 V (after 20 min.)

After that I put the SVR-T for about 15 min. to stable 27.9 - 28.1 °C also 61% rel. humidity
meassured DC offset at start was -0.3µV at start and -0.2µV when I stopped
9.999 979 7 V (after 15 min.)
9.999 981 1 V (after 20 min.)
9.999 982 3 V (after 30 min.)
9.999 983 1 V (after 40 min.)
9.999 983 5 V (after 50 min.)
9.999 983 8 V (after 60 min.)

This is around -15µV (= 1.5ppm of 10V) change with temp. delta of 3.5 °C.
That makes it overall a little better than 0.43 ppm/K.   

bye
quarks
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on August 07, 2013, 03:22:23 pm
Thanks quarks, that's goo data  :-+
Title: Re: Geller Labs SVR and SVR-T
Post by: branadic on August 07, 2013, 07:22:12 pm
Quote
1) The reference drifts from 10.00009 to 10.00015 V = 60uV for lets say 24 to 30 = 6 degrees C temp change. This is 10 uV pr degree. For 10V (10 000 000 uV, i.e. 10 million microvolts) magnitude it is 10 uV / 10 V = 1 part per million. So we have that 60 uV change in 10 V is 6 ppm for 6 degrees change of temp (1 ppm / degree C).

To be correct you have to use the start point what means 1ppm = 10,00009µV so in this example the drift is <6ppm and so temperature coefficient is <1ppm/K.
And one more detail, quarks you are from germany so you should have noticed that temperature differences are given in units of Kelvin, not degree Celsius. This is why only ppm/K is the correct physical unit for a drift.
It's sad that most manufactors do the same mistake, okay they are no physicist.
Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 07, 2013, 07:55:13 pm
Quote
1) The reference drifts from 10.00009 to 10.00015 V = 60uV for lets say 24 to 30 = 6 degrees C temp change. This is 10 uV pr degree. For 10V (10 000 000 uV, i.e. 10 million microvolts) magnitude it is 10 uV / 10 V = 1 part per million. So we have that 60 uV change in 10 V is 6 ppm for 6 degrees change of temp (1 ppm / degree C).

To be correct you have to use the start point what means 1ppm = 10,00009µV so in this example the drift is <6ppm and so temperature coefficient is <1ppm/K.
And one more detail, quarks you are from germany so you should have noticed that temperature differences are given in units of Kelvin, not degree Celsius. This is why only ppm/K is the correct physical unit for a drift.
It's sad that most manufactors do the same mistake, okay they are no physicist.


The quoted text and the errors/blame are all mine. As Pedantischer Besserwisser I politely advise you not to remove the quote's author ???
Title: Re: Geller Labs SVR and SVR-T
Post by: branadic on August 08, 2013, 08:27:24 pm
If people deal with physical units I demand that this is done in the correct way and yes, in this point I'm captious. It's the same like the difference between VA (volt ampere) and W (watt). Sorry about that.
I follow the device: "You want to learn something? You should learn it in the correct manner."
Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 09, 2013, 03:55:43 am
I have done wrong towards the SVR-T by my sloppy setup. It is much better than I reported earlier.

Yesterday I used some very short leads (about 20 cm), placed the board just in front of the 34401A, used a linear PS (wall wart unregulated 18-19 V with LP filter / cap bank around 2 meters away from the ref board and a LM317 to within 1% of 15 V placed only cm's from the ref board), placed the ref board with PS and the sensor from a cheap fleabay thermometer in an aluminum cake box, "sealed" the box with styrofoam, turned off all other electric appliances and lighting, waited a bit,  pushed the STAT button of the DMM with a 1.5 m long broomstick and left the room.

When I came down this morning the DMM read MAX 10.000 014, MIN 9.999 996 and MEAN 10.000 007 (STAT-mode gives one digit more than the normal display). The thermometer had MAX 29.6 and MIN 25.2. So probably less than 20 uV change for maybe 4 degrees change in temperature (to be somewhat conservative). And the meter seems to be just a few uV off the nominal 10 V (first and last calibration 1983 stored in meter  - calibration is before sales date and must be the initial calibration from Agilent).

I hardly believe it  :-* :-DD
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on August 09, 2013, 07:55:03 am
I hardly believe it  :-* :-DD

Your result is almost the same as mine, which is a very good confirmation.
So I think you can be happy and believe it.
Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 09, 2013, 08:34:57 am
Sure. It is just that it is the first time I get anything confirmed to around 0.001% (my other paid-for 3458A-measured reference from a competitor was off 0.004% off at delivery and has now drifted out of spec 0.01% in 3 months). I have many stable references with datasheet initial accuracy 0.01%, and I can now trust those that measure off 0.005% and less. I am also pleased that I do not need any calibration (I only need to trust DC 1 - 10V for references and precision voltage dividers, for everything else I am OK with a good handheld meter).
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on August 09, 2013, 10:58:46 am
I am very happy with my Geller SVRs!
There is only one little thing, that could make it even better for me, that is a statement of the uncertainty of the calibration.
But because I have my own "cal lab" and can compare to very accurate gear, I guess the absolute accuracy of my SVR-T is pretty close to 1ppm at the stated cal temp.
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on August 16, 2013, 02:09:07 am
Just for fun, I ran my Geller SVR-T against a Fluke 731B for the last day or so using the DC volts ratio feature of a 34461A.  Screen capture is attached.  Ambient temperature spread was 23.7 to 25.6 deg C.  Unfortunately, there was some kind of glitch early on (probably an AC power glitch that got through the power supply) so the minimum value isn't valid.

They appear to be tracking within a couple of ppm.

I too am happy with the SVR-T.  I suppose the 731B has the advantage of the 1V output, but it's not in cal to my knowledge so it is of limited use.

Orin.
Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on August 16, 2013, 04:50:58 am
Just for fun, I ran my Geller SVR-T against a Fluke 731B for the last day or so using the DC volts ratio feature of a 34461A.  Screen capture is attached.  Ambient temperature spread was 23.7 to 25.6 deg C.  Unfortunately, there was some kind of glitch early on (probably an AC power glitch that got through the power supply) so the minimum value isn't valid.

