Newbie Precision Reference Calibrated with the 3478A Dinosaur

[daqq]

Hi,

Great device! But how exactly do can you calibrate the 25 uA/ 2.5mA ranges with this?

Also, the 1uF cap (C3) directly on the op amp output could lead to some nasty stuff.

David

[hgg]

Hi,

The current reference was not designed to be calibrated... but if you can accurately measure its output,
then you can use the measurement for calibration.  It is dependent on the absolute values of the resistors.
The UT136C with its voltage range calibrated shows 2.50mA & 25.0uA.  UT61E with its high burden voltage
shows 2.472mA & 24.78uA...

Also, the 1uF cap (C3) directly on the op amp output could lead to some nasty stuff.

Like what?  Can you please explain?

[daqq]

Short version: It adds a direct capacitive load to the op amp output, which it might not like and start oscillating - essentially, going into constant overshoot, desperately trying its best to regulate the output. It might not happen and it might not apply, but it might. You probably wouldn't see it on the output on the HP, but you might see it on a scope.
Long version: http://www.analog.com/library/analogdialogue/archives/31-2/appleng.html

[daqq]

If you have LTSpice, run attached simulation. Look at the output voltage.

[hgg]

Thats exactly what happened when I did not use it... 

From my little experience, Chopper Amps are a bit picky and when I was powering the circuit
from my power supply that was connected to the mains, I had some problems with it which
disappeared when I used the capacitor.  There is no problem with the battery and if you look
the output voltage on the oscilloscope (ac coupled) it is a very steady flat line down to 2mV
which is the limit of the scope.

[Kevin.D]

If you have LTSpice, run attached simulation. Look at the output voltage.

Hi George .
I  remember you had a thread about this opamp  'ltc1250' . It's not very good as a buffer amp because it's got very little phase margin at unity gain,( it can't source much current either about 5mA? (bet the ref can source that). It's got  About ~ 10 deg Ph margin, so it's not very stable at a gain of 1, even without any capacitive load. If you want to use an opamp like this at a gain of 1 with a large capacitive load  then we have to isolate the output of opamp from the CAP with a resistor say~ 20 - 100. Then use a resistor and smallish cap in the feedback to lower the bandwidth and so set a decent phase/stability margin.  If your having probs (it WILL oscillate  with a load cap that big on that opamp, your's might only just about be stable due to a high esr cap). To ensure stability try something like a 20R on the output and 1k and 47n in feedback like below .( Vi,Ii are just  probes)

[hgg]

Hello Kevin,

I had some difficulties finding a chopper amp that could easily source more than 2mA
and have a low drift as well.  The LTC1250 can output full swing into 1k load and can
easily drive the 2.5mA that I needed.   The output voltage is pretty stable as it is right
now, but I will try your suggestion as well.  I remember that it was stable with a much
smaller capacitor as well.

Maybe I was lucky but the current circuit is rock solid with repeatable and accurate
measurements down to 5 significant digits after 5 months. 
(always at the same ambient temperature though...)

[Andreas]

I had some difficulties finding a chopper amp that could easily source more than 2mA
and have a low drift as well. 

You could try the LTC2057.

I've checked again the reference with the 3478A and after 5 months its still showing 2.50005V
Either its very stable or both the reference and the 3478A are drifting together.

The MAX6325 is a buried zener reference -> is relative stable over time.
I have here similar (5V) references (MAX6350) from the same series which have drifted about 25-30 ppm over 800 days.
(the blue line is the drift of my "instrument" against a LTZ1000A reference).
So with your instrument resolution of 4 ppm for the 2.5V you will need some further time to recognize the drift.

BTW: which reference is used in the HP3478A?
In the service manual it seems to be a LM399.
So if its well pre aged the drift of the instruments reference should be even lower.
(buried zener within hermetically sealed package).

With best regards

Andreas

[Kevin.D]

   The output voltage is pretty stable as it is right
now, but I will try your suggestion as well.  I remember that it was stable with a much
smaller capacitor as well.

Without the extra compensation it looks to be stable loop as long  as the esr of the load cap > ~ .5 Ohm . So  1 uF electrolytic should be stable (esr 2-8ish)  .But change it for a low esr  ceramic and it will start  . 

[hgg]

Hello Andreas,

Nice graph.  You tracked the drift pretty nice.  The red graph is interesting because I read that
the ageing of the zeners is unidirectional, meaning that if it starts lets say positive, it will stay
positive.  In your graph although it starts negative it then goes positive for some reason.

The HP3478A is well pre aged but I don't remember which reference it has.  I would have to check.
Maybe I will also check my reference once a month and make a crude graph.


Kevin all three capacitors used on the reference are 1uF tantalums which have an ESR of around 1 Ohms,
so maybe thats why its stable.

[Andreas]

Nice graph.  You tracked the drift pretty nice.  The red graph is interesting because I read that
the ageing of the zeners is unidirectional, meaning that if it starts lets say positive, it will stay
positive.  In your graph although it starts negative it then goes positive for some reason.

Hello,

dont forget that this may only be valid for the bare zener. (usually 7V references).
A monolithic 2.5 or 5V reference with plastic housing has several ageing effects:
- the zener (+ the current source).
- the resistor divider which may oxidize due to humidity
- the (plastic) housing (or the epoxy pcb) which swells with humidity and creates mechanical stress to the chip
  on some chips you can see this as seasonal changes (autumn / fall)
- and finally the die attach which is used to cement the chip to the lead frame
.....

with best regards

Andreas

[blackdog]

Hi,

I want to show yo guys a way of not loading your chopper opamp.
Look @ the schematic, it uses various techniques to kill noise and to be shure not to load the opamp.
The J310 HF Fet is cheap. and most of the time it can source 8 a 10 mA.
The opamp sees no load and you wil get a better output commonmode, dropout is about 0,2V.

There a to techniques in this schematic to kill noise, he first one is the "bootstrap" capacitors C1 and C2, its a low leakage current
lowpass filter of about 3,5Hz.
There is almost no leakage current through C1 because the voltage over C1 is very low(serveral mV).

The second filter is C3 and C4 and R8.
This is just an schematic i'am testing at the moment to see how much noise is filtered.
The LM10 is not good enough to be ussed as a good Reference circuit.
But... Use for the 5V section a LT1021, Max6126, LTC6655, and for the second stage a LTC2057, LTC1250 and you wil have a super reference :-)

Happy building!



Kind Regarts,
Blackdog

[hgg]

@Andreas,
So, its not so simple...
Thank you.

@Blackdog,
Thank you for the schematic.  I like the JFET solution but I have some questions.
Have you built the circuit?  Why all this filtering?  What is the source of the noise?
Have you measured the noise you are trying to filter?
When you say "dropout is about 0.2V", dropout voltage of which component? 
Do you fine tune the unity gain of the op-amp with the P1?
Is the long term stability of the reference now also dependent on resistors R10 & R5?

Regards,
George.

[blackdog]

Hi hgg,

A fast reply, buissy... ;-)

Yes i build it, its testing now, for temperature stability, and no, the LM10 is not a good Reference.
Why, the reference voltage is multiplied by 50 times!.

I made this, to do measurements on the noise behavior, and still have to do this measurements, but no time, work, work, work..

To pictures, the first one is the "stability" in 16,5 hour, the upper meter is the 5V output, the lower meter is the 10V output.
The temperature shift is about 10C, lower voltage is colder, Temp coefficentie is about 0,5mV/C @ 10V DC output.



And this is the box and the test circuit.


Later more...

Kind regarts,
Blackdog