Ultra Precision Reference LTZ1000

[MiDi]

Short update to the PS issue, will give details tomorrow.

It was oscillation (1Vpp@200kHz) between my SMPS and LT1763 - with linear PS everything is fine.

A quick look at the datasheet gives ~30dB ripple rejection @200kHz, so 1Vpp should turn into 30mVpp at the output - nope: there is ~50mVpp at FB-Pin (should be ~3mV).

This reminds me of something stated in datasheet of LT3045 - if I remember correctly:
The input cap in conjunction with SMPS should be placed a good portion away from the linear regulator to avoid interference - now I have seen why!

And here things are even worse, the regulator screwed up completly 

[branadic]

Good that you found the reason. For me it's a complete no go to combine precision circuits and SMPS. However, I suggest reading AN101 by Linear Technology for  using switchers and LDOs.

I'm currently soldering my boards. Some components arrived today. Could finish two boards by now, but it's to late to power them up. Another 3 boards are presoldered and another 5 are preassembled. Still waiting for some parts.

Please keep an eye on C7, C8 and C9, they should be soldered as far as possible from the edge, so the board fit's without problems to the case. Further you have to bend Q1 somewhat. Cut the contacts of the DC jack after soldering. If you consider that points the board fits without problems into the first slot of the Hammond case. However, I recommend to use some Kapton/Polyimid tape for the edge, before you slide the boards powered into the case. Solder maks is not a reliable isolation.

All capacitors are MKS2 (Wima) except C7 and C8, this are MKP2 (Wima). For the NTC choose whatever you prefer, it's up to you.

Pictures coming soon as well as drawing for the case.

-branadic-

Edit: All boards, except 6 which have been reserved for one use, have been shipped.

[MiDi]

Good that you found the reason. For me it's a complete no go to combine precision circuits and SMPS. However, I suggest reading AN101 by Linear Technology for  using switchers and LDOs.

Yes, it was a lesson learned, even for initial operation I will avoid SMPS in future
Ultra precision circuits and SMPS are tough to combine while maintaining the precision, but e.g. keithley did in DMM7510
Thanks for AN101, it was a good bed reading to get an idea of realistic numbers.

Hopefully last OT post, here are the promised details of investigation on LT1763 screw up (follow up of 2375, 2383 & 2388).

The temperature of LT1763 was measured with thermocouple with pcb upside down in free air @15V supply:
room temp: 21.4°C
max @warm up: 30.9°C (dT 9.5°C) - certainly not in equilibrium
min @steady state: 28.5°C (dT 7.1°C)

The reference is GND at power jack for all measurements:

1. Screw up on warm up due to resonance between SMPS and LT1763.
    LT1763 dropped from 12V to 10V output with 15V input and ripple of 1Vpp@200kHz.
    Same result with 16V supply, so this seems to be an issue, or is it a feature (expected behaviour)?

2. Apparently normal operation at steady state with expected 12V output - input ripple is reduced to 0.76Vpp.

3. Same at steady state with closeup of 12V ripple: 0.76Vpp/0.076Vpp gives attenuation of 10 (20dB), Datasheet states ~30db@200kHz - so it is a bit off.
    Interestingly the feedback has 40mVpp ripple, but should have only 1/10th of 12V ripple (7.6mVpp).
    There is a lot of current flowing between SMPS and input cap and as the probes are referenced to GND at power jack, there is a bit of voltage drop between GND at power jack and GND at LT1763 output involved.
    But luckily we can get the voltage drop due to excessive current from -Ref line, which is ~20mVpp, so ~20mVpp remaining at feedback.
    Should be in the ballpark as there was no appropriate probing possible (normal GND leads at probes).
   
4. Added 0.5m leads to SMPS - resonance has gone and turned into expected ripple for SMPS (down to 1/10th at ~100mVpp).
    Output now is 12V on warm up & steady state.
    So I had the luck to hit the resonance point of the system with choosen leads
   
5. For comparison: same with linear PSU 

[exe]

Yes, it was a lesson learned, even for initial operation I will avoid SMPS in future

Imho it's not about SMPS, it's about loopback compensation. A linear PSU is no different in this regard.