They appear to be tracking within a couple of ppm.

I too am happy with the SVR-T.  I suppose the 731B has the advantage of the 1V output, but it's not in cal to my knowledge so it is of limited use.

Orin.

Just a metrological hint:
measuring both references back to back, i.e. their difference, would provide much faster (2x) readings and higher resolution, directly detecting changes in the sub ppm range.

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on August 16, 2013, 05:36:11 am
Just for fun, I ran my Geller SVR-T against a Fluke 731B for the last day or so using the DC volts ratio feature of a 34461A.  Screen capture is attached.  Ambient temperature spread was 23.7 to 25.6 deg C.  Unfortunately, there was some kind of glitch early on (probably an AC power glitch that got through the power supply) so the minimum value isn't valid.

They appear to be tracking within a couple of ppm.

I too am happy with the SVR-T.  I suppose the 731B has the advantage of the 1V output, but it's not in cal to my knowledge so it is of limited use.

Orin.

Just a metrological hint:
measuring both references back to back, i.e. their difference, would provide much faster (2x) readings and higher resolution, directly detecting changes in the sub ppm range.

Frank


Good point.  What was I thinking?  I remember; the tempco of the 34461A is pretty much eliminated in this configuration.  Of course, a tempco of one or two ppm isn't going to matter much in a differential measurement of a few 10s of uV and I need to worry about thermal EMFs instead.

I had considered using my 845AR's recorder output while using the 10uV scale, but decided the 845AR probably drifted too much.

I'll go and reconfigure the connections for the differential measurement and give it another day or so.  Thanks for the hint!

Orin.


Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on August 16, 2013, 03:02:49 pm
Just a metrological hint:
measuring both references back to back, i.e. their difference, would provide much faster (2x) readings and higher resolution, directly detecting changes in the sub ppm range.

Frank

Would someone pleas elaborate this "back to back" ??

How would one conect that for the above measurement ?

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: alm on August 16, 2013, 03:27:23 pm
Connect the negative terminals of the references together. Connect the DMM between the two positive terminals. It now measures the difference between the two references with a higher accuracy and resolution because it can be on a lower range.
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on August 18, 2013, 06:53:17 am
So I set everything up for differential measurement, then yesterday morning, everything went crazy when I nudged something on the bench.  Results all over the place.  I removed the possible culprits - a monitor and keyboard (pushed into a speaker and on auto-repeat) and now everything is back to 'normal'.

The first attachment ...6305C is with an HP 6205C power supply (typo in the filename).  After a while, things got kind of noisy.  I suspected the power supply*, so I breadboarded an LM317L** set to 15.00V and fed it 18.5V from a different supply.  Much better; see second attachment.

In both cases, the SVR-T and 731B track each other within a couple of ppm (+/- 0.35ppm meter uncertainty; 100mV range 90 day spec for the 34461A).

This is not a best case scenario for the SVR-T.  It's in the open on the bench, so it's possible the temperature sensing thermistor that is under the board sees a different temperature than the chip that's on the top of the board.  I'm going to put the SVR-T in an enclosure and see what effect, if any, that has.

I still have to work out what causes the glitches.  With the 100 plc integration, I'm sure they are worse than they seem.

Got to love the 34461A for this.  I'm using the web interface to monitor it from the other end of the house.  It will download a screenshot from the meter on demand, or show the current reading/statistics in real time.  No GPIB necessary, just using the ethernet interface (admittedly bridged over wireless by a laptop in the 'lab', but that just makes it more amazing that it all works).

Orin.

*The 6205C barely worked when I got it and needed much contact cleaner in the slide switches.  I suspect I didn't clean them well enough.

** See the TI datasheet.  10uF tantalum across the "R2" resistor and 1uF tantalum on the output.  "R1" is 243 ohm, R2 is 2610 in series with a 200 trimpot.
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on August 27, 2013, 06:04:49 am
Well, the SVR-T is now in a nice metal box and here is a couple of days of comparison with the Fluke 731B, differential measurement, Fluke '+' to 34461A Lo and SVR-T '+' to 34461A Hi. Fluke '-' to SVR-T '-'.  Temperature spread was 20.4 to 24.2 C.

They are within 1ppm of each other over 48 hours and 3.8 deg C.

So, the SVR-T at $75 is a great deal IMO.  Add a box and some clean power and it's still less than the bargain basement, uncalibrated, batteryless 731B and I shudder to think what Fluke would want to calibrate the 731B!  (Fortunately, it was only $30 for a new NiCd pack for the 731B at the local BatteriesPlus.  You don't want to run a 731B without a battery as the reference section then gets powered by unfiltered DC from the battery charge circuit.)

Orin.

Title: Re: Geller Labs SVR and SVR-T
Post by: Electro Fan on August 27, 2013, 06:38:07 am
How critical is the quality of the power supply to the results produced by the Geller devices?

(What are the minimum attributes/specs for a power supply that Geller users would recommend in order to make it worthwhile to use a Geller?)

Thx, EF
Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on August 27, 2013, 09:10:28 am
Orin,

wow, I did not expect the Geller reference, nor the 731B to be that stable over time and temperature.

Both seem to have a T.C. below 0.1ppm/K, and 2 days stability below 0.5ppm, as your nice graphic shows.

Respect!  :clap:

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: quantumvolt on August 27, 2013, 10:43:59 am
How critical is the quality of the power supply to the results produced by the Geller devices?

(What are the minimum attributes/specs for a power supply that Geller users would recommend in order to make it worthwhile to use a Geller?)

Thx, EF

Short term transfer accuracy applies in the short term at the reported temperature and at the specified power supply voltage (15.0V for 10V SVR references) after of warm up period of 30 to 60 minutes, <10 ppm within 5 minutes.

The SVR requires an external (not supplied) single 15 VDC power supply, preferably using a linear regulator. With no load on the SVR board, the board draws on the order of 2 mA. A 15 VDC linear power supply rated at 80 milliamps or more is recommended. For relatively low accuracy applications (.01%), two 9V batteries in series can power the board at about 18 V.

...

The reference chips we use are rated at 100 uV / V (power supply), however most perform much better.

...

