Project KX : DIY calibrator / reference source/meter

[Andreas]

Hello,

For the voltage divider I would at least use some UPF25 resistors (5ppm/K) or even better.
All what is not below 1ppm/K spoils the accuracy of the LM399.
I measured around 16uV drift for one thermally shielded LM399 over a 8-46 deg C environment range.

Instead of R15/R16 connecting to pin 3 of the LTC2057, I would connect 2 individual resistors from output of LTC2057 directly to the PIN1 of each LM399.

With best regards

Andreas

[TiN]

Well, there are two reasons for choosing worse resistors, than most would use.

1. I don't want to wait another half year for VPG resistors and have 300usd BOM for 1 board. I have LTZ1000 version for that
2a. Whole board will be in constant temperature box, so thermal gradients will be constant and controlled. There is temp sensor IC between VREFs, so I can play with some software correction for ratio as well.
2b. I already have handful of PTF and RNC55/60 resistors to play with.

Layout placement



As of R15/R16... May share a little more light why such approach is better?
I also want use same circuit to boost 7.xx V zener voltage to 20.0000V (for my K2001's calibration reasons I need 2V and 20V)

I attached modified circuit from simulator, with one zener for clarity, to confirm I got idea correctly.

[Andreas]

In your cirquit the 1mA current for the LM399 is flowing through the series resistors R3 R23.
so you are not sensing+amplifying the zener voltage but the zener voltage + 500mV (500R * 1mA).

With best regards

Andreas

[Andreas]

2a. Whole board will be in constant temperature box, so thermal gradients will be constant and controlled.

I attached modified circuit from simulator, with one zener for clarity, to confirm I got idea correctly.

In this case a AD587 might be the better solution. (with carefully trimmed resistors).

Sorry but this cirquit looks for me totally different to the dual version.

With best regards

Andreas

[wiss]

In your cirquit the 1mA current for the LM399 is flowing through the series resistors R3 R23.
so you are not sensing+amplifying the zener voltage but the zener voltage + 500mV (500R * 1mA).

With best regards

Andreas

There's no DC-feedback.

[TiN]

That what I got confused too, considering suggestion below...

Instead of R15/R16 connecting to pin 3 of the LTC2057, I would connect 2 individual resistors from output of LTC2057 directly to the PIN1 of each LM399.

[Andreas]

There's no DC-feedback.

You are right its a DC-feed forward.

@TiN
Im not shure wether we are talking about the same cirquit. (cutout attached below)

If you look at the cirquit mentioned above you will clearly see that the LM399 zener has 2 current paths for the 1 mA total zener current.
One is thru the R2 pull up with below 0.1 mA. (only for Start-up)
the other is from the 20V output through 7K5 (R15-R17) 2mA and through R3 + R23 to the both zeners. (1mA each)
Of cause you will have to adapt R15-R17 if you have only a 10V output.

And of cause you will have to do a carefully design that the zener current does not change by more than 0.1% over ageing + temperature.
So in my opinion the start up current through R2 is too high if the 24V are not extremely well stabilized.

With best regards

Andreas

[MK]

another way to solve the self-start issue is to use a fet between the output of the opamp and the zener bias chain, then the 200k resistor is no longer needed ans the bias current is more closely controlled.

[wiss]

You mean JFET?

[Andreas]

another way to solve the self-start issue is to use a fet between the output of the opamp and the zener bias chain, then the 200k resistor is no longer needed ans the bias current is more closely controlled.

How do you stabilize the zener current in this approach?
The solutions that I know need a additional op amp + resistors for the current stabiliziation.

With best regards

Andreas

[wiss]

How do you stabilize the zener current in this approach?
The solutions that I know need a additional op amp + resistors for the current stabiliziation.

OpAmp-out -> JFET-gate,
JFET-drain to V+,
negative feedback from JFET-source,
current-setting resistor from JFET-source to Zener.

When OpAmp-out is zero the JFET will conduct, ie. the OpAmp-out will sit several volts below the 10 (20) V output

[MK]

How do you stabilize the zener current in this approach?
The solutions that I know need a additional op amp + resistors for the current stabiliziation.

OpAmp-out -> JFET-gate,
JFET-drain to V+,
negative feedback from JFET-source,
current-setting resistor from JFET-source to Zener.

When OpAmp-out is zero the JFET will conduct, ie. the OpAmp-out will sit several volts below the 10 (20) V output

Bingo, the zener will always self start, and max current available is limited by IDSS, seems a good anwser to the self starting behaviour to me.

[TiN]

Assembled LTZ-based 20V ratio amplifier to test, breadboarded before PCB layout.



Signal path:

Amplifier raw LTZ output->20V : 3 x 9V LiPo battery power -> BNC input -> LTC2057 with 25ppm 0.1% resistor arrays -> BNC. Whole rig put into metal can (non-hermetic). BNC output goes directly to Keithley 2001 for data collection.

LTZ reference : Keithley 2400 +15V as power source -> BNC -> hermetic IKEA can -> SMA+/SMA- -> Amplifier BNC input. Reference is in striped metal can.



