A Voltage Reference Lab Experiment: ADR441 "Burn-In" 1000 Hours (10 ICs)

[CaptnYellowShirt]

Any news on this?

I got tired of discussions of my somewhat sloppy method, so I have cleaned up the PSU common connections and started over.

Yeah I glossed over a lot of that crap. I think you said it best in your first post: "There is quite a lot of (different ?) 'knowledge' and experience distributed on this forum. I for one am in for learning ..."

There's really only one way to know something and that's to try. Everything else is just feather preening.

For one, I'll be interested in hearing how it went.

[macfly]

[/quote]

Mhm,

On the other side you could try to improve the EMF sensitivity by placing a 1-100nF capacitor at the output of the SVR reference.
On your ADR441 devices you have 100nF at each output if I interpret the white board correct.
I do not know wether the AD587 will be stable with capacitive loads larger than 1nF because 1nF is the maximum mentioned in the data sheet. Perhaps the best filtering will be a 100 Ohms between the SVR output and the 100nF capacitor going to the high impedant input of the 34401A.

With best regards

Andreas
[/quote]

I agree with Andreas. But cap's are not the only way. EMF sensitivity drives me sometimes near crazy. For my Geller Reference, I made a serious solution, to get rid of all these unwanted RF:

Metal housing, battery operated, no connection wires to the DMM:

[Andreas]

Metal housing, battery operated, no connection wires to the DMM:

Hello,

there are still (short) connections to the DMM.
Try to put the DMM into the metal housing.
Otherwise you will have to filter the both outputs against the housing. (Common mode)
And the + against the - (Differential mode).

With best regards

Andreas

[macfly]

Metal housing, battery operated, no connection wires to the DMM:

Hello,

there are still (short) connections to the DMM.
Try to put the DMM into the metal housing.
Otherwise you will have to filter the both outputs against the housing. (Common mode)
And the + against the - (Differential mode).

With best regards

Andreas

Hi Andreas,

I agree with you. But unfortunately my DMM, a HP 3457A, has no 'guard' connector, as
e.g. the 3456A has.

Regards,

macfly

[bingo600]

I agree with you. But unfortunately my DMM, a HP 3457A, has no 'guard' connector, as
e.g. the 3456A has.

Regards,

macfly

Hmm strange

HP specificaly mentions it in this publication (Page. 15 ff.)
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1986-02.pdf

Actually a nice walkthrough of the 3457A

/Bingo

[CaptnYellowShirt]

I agree with you. But unfortunately my DMM, a HP 3457A, has no 'guard' connector, as
e.g. the 3456A has.

Regards,

macfly

Hmm strange

HP specificaly mentions it in this publication (Page. 15 ff.)
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1986-02.pdf

Actually a nice walkthrough of the 3457A

/Bingo

The 3457a has an internal guard shield (e.g. not mains-earth grounded), but no way to hook into that from the outside like many other high-resolution dmm's offer.

[macfly]

[/quote]

Hmm strange

HP specificaly mentions it in this publication (Page. 15 ff.)
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1986-02.pdf

Actually a nice walkthrough of the 3457A

/Bingo
[/quote]


Thanks for the link - indeed, a very interesting article!

Regards

macfly

[JoeN]

I can't believe no one asked this.

How did you get a sample of 10 ICs?

Analog will only ever send me two.  Thought that does work for me.

[quantumvolt]

As I wrote in Reply #4 in this very thread:

"When I asked for these devices I think I said 'a handful different references from the ADR44x family'. I was planning to compare them to my LTC6655 2.5V and 5V references."

Obviously - as is very, very common these days - someone didn't read what I was asking for, and sent me 10 equal devices.

Concerning begging method: I just sent a polite and well formulated email to ADI ("Contact us" on the web site). With no promises on what I would do with them, and stating clearly that I would only accept them if ADI demanded nothing in return.

If one learns to read, think and write - and has some (even just a private) cause - I guess one can sample a 3458A ...

[JoeN]

As I wrote in Reply #4 in this very thread:

"When I asked for these devices I think I said 'a handful different references from the ADR44x family'. I was planning to compare them to my LTC6655 2.5V and 5V references."

Obviously - as is very, very common these days - someone didn't read what I was asking for, and sent me 10 equal devices.

Concerning begging method: I just sent a polite and well formulated email to ADI ("Contact us" on the web site). With no promises on what I would do with them, and stating clearly that I would only accept them if ADI demanded nothing in return.

If one learns to read, think and write - and has some (even just a private) cause - I guess one can sample a 3458A ...

In my case, they allow you to register and then allow you to order online without resorting to emails.  I've been quite successful this way, it's just you can't order more than 2 devices per part number and 4 different part numbers per order.  You can place two orders per month.  On the other hand, you can order two of the $70 devices too and they have never denied me.  Way back in the day I begged U.S. Robotics for an 14.4K HST modem on a BBS system operator discount though I wasn't running a BBS at that moment (I had for years).  I just told them I was intending to set a new one up.  That had to be done via old fashioned letter.  I ended getting one of their $600 modems for something like $250.  The prices back then were really insane.

I guess someone made a judgment call since you were ordering a super cheap part, why not give you a bunch of them?  My last order had a AD5791 DAC, a $60 part.  I doubt I could sweet-talk them into sending me 10 unless I had an Agilent or Tektronix email address.  OTOH, 2 are enough for my purposes anyway.  They sent it no problem.

[quantumvolt]

Well - I am sorry I misunderstood you. I do not register and try to sample as much dollar worth as often as I can. I guess the real cheapos double-sample by also making an account in their little sister's name ...

I got a few refs that I do not need (I have more precision refs than I will be able to power up before I die) for an experiment. The experiment is closed for now.