The details of high precision use, 30 to 60 minute warm-up, use as close the temperature where we did the calibration, power supply voltage of 15V +/- 0.1V are all just to squeak the last ppms possible (to get as close as possible to our 732B voltage) using this relatively inexpensive method. At relatively low precision (e.g. 0.01%), these are all non-issues.


Quotes from / more at http://www.gellerlabs.com/Voltage%20References.htm (http://www.gellerlabs.com/Voltage%20References.htm)
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on August 28, 2013, 03:58:21 am
Orin,

wow, I did not expect the Geller reference, nor the 731B to be that stable over time and temperature.

Both seem to have a T.C. below 0.1ppm/K, and 2 days stability below 0.5ppm, as your nice graphic shows.

Respect!  :clap:

Frank

Thanks!  It's Agilent's nice graphic really - the 34461A is connected via Ethernet and I access it remotely with a web browser.  It will show the meter's display either numerically or graphically.  I just take a screenshot and cut the interesting part out.

For the record, I attached pictures of the SVR-T & its new home.

Orin.

[Edited: stupid typo]
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on September 21, 2013, 09:36:03 am
Just got my 3457A back from cal.

Room temperature is 22.8 deg Celcius , SVR's calibrated @23.0C.

Freshly (7days ago) calibrated HP 3457A , connected to Geller SVR-T  (Turn on ~2.5 hr's)
May-2011 calibrated keithley 2015-THD , connected to Geller SVR  (Turn on ~4 month)

The SVR's are connected to a HP E3610 , but not in a box (yet)

Edit: I really like my SVR's  :-+

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on September 21, 2013, 01:05:28 pm
 :-+ perfect results!
Title: Re: Geller Labs SVR and SVR-T
Post by: saturation on September 21, 2013, 02:21:15 pm
Thanks bingo6000, that is good news.
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on September 21, 2013, 02:35:36 pm
@quarks

I got a very nice offer from a Cal-Lab in Germany.

They offered a ISO9000 Cal. with Justage , for 99€ + VAT and return shipping to DK (UPS).
Total 153.00€

Contact info:
Quote
Beatrice Jauch

Esenwein GmbH
Porschestraße 17
D-73269 Hochdorf
Tel. 07153 . 30 80 951
Fax. 07153 . 30 80 977
eMail: jauch at <same as webpage>
web: kalibrierlabor.de

Thats not a bad price.

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on September 21, 2013, 07:53:39 pm
Just in case someone missed this info
http://www.gellerlabs.com/DMMtempco.htm (http://www.gellerlabs.com/DMMtempco.htm)
Title: Re: Geller Labs SVR and SVR-T
Post by: Galaxyrise on December 26, 2013, 05:13:01 pm
Too good a deal to last, it seems:

Quote from: http://www.gellerlabs.com/
However, the sad truth is, our present business plan is not commercially viable. It is not bankruptcy, just an ideal time to make a good business decision.  The present plan is to wind the current business down around the end of this year.

Sad I missed my opportunity!  I had a bunch of local experiments I wanted to do before I was "ready" for something calibrated.
Title: Re: Geller Labs SVR and SVR-T
Post by: bronson on December 26, 2013, 05:49:02 pm
Dangit, that's too bad.  I gotta say, the SVR-T seemed a great deal for what you get.  I wonder if Joe could have charged twice as much for them, and if that would have made his home lab more viable...?

Here's a random pic of my SVR-T from a few months ago.  Random Pomona banana cords, random old power supply from the 80s.  Ignore the Fluke -- I already knew it needs a recal.

Drat, I just noticed the 3457 is only displaying to 5 1/2 digits.  How did I do that?  Guess I'm going to have to redo the test.  And that requires waiting for the shop to warm up...
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on December 26, 2013, 10:58:39 pm
Note that Joe told me it wasn't a good idea to run several meters from the same SVR.

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: Vgkid on December 27, 2013, 07:28:26 am
Bingo, how is the E-100?
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on December 27, 2013, 09:38:47 am
Bingo, how is the E-100?
I haven't had much time to play with it yet ....
But it seems to be functioning , and not "far off"

I have gotten the Deox-IT for the wheels , but  still need to open it up and write down all the ELCO' values/voltages for a (recapping).

Hope to do this in Q1 of 2014.

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: bronson on December 28, 2013, 02:28:15 am
Yea, for absolute precision I agree with Joe: hooking multiple meters in parallel seems to destabilize things a bit.  In this case, at 5 1/2 digits, I don't think it matters.

I'd like to know why though...  Seems like hundreds of gigaohm input impedance in parallel with teraohm input impedance should be fairly undetectable.  Oscillation between the charge reinjection circuits?  Constructive interference of the sampling noise?

Did Joe say anything more?
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on December 28, 2013, 11:02:06 am
@Bronson


How is your 3457A doing with temperature deviations ?
My 3457A seems to be extremely temp sensitive.

I have a SVR-T on the 3457 , and a SVR on the Keithley 2015

When the room temp changes (2 degrees) the 3457A shows a 4..5 ppm change , even with a SVR-T connected.

If i then swap the SVR & SVR-T , the Keithley is still "spot on"

This makes me think that the environment temp has a big influence on the 3457A.

I am wondering if i should open the 3457A , and "isolate" the V-Ref board with some foam padding.


/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: bronson on January 10, 2014, 08:26:37 pm
Hi bingo600, I've been meaning to run some tests but haven't gotten around to it.  I'll reply when I do.

You saw this of course?  http://www.gellerlabs.com/DMMtempco.htm (http://www.gellerlabs.com/DMMtempco.htm)

My problem is that I have my single SVR-T and the 3457...  I'm sure I'll see tempco swings but I won't know who to blame. :-//  Given the age and pedigree of each device I'm inclined to think most of the swing would be the 3457 but dunno.

Also I need to set something up to log ambient temp...  Just stick a K thermocouple into the Fluke and GPIB it?  Or is there an easier way that I'm overlooking?