[Macbeth]

Is the LT1763 input capacitor C12 1uF/25V being an 0805 MLCC under specified? The data sheet calls for anything between 1uF-10uF, but this IC is usually used for <=5V (at least it's fixed versions are in that range). Perhaps a cap closer to 10uF would be better for the higher input voltages?

Any old MLCC capacitor could present a drastically lower capacitance at input voltages as large as 15-18V.

See Temperature and Voltage Variation of Ceramic Capacitors, or Why Your 4.7µF Capacitor Becomes a 0.33µF Capacitor

[branadic]

I use 0805 1µF X7R/50V capacitors. If you are afraid you can bodge a MKS2 1µF cap instead.

-branadic-

[exe]

Any old MLCC capacitor could present a drastically lower capacitance at input voltages as large as 15-18V.

I don't think it's  about age. It's about dielectric material and volumetric capacitance (bigger caps of small size tend to have more dc bias problems). The only way to know possible variation is (apart from measuring) to refer to datasheet or simulator. Big vendors like murata, kemet, avx provide dc bias curves. The effect can be very pronounced even at voltages below 5V.

Although, there is an aging effect as well: https://www.murata.com/support/faqs/products/capacitor/mlcc/char/0006

[Cerebus]

Any old MLCC capacitor could present a drastically lower capacitance at input voltages as large as 15-18V.

I don't think it's  about age. It's about dielectric material and volumetric capacitance (bigger caps of small size tend to have more dc bias problems). The only way to know possible variation is (apart from measuring) to refer to datasheet or simulator. Big vendors like murata, kemet, avx provide dc bias curves. The effect can be very pronounced even at voltages below 5V.

Although, there is an aging effect as well: https://www.murata.com/support/faqs/products/capacitor/mlcc/char/0006

<Chuckle> In idiomatic English "Any old x" is just an emphatic way of saying "Any x" possibly with a side suggestion that it's not new, novel, or in any way special.

[branadic]

Since one user didn't answer by now there are 5 boards left for the community. So for those of you that have been late, here is your last chance to participate. Simply write me a PN with your email address and postal address for further details + shipping and how many boards you need.

-branadic-

[branadic]

Sorry to say that, but all boards have been sold.

-branadic-

[exe]

I'm curious why LM399 and ltz1000 are not bypassed by a cap? Neither directly, nor with an RC-filter. Is this because of potential problems (with, e.g., leakage current, or opamp's input bias) or because this is not really needed?

[MiDi]

I know at least one dmm with LM399 that uses a bypass cap, think it was 0.33uF.
For LTZ1000 there is a hot discussion about bypassing.
If used good caps e.g. MKP there is no issue with leakage.

[BradC]

Sorry to say that, but all boards have been sold.

You say that like its a bad thing.

[Andreas]

I'm curious why LM399 and ltz1000 are not bypassed by a cap? Neither directly, nor with an RC-filter. Is this because of potential problems (with, e.g., leakage current, or opamp's input bias) or because this is not really needed?

Hello,

Most designs are rather old. So mobile phones, WIFI etc. did not play a role.
On LM399 a 100nF cap improves massively EMI behaviour.
Leakage is negligible compared to 1 Ohms output impedance.

On the datasheet cirquit of the LTZ1000 you will get a instable output without further components.
(so use either Dr. Franks (branadic PCB) or my design to improve behaviour).

When you look at the Fluke 732B specification it is a 0.18V/m specification without being out of tolerance.
Industry standard requires to withstand 10V/m so you will need a shielded RF cabinet to operate a Fluke 732B according to spec.

with best regards

Andreas

[branadic]

Little volt nut meeting with 9x LTZ references and 2x AD587 references on Keithley2002 and K3458A and lot of good discussion today in Stuttgart, together with some homework for the next meeting.