This thread was about 'burn-in' / drift in voltage references. Please post your sampling advice somewhere else. Thank you.

[JoeN]

Well - I am sorry I misunderstood you. I do not register and try to sample as much dollar worth as often as I can. I guess the real cheapos double-sample by also making an account in their little sister's name

It is not my intention here to criticize your sample acquisition method.  It's just that your way of doing it was unknown to me and I appreciate that you explained it.  In return, I explained how I do it in case you were not familiar with it.  In any case, I didn't order that part just to maximize the published price of the chips I am getting.  I took it specifically because AD has a very nice circuit note on how to construct a 1PPM voltage source from this part:

http://www.analog.com/en/circuits-from-the-lab/cn0191/vc.html

And I am building it.  It's almost done.  It is being built around an ATTMega328P controller and an Analog voltage reference (interestingly, they suggest a Linear voltage reference).

On that order I also ordered the AD8675, AD8676, ADR445.  These are all far less expensive parts (buffer amps and a precision voltage source) needed to construct the circuit.

[CaptnYellowShirt]

Way back in the day I begged U.S. Robotics for an 14.4K HST modem on a BBS system operator discount though

Its sad to think some people today don't even know what a modem handshake sounds like.

[quantumvolt]

... AD has a very nice circuit note on how to construct a 1PPM voltage source from this part:

http://www.analog.com/en/circuits-from-the-lab/cn0191/vc.html

And I am building it.  It's almost done.  It is being built around an ATTMega328P controller and an Analog voltage reference (interestingly, they suggest a Linear voltage reference).

On that order I also ordered the AD8675, AD8676, ADR445.  These are all far less expensive parts (buffer amps and a precision voltage source) needed to construct the circuit.

Very good. Many people here are interested in AD5791 and the LT ref that has an unnamed footprint on the ADI demo board - namely LTZ1000.

I suggest you show your project either in the LTZ1000 thread, the Seven Decade Voltage Calibrator thread, or make a new one.

I would be very interested in your project. Good luck.

[Andreas]

(interestingly, they suggest a Linear voltage reference).

 These are all far less expensive parts (buffer amps and a precision voltage source) needed to construct the circuit.

Which Linear reference?

I would use a 10V reference (AD587 in CERDIP for long term stability) for this circuit together with a temperature sensor for correction.
Otherwise the most expensive parts will be the resistors to scale the 5V to +/- 10V.

With best regards

Andreas

[quantumvolt]

EVAL-AD5791SDZ_LTZ1000A.pdf (145.8 K)
EVAL-AD5791SDZ Bill_Of_Materials.xls (40.5 K)
AD5781_91_Gerbers.zip (447.6 K)

Location: http://ez.analog.com/message/67799#67799

[JoeN]

(interestingly, they suggest a Linear voltage reference).

 These are all far less expensive parts (buffer amps and a precision voltage source) needed to construct the circuit.

Which Linear reference?

I would use a 10V reference (AD587 in CERDIP for long term stability) for this circuit together with a temperature sensor for correction.
Otherwise the most expensive parts will be the resistors to scale the 5V to +/- 10V.

With best regards

Andreas

Thanks for the input.  The part that AD suggests is LTZ1000 like quantumvolt wrote.  On the eval board that you can *buy* from Analog, there is an unpopulated space for this part, or you can provide your own reference.  I've included some circuit diagrams and the image of the board they sell.

I chose an ADR445 as the reference.  Was this a bad move?  My main motivation was that it was free but still seems to have very good specs.  Also, I don't have the precision resistor values needed to precisely double the voltage but I do have matched pairs of .01% Vishay resistors (odd value - 681 ohm) that are allowing me to create a precise -5.000V off the 5.000V source.  So I am giving up on -10V/10V and going with -5V/5V, at least the first go-around.  I know this is all fishy and is extremely unlikely to get me to the promised land of 1PPM but hopefully that reference is at least stable so I can trim it out in software.

First image, the unpopulated reference on the eval board.  Second image, the eval board I am trying to somewhat clone.  Third image, general suggested application schematic for the DAC.

[Andreas]

I've included some circuit diagrams and the image of the board they sell.

I chose an ADR445 as the reference.  Was this a bad move? 

Hello,

now its clear why they recommend a LT device. The LTZ1000 is the best of all references.

I have only tested the ADR4550B which is specced 25ppm/khr against 50ppm/kHr ageing drift of the ADR445.
The drift of the ADR4550B (ADC18) was about 90ppm for the first half year. And even after this there are some "jumps" (due to humidity changes?) of a few ppm each week that I do not  have on a well aged hermetically tight device like the AD586LQ (ADC13).

With best regards

Andreas

[JoeN]

now its clear why they recommend a LT device. The LTZ1000 is the best of all references.

I just checked and Linear won't sample this part.      

[Andreas]

I just checked and Linear won't sample this part.      

They know that you cannot afford the 5 precision resistors at $30 each ;-)

The 2nd best is the LM399 (at stock @ digikey)

With best regards

Andreas

[JoeN]

The 2nd best is the LM399 (at stock @ digikey)

OTOH that part can be sampled.   

[Andreas]

Hello,

even a clean PCB absorbs some humidity and swells.
The same is with plastic DIP packages.
At least so much that you can see it in the ppm-range.
My LT1027CCN8-5 have a humidity sensitivity of 0.5ppm/% rH.

The only way to avoid this is to use hermetically packages (metal or ceramics) and seal the pins with glas.
But even in this case you have to carefully decouple the package from the PCB.
Otherwise the humidity change of the PCB will create similar changes in the range of about 0.5ppm/% rH.

With best regards

Andreas