Seems like this would be an easy project since my EE shop is an unheated detached single car garage and the temp swings have been huge lately...
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on January 10, 2014, 09:30:41 pm
@Bronson

I'll take back the "Bad Mouthing" about the 3457A .....
I have goofed , and didn't put the Keithley out of "AutoRange" , so it shifted to the 100v range , when the SVR hit 10.xx volt and stayed there. And i didn't even notice that a digit was missing  :palm:

After i have put the Keithley on manual 10v range , they are both affected my temperature.

And in the 2 days since i have discovered this error , it seems that they are quite equal in the temperature drift.

(Sorry HP)   :'(

But i just finished my GPIB logging setup , and have written a loggerprogram in python
https://www.eevblog.com/forum/reviews/linux-gpib-and-python-keithley2015-hp3457a-10v-logging/ (https://www.eevblog.com/forum/reviews/linux-gpib-and-python-keithley2015-hp3457a-10v-logging/)

Now i can really start to collect data.

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on March 31, 2015, 04:00:59 pm
Hello,

I have had got the possibility to make some measurements on a Geller SVR-T, which HighVoltage lent me recently.

I'd like to share stability  measurements, I've done in comparison with my 3458A, NPLC100 (2sec aperture).
The room temperature during this 29h run was exceptionally stable, <= 0.2°C change.

It turns out, that the Geller SVR-T competes even with the LTZ1000 references in terms of noise ~ 400nV, standard deviations for 10min. transfer stability sigma = 0.06ppm , 2h sigma = 0.07ppm and 24h sigma = 0.1ppm.
Also see stability figures, noise ~ 200nV, and Allan deviation from a LTZ1000 in comparison.
(The room temperature was less stable, about 0.7°C)

These Allan diagrams were normalized to ppm.

I also did a rough, estimate measurement on the T.C. of the Geller SVR-T, see my usual beer box setup.
I heated the whole box to about 31°C, from originally 20.4°C, measured by a thermo couple and a pyrometer, on the surface of the reference aluminium box, after temperature stabilization.

The voltage deviation vs.  time graph reveals several effects concerning temperature change and lagging, but clearly demonstrates  the very low T.C. of around -0.1ppm/K over a 10°C range .. that also competes very nicely with the LTZ reference.


Concerning the absolute long term stability:

I got a poor-mans-calibration from quarks at about September 2013.
His gear (actually his 8508A) had a brand new calibration, and my 3458A was 0.3ppm low compared to his instrument.

Since 5 years, I monitor the drift of my 3458A against two external LTZ1000, and one 5442A, and the overall drift between all these four references is <= +/- 1.5ppm over the whole time period.

Therefore, this recent measurement of HighVoltages SVR-T should be within 2ppm absolute uncertainty, as estimated.

This also is feasible, as it turns out, that the SVR-T is measured to 10.0000223V (+/-100nV) , only 2ppm off from the intended calibration value from Joe Geller, 2 years ago.
I measured no bigger differences than 1µV to that value, but I omitted to measure the warm-up behaviour.

HighVoltage promised to present comparative measurements with his new 34470A, which fall really very well within assumed limits. (Can't be accidentally)

As a resumée, the Geller SVR-T is really a very stable reference, very precise and also long term stable.. A pity that Joe Geller went out of business.

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on March 31, 2015, 07:20:47 pm
A pity that Joe Geller went out of business.

Hello Frank,

thats right. But on the other side:
Joe used the AD587LQ (comercial grade) references for his SVR-T.
They are obsolete now (perhaps thanks ROHS).

The only alternative would be the AD587UQ (military grade).
But I have heard that this one behaves much worse near room temperature than the LQ reference.
(they are trimmed to the temperature extremes).

By the way: in your diagram you write "AD586" = 5V reference instead of "AD587" = 10V reference.
In my ADCs  I use the AD586LQ as 5V references. (Which are still available).

With temperature compensation and after a run in phase of 6-12 months some of them are very stable
(within some ppm / year).

With best regards

Andreas



Title: Re: Geller Labs SVR and SVR-T
Post by: quarks on April 01, 2015, 09:34:48 am
As a resumée, the Geller SVR-T is really a very stable reference, very precise and also long term stable.. A pity that Joe Geller went out of business.

Hello Frank,
that is excellent information, thanks a lot for sharing :-+

And big thanks also to Joe Geller for his great work.

bye
quarks
Title: Re: Geller Labs SVR and SVR-T
Post by: JohnnyBerg on April 01, 2015, 09:53:37 am
I've got some voltage refs ready in a week or two. I could spare a few for testing.
Would be nice to see how they compare to the Geller SVR?

So if anyone is interested to do some testing, send me a PM.
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 01, 2015, 04:57:24 pm
A great thank you to Dr.Frank for testing my Geller Labs SVRT in such detail and creating a baseline for comparison.

I had the SVRT running in my lab for several days now.
And today I took the first data with my new Keysight 34470A and a few other meters in comparison.

Soon I will hook up the Pro Version of BenchVue and take some long term data.
But I think I would need to place the SVRT in a protected box to remove temperature changes.
Right now the temperature in my lab is between 22 and 24 degrees, depending on the time of the day.

And a big thank you also to Joe Geller to produce such a great reference.




Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on April 01, 2015, 06:25:09 pm
[b...and creating a baseline for comparison.[/b]

I had the SVRT running in my lab for several days now.
And today I took the first data with my new Keysight 34470A and a few other meters in comparison.

Soon I will hook up the Pro Version of BenchVue and take some long term data.
But I think I would need to place the SVRT in a protected box to remove temperature changes.
Right now the temperature in my lab is between 22 and 24 degrees, depending on the time of the day.

And a big thank you also to Joe Geller to produce such a great reference.

HighVoltage,

yep, the baseline for the SVR-T is much more stable than what you see,
so the sad truth about the 34470A is, that it is less temperature stable thane the Geller Ref, and despite its inner LTZ1000A reference.

They simply used less sophisticated components for the A/D voltage reference amplifier, obviously, as specified with 1ppm/K ..

I think, your 34470A needs a box to protect its inner temperature..

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 01, 2015, 06:49:00 pm
Hello Frank

Sadly I think you are right and the Geller Labs SVRT seems more stable than the brand new Keysight 34470A.
Sometimes I see a flat line in the display of the 34470A for like 30 min or so and then suddenly for no apparent reason, I get a drift up or down.