Second and third LTZ reference burning on another rig with open KI2001.
Power source: Linear +15V power supply powered from mains 110V.



Going to order fixed revision for LTZ PCB and LM399 version monday.

[branadic]

Hi TiN,

what about your offer you've once made to sell assembled LTZ reference boards? Is this offer still valid?

branadic

[TiN]

Yes, if long assembly time does not scare people. VPG resistors are not quick to get, last time it took almost 4 month.
Z202 plastic ones should be around a month to buy. So, if that's ok, modules will be around jan.
Also making dual LM399 version proto as discussed above on previous page. That will be way cheaper as well, and provide 10/20V output too.

[TiN]

Layout, same format as LTZ one.

[branadic]

As far as I remember you wanted to provide a complete set with pcb and all parts, won't you? I also remember that I've signed up for such a set. Am I'm missing something?

branadic

[TiN]

No, you not missing anything
The only issue was a timeframe, as it took half year to get resistors here, which is kinda longer than expected.
And I just got first prototypes (5 boards, 3 with A and 2 non-A) debugged and started aging.

Also it was not a set, but a assembled units promised, as well as bare PCBs (no parts).
But I don't think there would be much people willing to pay 300USD for jumpwires and bodges, so I hold off on selling any prototypes to
those few who asked for complete LTZ reference module.

So to clear things out:

* Fixed revision board is on it's way, no bodges or jumpwires.
* Resistors to be used: Z202 120R, 1K, 13K 0.1%, VHP202Z 70K 0.1% (need order)
* LTC2057 opamp (I have onhand)
* LTZ1000A or LTZ1000 (Need order, takes a week to get from Linear webstore)
* At BOM cost + shipping (EMS Express, a week to US or Europe usually)

Also list of all people who wanted anything are readily available.

[TiN]

Test prototype with wacky connections as a concept test (fail test?)

LTC2057 with array of 25ppm resistors.



1 week of data log.



Powered by 3 x 9V LiPo batteries, measured voltage about +25VDC.
Single foil resistor instead of bunch of PTF56's seems worse (graph after 75000 x-axis point).

[TiN]

Happy Xmas for all voltnuts.
Just today got my gift from santa. 



Top left: Dual LMx99 with LTC2057 7V->10/20V output. Differential inputs/outputs via SMA. FR4 4L
Middle top left: Old LTZ1000 alpha-version
Top right: Revised Rev B of LTZ1000. Fixed heater circuit, fixed U5 opamp power layout. Minor tweaks.
Bottom: Bananna plug boards, 5 boards shorting all plugs by various configuration, B-SMA is adapter to SMA jacks.

Back of boards:



Aren't they beauty with gold plating?



Guards and ground openings.



Top size of fixed LTZ1000 reference PCBs





Center have minimum copper around pins to isolate zener thermally.



All PCB are 4-layers, stackup shot:



Can't wait for NY holidays to play with these toys  O0

[Vgkid]

Those are truly gorgeous. 

[AudioplatinumService]

Very nice!!

[TiN]

Assembled one of LTZ references.
Works a charm, right from the start.  O0

Front side.



Measurement log start:



Configuration:

Power source: 3 x 9V Li 800mA accu's, +25V output
Opamp: LTC2057 x 2
Zener: LTZ1000CH WW16 2014
Resistors: VPG VHP203Z 70K x 2 , Z202 13K, 1K, 120R

[TiN]

Hm, guess spoke too fast about working from a start
Set up a measurement/comparsion setup with 2 first revision LTZ references (which were 24/7 on during last few month) and one fresh black LTZ.

Measurement : K2002 with 2001-TCSCAN card. All references connected by ribbon cable.

Channel 4 - New black LTZ module
Channel 5 - Old LTZ 3
Channel 6 - Old LTZ 4

Data of 24 hours:



Vertical scale is +/- 5ppm on each reference voltage accordingly. Something fishy going on channel 4, right?

Hooked scope to output, and see 125mV of 54kHz oscillation on output. Older references don't have it.



Setup (overall photo clickable):



What is could be? Usual suspects like SMPS supplies nearby, LED lights, other test gear are ruled out by turning them off. Putting ref into metal can have no effect either.

[chickenHeadKnob]

Is unit 4 one with ltc2057 chopper amp? Think 54Khz is just above chopper main frequency according to LT chart: Input voltage noise VS freq.
The data sheet implies chopper freq around 48Khz but with some randomization or dither.

From the data sheet
www.linear.com/LTC2057

Input Voltage Noise
Chopper stabilized amplifiers like the LTC2057 achieve low offset and 1/f noise by heterodyning DC and flicker noise
to higher frequencies. In a classical chopper stabilized amplifier, this process results in idle tones at the chopping
frequency and its odd harmonics. The LTC2057 utilizes circuitry to suppress these spurious artifacts to well below the
 offset voltage. The typical ripple magnitude at 100kHz is much less than 1?V RMS. The voltage noise spectrum of the
LTC2057 is shown in Figure1. If lower noise is required, consider one of the following circuits from the Typical Applications
section: "DC Stabilized, Ultralow Noise Amplifier" or "Paralleling Choppers to Improve Noise."