-branadic-

[exe]

What's the name of the power supply on the photo? (the one below Keithleys).

[2N3055]

What's the name of the power supply on the photo? (the one below Keithleys).

Looks like Hameg HM8143

[branadic]

What's the name of the power supply on the photo? (the one below Keithleys).

Looks like Hameg HM8143

Exact, it's a HM8143.

-branadic-

[casinada]

After all the effort on expensive multi-meters and nice voltage references, do the cables make sense? 

[ManateeMafia]

Looks like you have plenty of data collection going on there.

A couple of questions:
1. Do you have a gold standard reference? One that has been recently calibrated or measured against a calibrated source?
2. Do you plan on shipping a reference around for measurements against house standards?
3. The 2182A is the best meter there for series-opposition measurements. Is there a plan to create a special cable for it so you can make measurements against the others?
4. Do any of you have access to a Data Proof scanner?
5. Will the measurements be compared between line power and battery operation?

Good luck. I am sure more will want to join you by your next meeting.
BTW, I thought meetings in Germany included beer. I see no evidence of such. 

[Andreas]

Looks like you have plenty of data collection going on there.

A couple of questions:
1. Do you have a gold standard reference? One that has been recently calibrated or measured against a calibrated source?
2. Do you plan on shipping a reference around for measurements against house standards?
5. Will the measurements be compared between line power and battery operation?

Hello,

yes around 12.2 Mb raw .csv data. It will need some time to evaluate all.
1. the 3458A was recently calibrated at Keysight Böblingen, so we are within 90 days specs.
2. I plan to ship (within EU) my AD587LW#02 after having a shippable read out interface. (but this will be part of a new thread).
5. My references are only specified for battery operation.
    Branadics reference LTZ1000#3 has been tested for both without and with line power.
    (I can not see any significant difference before and after the glitch due to connecting).

Some impressions:
My LTZ#4 on HP3458A some minutes: Standard deviation 0.2 uV
A bit more noisy the travel standard AD587LW#02: Standard deviation 0.6 uV (offtopic here)
Branadics LTZ1000#3 (10V option unadjusted) battery and line powered (after the glitch)

with best regards

Andreas

[branadic]

After all the effort on expensive multi-meters and nice voltage references, do the cables make sense? 

You are right, good cables are important. But to be honest, 100€ and more for a ready to go 1m PTFE cable with gold-plated spade lugs or simple Multi Contact (former Staubli) connectors are half the budget of a LTZ reference und pretty expensive. I'm still searching for an economic solution or a source for good twisted pair, shielded PTFE cable. This is what I took for homework to be done until next meeting. 

2. Do you plan on shipping a reference around for measurements against house standards?
3. The 2182A is the best meter there for series-opposition measurements. Is there a plan to create a special cable for it so you can make measurements against the others?
4. Do any of you have access to a Data Proof scanner?

Good luck. I am sure more will want to join you by your next meeting.
BTW, I thought meetings in Germany included beer. I see no evidence of such.

2. The maximum I think of is an intercomparison within Germany where I can travel to.
3. Well, the 2182A needs some calibration first, but yes the plan is to use it in further measurements. Cables are already at hand for it.
4. No, I don't have a Data Proof scanner. I once had a Prema 2080, but sold it by stupidity once I got my Prema 5017 SC, that was before we bought the K3458A.

No drinks in the lab which is full of expensive gear and electronics. Since Andreas had to drive, beer was prohibited But we had coffee instead.

-branadic-

[TiN]

Sorry, those cables and laptops next to sensitive equipment bothers me enough to leave this post here.. 

[branadic]

Sorry TiN, but not everyone can wastefully spend money like you do. This is fine for you, but please don't complain that not everybody is in your situation. Somewhat more social competences would even be good for you and is more worth then money

-branadic-

[GregDunn]

Beer + calibrated references + expensive DMMs = the European equivalent of "Hey, y'all, watch this"?