Once I have BenchVue all setup and ready to go, I am planning to hook up 4 DMM's at the same time, since BenchVue supports 4 instruments at one time.

This is my plan:

1 x 34401A
1 x 34410A
1 x 34461A
1 x 34470A

I will have my second 34461A monitor the case temperature of the Geller Labs SVRT with a stable 5k Thermistor.

Do you think it is wise to have all 4 multimeters in the same box and the SVRT in a different box for temperature stabilization?
Or do you have any suggestions.

This way I would get a good and reliable correlation between the instruments and the SVRT.





Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on April 01, 2015, 07:57:23 pm
Dear HighVoltage,

for sure it would be possible to determine which of the  DUTs is drifting, by thermally separating source and DMM...

Whether these lower grade 6 1/2  instruments really give some useful information, I doubt that.

Is it possible on the 34470A, to read its calibration temperature?
(The HP3458A got this feature  ;D)

Anyhow, both of our instruments, 3458A and 34470A are apart a few ppm only.. again demonstrating the stability of the Geller ref, as a transfer standard.

Btw.: Don't you own a nice, dry and temperature stable basement, as in each good German house?

The results would be much more stable.

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 07:31:57 am
Btw.: Don't you own a nice, dry and temperature stable basement, as in each good German house?
The results would be much more stable.
Frank
Dear Dr.Frank
No basements where I live, because we are close to a small river and basements would overflow when the storm comes.

But what I noticed is that a close by WiFi and Cell Phone activity will have a massive influence on the data. Much more than I would have expected, even is the cell phone is 2 to 3 m away.

With this information in mind and all wireless activities turned off as far as possible, I did another test on the SVR-T that you had measured.

Here are the results from 3 Agilent 34461A's and the new Keysight 34470A.
All 4 pictures are from 4 different instruments, hooked up in parallel to the SVR-T
I would suspect the slight difference to be just a calibration issue between the instruments

Screen_1 = 34461A
Screen_2 = 34470A
Screen_3 = 34461A
Screen_4 = 34461A

I think what this proofs one more time the ultra stable SVR-T

Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 07:38:54 am
And here are the same data as trend charts

Screen_A1 = 34461A
Screen_A2 = 34470A
Screen_A3 = 34461A
Screen_A4 = 34461A

Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on April 19, 2015, 10:47:37 am
But what I noticed is that a close by WiFi and Cell Phone activity will have a massive influence on the data. Much more than I would have expected, even is the cell phone is 2 to 3 m away.

Hello,

did you already try a blocking capacitor (10-100nF) at the output of the SVR-Reference. (near pin 4+6 of the AD587).
I always try to place one (as long as the reference does not oscillate) to reduce RFI effects.

With best regards

Andreas


Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on April 19, 2015, 11:08:44 am
Hi HighVoltage,

All your 4 instruments confirm the very good stability of the SVR-T, essentially the same magnitude as with my 3458A.

All these references, especially the LTZ, are very sensitive to AF noise, especially from switch mode PSUs.

So in my basement, where my analogue lab is situated, there are conventional lamps (50Hz) only, no WLAN, no cellular phones and no DECT station.

The steel wires in the concrete roof also filters most of radiation from outside.
(FM radio can be received, through a small window, anyhow.)

So, a (good German) basement gives a near perfect shielded lab.

A pity, that in many areas of northern Germany , you don't have basements in the houses.
(My brother in law doesn't have one, either, he's living in the Oldenburger Münsterland, where they have moorland. So he use to bury many of his tools in his 1/2 hectare garden..   |O)

PS: When will you have the opportunity to compare the SVR-T etc to a DKD lab?

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: bingo600 on April 19, 2015, 01:35:07 pm
But what I noticed is that a close by WiFi and Cell Phone activity will have a massive influence on the data. Much more than I would have expected, even is the cell phone is 2 to 3 m away.

Hello,

did you already try a blocking capacitor (10-100nF) at the output of the SVR-Reference. (near pin 4+6 of the AD587).
I always try to place one (as long as the reference does not oscillate) to reduce RFI effects.

With best regards

Andreas

Joe Geller warned me about Cellphones nearby the SVR's , he had the same discovery.

/Bingo
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 03:22:16 pm

Joe Geller warned me about Cellphones nearby the SVR's , he had the same discovery.

/Bingo

Very interesting, especially since I have a the SVR-T in a fully enclosed metal housing
I will make a picture later of  trend chart with a cell phone next to it
It is unreal how bad it is.
 
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 03:28:47 pm

PS: When will you have the opportunity to compare the SVR-T etc to a DKD lab?

Frank

Hello Frank,
The SVR-T and the LTZ1000A will go to the DKD lab tomorrow morning
They are already running for a few days on 24V battery power and will be transported to the DKD lab in an ON state.
They will stay in the DKD lab at 23 degree C for another day, before they will start measurements with a HP 3458A and a Fluke 8508A
Then I will have two confirmed references and will send you the LTZ1000A

Will keep you informed of the progress.
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 03:35:34 pm
But what I noticed is that a close by WiFi and Cell Phone activity will have a massive influence on the data. Much more than I would have expected, even is the cell phone is 2 to 3 m away.

Hello,

did you already try a blocking capacitor (10-100nF) at the output of the SVR-Reference. (near pin 4+6 of the AD587).
I always try to place one (as long as the reference does not oscillate) to reduce RFI effects.

With best regards

Andreas

Hello Andreas,

The SVR-T has not been touched and nothing has changed, since it left Joe Geller.
Except, I placed it in an extruded Aluminum housing and added gold plated MC connectors.

Do you think, the blocking capacitance will help, with the cell phone RF sensitivity, although the SRV-T is already in a good metal housing.
May be the RF energy is coupled in to the wires from the reference to the multimeters?
I will post a screen shot later with an iPhone 1 m away from the reference
It is so bad, that at first I thought something is wrong with the reference


Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on April 19, 2015, 03:55:16 pm
Remembering teardown of a 732B, they placed ferrite rings around the output jacks.
Maybe that helps even better..
Capacitor on the output may cause oscillation.

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on April 19, 2015, 04:20:58 pm
That is a great idea for testing.
You think each of the wires should have their own ferrite ring or should both go through the ring as a twisted pair?

I will do some test's after the reference is coming back from the DKD lab.
Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on April 19, 2015, 06:41:52 pm
Do you think, the blocking capacitance will help, with the cell phone RF sensitivity, although the SRV-T is already in a good metal housing.
May be the RF energy is coupled in to the wires from the reference to the multimeters?

Hello,

That is exactly the way where the RF-field comes into the housing.
Depending on parasytic capacitances the metal housing may also increase the effect.
The best method finding the sensitive lines is to put your hand on the input/output lines or the batteries or the housing.
Another would be setting a non isolated metal tip on every pin of the reference under field effect or using a switchmode power supply.

With my precision measurements I now have nearly at every IC-Pin going to a external line some type of filtering.

Capacitor on the output may cause oscillation.

References usually are not sensitive to capacitive loading (except very low power devices like ISL21009).

You think each of the wires should have their own ferrite ring or should both go through the ring as a twisted pair?

Both will help against external RF fields in the same way. The twisted pair method is only necessary if there are large frequencies on signal lines which should not be dampened in the same way as the RF interference.
Essential is the length of the wire within ferrite material. (high RF impedance against Z0 = 377 Ohms).
So best would be if you can do both: a ferrite + the blocking capacitor. A additional feed through capacitor on each pin through the housing (for power supply and the output) would give further improvement.

With best regards

Andreas

Title: Re: Geller Labs SVR and SVR-T
Post by: Dr. Frank on April 19, 2015, 08:45:09 pm
That is a great idea for testing.
You think each of the wires should have their own ferrite ring or should both go through the ring as a twisted pair?

I will do some test's after the reference is coming back from the DKD lab.

Fluke calibrators have 3..4 jacks, and each one had its own fat ferrite ring.. no twisted pair (how should  that work, anyhow?)

The photo was somewhere on bbs38hot, I suppose

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: lars on April 20, 2015, 08:20:31 pm
On my eight SVR boards, that I have put into boxes, I have added ferrite beads both to the banana jacks and PS connector. I can't for sure say that it helps at all. If I have a call on the cell phone within 1meter I see a disturbance on the reading. I'm not sure if the SVR board or the measuring device is the problem.

Probably I should have put some capacitors to chassi on the box also.

Setting up a test is not easy. For what frequencies? What field strength? At work we have a EMC room but even if I could set it up where I need a way to measure it without error due to the measuring device being disturbed.

/Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: lars on April 20, 2015, 09:17:10 pm
More than a year ago I bought a AD587UQ and AD587JQ from Digikey. Now I have finally made a temperature test on them. As said in an earlier post the AD587UQ seems to have a high temperature coefficient at room temperature. I got +11ppm/C at 23°C. The JQ was +2ppm/C. Both of them had the familiar "U"-shaped curve up side down. The second order tempco was -0.03ppm/(C*C). This is similar to what I got on the SVR boards from Geller labs that I added a NTC+resistor to compensate for both first and second order temperature coefficients. The AD587LQ seems to have had around +2ppm/C very often and that was perfect to use one non-linear NTC resistance to compensate for both first and second order. I did quite rough compensations but I'm quite sure Joe made them even better. On my last four SVR's Joe had already  put the NTC's (SVR-T's  ;) ).

The +2ppm/C I got for the JQ is perfect to compensate with an NTC. But the sad thing about the about 15-20 JQ's I have checked is that about one fourth have had several ppm p-p noise and about three quarters have unusable temperature coefficients.

I don´t have a Fluke 732 or similar stable voltage reference but I have four temperature compensated REF102CM that have been powered since 2000. With external checks 2003 to 2013 they have drifted almost nothing (with uncertainty +-10ppm) over 10 years. I have been fortunate to check against several DMM and calibrators during the last 15 years. The last good check was against a Fluke 8508 . My SVR board was +0.2ppm. Of course this was pure luck. The expanded uncertainty of the 8508 was 2.7ppm. I myself think I have an expanded uncertanty of 5ppm for my 10V references.

The eight SVR has stayed within 2ppm from the REF102CM's the last years. The regression lines is within 0.2ppm/year.

What I have seen on almost all SVR boards and also the REF102CM is about 0.8ppm jumps between two levels. On a HP34970A I have followed for almost 20 years I still see jumps of 1.5ppm that I conclude comes from the LM399. By the way it still drifts about 1ppm/year (internal LM399 almost always powered on).

Geller labs specified 5ppm as transfer accuracy for the SVR if I remember correct. My first SVR board was +4ppm from what I thought it should be. I was happy to take it on the plane to a sub-supplier I visited the week after I received the SVR board. They have a Fluke 5700 that is calibrated every year by Fluke UK. Comparison with that showed about 3-4ppm high for the SVR board. Half a year later I sent it back to Geller labs and it was +4ppm on the protocol. All the other seven has been within 1ppm from my "10V" (except one that I destroyed at power up due to reversed wiring and I put in a AD587KQ and selected a new NTC+resistor).

/Lars

Title: Re: Geller Labs SVR and SVR-T
Post by: kubatyszko on August 16, 2015, 11:14:13 pm
Orin,

wow, I did not expect the Geller reference, nor the 731B to be that stable over time and temperature.

Both seem to have a T.C. below 0.1ppm/K, and 2 days stability below 0.5ppm, as your nice graphic shows.

Respect!  :clap:

Frank

Thanks!  It's Agilent's nice graphic really - the 34461A is connected via Ethernet and I access it remotely with a web browser.  It will show the meter's display either numerically or graphically.  I just take a screenshot and cut the interesting part out.

For the record, I attached pictures of the SVR-T & its new home.

Orin.

[Edited: stupid typo]

Hi Orin, I'm looking at building a couple of references similar to SVR-T , but when I look at your picture, it looks exactly like SVR with a perfboard on the side, can you perhaps elaborate what is its purpose ?
From the SVR-T schematic it seems that the THR trimming should literally be a thermistor and a resistor but your board has a few more components.

Thanks
Title: Re: Geller Labs SVR and SVR-T
Post by: orin on November 28, 2015, 06:12:31 am
Orin,

wow, I did not expect the Geller reference, nor the 731B to be that stable over time and temperature.

Both seem to have a T.C. below 0.1ppm/K, and 2 days stability below 0.5ppm, as your nice graphic shows.

Respect!  :clap:

Frank

Thanks!  It's Agilent's nice graphic really - the 34461A is connected via Ethernet and I access it remotely with a web browser.  It will show the meter's display either numerically or graphically.  I just take a screenshot and cut the interesting part out.

For the record, I attached pictures of the SVR-T & its new home.

Orin.

[Edited: stupid typo]

Hi Orin, I'm looking at building a couple of references similar to SVR-T , but when I look at your picture, it looks exactly like SVR with a perfboard on the side, can you perhaps elaborate what is its purpose ?
From the SVR-T schematic it seems that the THR trimming should literally be a thermistor and a resistor but your board has a few more components.

Thanks


The perfboard is just a voltage regulator using an LM317L to produce 15V for the SVR-T.  I feed it about 18.5V.  Details are in an earlier post.

Orin.
Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on November 28, 2015, 11:28:58 am
Just a few days ago, I hooked up my 34410A again to the SVR-T that was powered on constantly since about August 2015.

August 2015:
Lab temperature: 23 to 24 °C
34410A showed 10.000,02 V with a 1.3 uV StdDv.

November 28, 2015:
Lab temperature: 20 to 21 °C (20.6 °C right now)
34410A shows 10.000,05 V with a 1.237 uV StdDv.
MIN: 10.000,04
MAX: 10.000,05

So, there is a 3 uV drift since August.
But it is probably contributed to a drift in the 34410A and the SVR-T
Or is this contributed to the 2 degree difference in temperature?

I still think that the SVR-T is outperforming all my 6 1/2 digit meters and also outperforming the 34470A 7 1/2 digit meter, as we have seen before.

 
Title: Re: Geller Labs SVR and SVR-T
Post by: F64098 on December 20, 2015, 07:02:35 pm
Hello,

i'm just trying to build a SVR-T, using a chinese AD587LQ.
Does anybody know the dimension of the fixed resistor R04 who is in series to the 100k thermistor THR1?
Are both resistors SMD?
I tried to use a little glas-thermistor in series with a 1 meg trimmer but i got no success, the output voltage ist drifting like a rally car.
Without that "compensation" the output is rocksolid for the actual temperature...

Thanks in advance

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on December 20, 2015, 10:36:52 pm
Hello,

The picture (of SRV) is on page 1 of this thread.
I do not see any SM-components.

R04 has to be adjusted to the T.C. of the uncompensated reference.

With best regards

Andreas


Title: Re: Geller Labs SVR and SVR-T
Post by: HighVoltage on December 20, 2015, 10:40:46 pm
Hello,

i'm just trying to build a SVR-T, using a chinese AD587LQ.
Does anybody know the dimension of the fixed resistor R04 who is in series to the 100k thermistor THR1?
Are both resistors SMD?
I tried to use a little glas-thermistor in series with a 1 meg trimmer but i got no success, the output voltage ist drifting like a rally car.
Without that "compensation" the output is rocksolid for the actual temperature...

Thanks in advance

Frank

You can find descriptions here:
http://www.gellerlabs.com/svr%20series.htm (http://www.gellerlabs.com/svr%20series.htm)
Title: Re: Geller Labs SVR and SVR-T
Post by: F64098 on December 20, 2015, 11:16:26 pm
Hello,

The picture (of SRV) is on page 1 of this thread.
I do not see any SM-components.

The description says: "After installation of the thermistor and gain resistor on the underside of the board,"
I think, they are SM-components.

Quote
R04 has to be adjusted to the T.C. of the uncompensated reference.

I know, but i'm not sure, if R04 must have 50kOhms or 200kOhms e.g.
I tried some constellations but i got heavy drifting, maybe produced by the little thermistor which i placed between the socket and the housing of the AD587 and the resulting bad thermal coupling.
A SMD-thermistor, solded near the pins of the AD587 would be the better solution.

Best regards

Frank




Title: Re: Geller Labs SVR and SVR-T
Post by: Gyro on December 20, 2015, 11:18:37 pm
The thermistor and resistor on the SVR-T are mounted on the back of the PCB, the resistor is a fixed value T/H (like the other black ones on the board) rather than a trimmer (of dubious TC). Joe selected the resistor value during cycling in his oven. Once tested, he covered both parts with blue thermally conductive resin.

The temperature trimming process is described here:

http://www.gellerlabs.com/SVR_TempcoTest.htm (http://www.gellerlabs.com/SVR_TempcoTest.htm)

EDIT: The temperature compensation was adjusted at the customer requested operating temperature - it is only good for a few degrees either side.
Title: Re: Geller Labs SVR and SVR-T
Post by: lars on December 21, 2015, 02:59:04 pm
Hello Frank

As I have been through this, my recommendation is to first check what temperature coefficient you have before adding the NTC + resistor. Normally the useful range for temperature compensation is up to 3ppm/C in the SVR-T. That is also the normal TC´s for the LQ as it has the S-shaped temperature curve adapted for 0-70C, this is probably also the reason the UQ is not so well suited for the room temperature range.

Next step is to check with a 1Mohm resistor instead of the NTC + resistor. It shall move the output voltage 25+-5ppm. If so a rough estimate are to use a resistor of about 250kohm for 1ppm/C, 150kohm for 2ppm/C and 100kohm for 3ppm/C. I set up an Excel sheet for this. Joe used Labview for his automated procedure.

Using a small NTC is not recommended if no other way is used to get a long time constant. I used a through hole NTC like Digikey BC2302-ND. I think Joe also did that. I attached it close to the 587LQ and also boxed the SVR board. Joe used epoxy to get a close coupling to the board that gave more time constant.

Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: ltz2000 on December 21, 2015, 03:49:46 pm

The same basic circuit, but more compact layout. The thermistor glued on the top of the reference gives very good thermal tracking.

https://www.eevblog.com/forum/projects/diy-compact-dc-voltage-standard/msg805614/#msg805614 (https://www.eevblog.com/forum/projects/diy-compact-dc-voltage-standard/msg805614/#msg805614)


(https://www.eevblog.com/forum/projects/diy-compact-dc-voltage-standard/?action=dlattach;attach=183357;image)
Title: Re: Geller Labs SVR and SVR-T
Post by: Gyro on December 21, 2015, 07:35:51 pm
Hello Frank

As I have been through this, my recommendation is to first check what temperature coefficient you have before adding the NTC + resistor. Normally the useful range for temperature compensation is up to 3ppm/C in the SVR-T. That is also the normal TC´s for the LQ as it has the S-shaped temperature curve adapted for 0-70C, this is probably also the reason the UQ is not so well suited for the room temperature range.

Next step is to check with a 1Mohm resistor instead of the NTC + resistor. It shall move the output voltage 25+-5ppm. If so a rough estimate are to use a resistor of about 250kohm for 1ppm/C, 150kohm for 2ppm/C and 100kohm for 3ppm/C. I set up an Excel sheet for this. Joe used Labview for his automated procedure.

Using a small NTC is not recommended if no other way is used to get a long time constant. I used a through hole NTC like Digikey BC2302-ND. I think Joe also did that. I attached it close to the 587LQ and also boxed the SVR board. Joe used epoxy to get a close coupling to the board that gave more time constant.

Lars

Lars,

Many thanks for posting the details. I had been wondering about the compensation relationship myself and I suspect several others had too.  :-+

Chris
Title: Re: Geller Labs SVR and SVR-T
Post by: F64098 on December 21, 2015, 09:18:04 pm
Hello Lars,

i say also thank you for this excellent explanation. :-+
Using a 220k resistor in series with the thermistor should perfectly work for my AD587.

I'm also working on an layout for a small PCB, fitting into a TEKO tin-box (45*48mm).
I will use a small capacitor on the output and connect the output wires directly to small solder pads,
putting some ferrite beads over them, to reduce EMI problems and avoid additional solder joints.
There are also additional pads for 3 types of precision trimmers and for paralleling the trim resistors R01 and R03.

What do you think about a complete ground plane for increasing the "heat" dissipation?
Joe's boards don't use a complete plane. Instead he uses 2 Oz. traces " to minimize errors caused by trace voltage drops".
The whole circuit runs at 2 mA, so i think doing this would be very volt-nutty?!

Best regards

Frank

Title: Re: Geller Labs SVR and SVR-T
Post by: lars on January 16, 2016, 08:18:31 am
Hello Frank

About the ground plane I think it is mostly for bypassing not so much for heat as it is so thin compared to the glass fibre.

Be careful about DC current not going in the output wires. I think it is a good idea to take them close to pin 4 respectively pin 6 of the AD587. But as you say 2mA is not a lot and also not changing if you don´t put load on the output as I think you aren´t.

Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: lars on January 16, 2016, 08:23:50 am
I just tested three AD587JQ I ordered from Digikey and all three seems to have low temperature drift. Of course it is a risk as the spec is so bad but my three at least is useful for experiments with SVR-T like boards.

As AD not makes AD587LQ and KQ anymore some JQ probably will be as good as the LQ and KQ I guess.

Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on January 16, 2016, 11:21:49 am
Hello Lars,

I also got recently 3 AD587JQ (from RS).
But have not done any measurements yet.
Mine are all with DateCode 1342 (Philippines)

Which datecode do you have?

With best regards

Andreas

Title: Re: Geller Labs SVR and SVR-T
Post by: lars on January 16, 2016, 01:22:30 pm
The date codes on the AD587JQ from Digikey was 1517 and Philippines.

Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: F64098 on January 16, 2016, 01:32:36 pm
About the ground plane I think it is mostly for bypassing not so much for heat as it is so thin compared to the glass fibre.
Hello Lars,

but i think it may help to egalize "hot spots" or to thermally connect grouped parts.

Quote
Be careful about DC current not going in the output wires. I think it is a good idea to take them close to pin 4 respectively pin 6 of the AD587.
When you look at my PCB in the other thread, you see a strictly star ground, centered on Pin 4 of the AD587.
Both groundplanes (pictures of the PCB not shown) are only connected to this point.
And also the trace to the output is placed on the top layer while the trace to the trim circuit is on the bottom layer.

Quote
But as you say 2mA is not a lot and also not changing if you don´t put load on the output as I think you aren´t.
All things that will be connected to the output, are only a load for my purse  :-DD

With best regards

Frank
Title: Re: Geller Labs SVR and SVR-T
Post by: zlymex on March 13, 2016, 12:54:56 pm
Hi,
Since this is AD587 and voltage standard related topic, I'll put something here.
Look at what I got today:
(https://www.eevblog.com/forum/testgear/geller-labs-svr-and-svr-t/?action=dlattach;attach=208133;image)
More than 300 recycled AD587KR. I've tested 100 of them and they are all good with voltages between 9.996V to 10.002V. I plan to use them in a two stage parallel(of 64 to 100 units) to achieve the lowest noise of a solid state reference :)
Title: Re: Geller Labs SVR and SVR-T
Post by: lars on March 14, 2016, 08:09:57 pm
Nice to parallel up to 100 AD587 to get low noise. I have done the same with 12 in parallell. With just 12 in parallell it is critical that all has the same low noise otherwise it is of no use. With 100 it may be less critical. As I have had several units from two different batches (via Digikey and RS) with very high noise I have tested all individuals but it takes very long time. The noise was about 4ppm pp random over seconds so even 100 in parallell may be affected. The noise was so high that I saw it with just a six-digit DMM during initial tests.

Lars
Title: Re: Geller Labs SVR and SVR-T
Post by: Andreas on March 14, 2016, 10:22:58 pm
Hello Lars,

from theory the typical noise of a AD587 should be around 6uVpp. (0.6 ppm peak-peak).
So what happened here with 4 ppmpp = 40 uVpp?

The 3 AD586 that I have measured (similar only 5V instead of 10V) are all below 3 uVpp (0.6 ppmpp).

with best regards

Andreas