Ultra Precision Reference LTZ1000

[MiDi]

Wow, didn't expect such a response. I think I will order at Elecrow and pcb cost will be in the order of $2 per board or less depending on the final quantity plus shipping cost (Airmail) plus shipping to you guys (simple envelope to keep cost down). Thus I changed the outer form to a rectangle of size 120 x 50mm². PCB color will be black with immersion gold, single layer board.
Attached is the final schematic.
Please send me an official email adress with the PN too. Thanks.

Anyone interested in group buy of precision resistor set?

Edit:

LTZ1000 Precision resistors group buy meeting point

[Dr. Frank]

I am very honored by all these experiments!

[Hermann W]

@Branadic: There is a little mistake in the circuit diagram. It is not the regulator LT1763-3.3, but LT1763 without -3.3

[branadic]

Sure it's the adjustable version, but I used the library of the fixed version. Pinout matches.

-branadic-

[pansku]

@branadic Pm sent 

[branadic]

The call ends today, last minute orders are welcome:

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg1840139/#msg1840139

-branadic-

[AlfBaz]

PM sent, hopefully I didn't miss the boat

[Vgkid]

Pm sent.

[branadic]

The call is now closed. I ordered 150 boards. So this is a total of $139.50 + shipping $36.97 + import turnover tax to Germany + shipping cost to each of you. Inside EU this is about 1.45€ and international probably 3.70€.

Can't say anything about turnover taxes by now, but each board is about $1.17 + taxes + shipping, I think that is fine for all of you?
Production time 4-7 days plus 3-7 business days for shipping to me.

I still need some email addresses before I can start to write all of you for detailed information on how to pay me (invoice or PayPal).

-branadic-

[Hermann W]

@branadic: thank you for the personal Email.
Looking at the 10V-divider I suppose it is not correct. R12=6k and R13=2k333.. So UR13 is 2,8V (not 2,688V ) and there is no space for R14 and R15. R13 has to be a little bit smaller.

[Macbeth]

@branadic: thank you for the personal Email.
Looking at the 10V-divider I suppose it is not correct. R12=6k and R13=2k333.. So UR13 is 2,8V (not 2,688V ) and there is no space for R14 and R15. R13 has to be a little bit smaller.

Hermann, you misunderstand. I was scratching my head and then realised you are working out the resistors in parallel. 

This is not the case - it's just the PCB footprints giving you the option of axial wirewound or metal foil resistors. One or the other. Also the R12 vs R12A placement looks like it requires a jumper depending on what type of resistor you go with. It has to be said branadics placement of the component labels on the PCB leave a lot to be desired  But reading the schematic and layout it can be worked out.

I'm not sure why there is a difference between the values of wirewound vs metal foils - I guess I have to find Dr. Franks original reasoning?

[branadic]

It has to be said branadics placement of the component labels on the PCB leave a lot to be desired 

It has to be said, that I got the Eagle files from Dr. Frank, added the voltage regulator and slightly changed the position of some components to make the board fit into the Hammond case. I didn't change components label or values, they where given as it. 

-branadic-

[Macbeth]

It has to be said, that I got the Eagle files from Dr. Frank, added the voltage regulator and slightly changed the position of some components to make the board fit into the Hammond case. I didn't change components label or values, they where given as it. 

Yes, but the .pdf board layout you provided in your email does not include the (human readable) silkscreen, it does contain all kinds of meta-data labels that are all over the place and overlapping and should not be part of the gerbers (I hope).

Just look at Echo88's Dr. Frank board a few posts above - proper silkscreen.

At the end of the day it doesn't matter providing the board is electrically sound. I'm not complaining, it's a first spin that we all jumped on afterall 

ETA: I see the fancy PCB is actually Cellularmitosis version of Dr. Franks. So he perhaps tidied it up and added the silkscreen? I do apologise!

[Hermann W]

I was scratching my head and then realised you are working out the resistors in parallel.

Ok, I take the circuit diagram how it was. Now i had to laugh too 

[mweymarn]

All,

fyi - Digikey appears to have the LTZ1000A back in stock. At the moment they show 288 pieces available...

Regards,
Martin

[MiDi]

Special offer for pww resistors Ultrohm Plus from Edwin till 2018-10-21

[branadic]

Order has been updated to: In production. That took quite long.

-branadic-

[BradC]

Quick question.
So I re-read this entire thread (and a few others) looking for answers to these questions and can't seem to find anything other than vague subjective handwaving. So I'll ask a couple of questions and hopefully not trigger WW3.

LTZ1000 vs LTZ1000A. I'm under the impression that the standard LTZ1000 is actually likely to age / drift slower if it's run at a low enough temperature (ie with a 12.5k/1k divider). Are there really any real-world advantages to the A model other than power consumption (says he who has just taken delivery of some A units)?

Chopper amps as buffers. There are a number of posts dotted through several threads indicating that if you follow an LTZ1000 with a chopper it might be advantageous to stick an RC network in between to insulate the zener against the chopper input current pulses. My gut reaction is that sounds entirely reasonable and I'm struggling to see a downside. Has anyone done it? What sort of magnitude RC network might one look at if someone was mad enough to hack one into an existing design?

[chekhov]

Chopper amps as buffers. There are a number of posts dotted through several threads indicating that if you follow an LTZ1000 with a chopper it might be advantageous to stick an RC network in between to insulate the zener against the chopper input current pulses. My gut reaction is that sounds entirely reasonable and I'm struggling to see a downside. Has anyone done it? What sort of magnitude RC network might one look at if someone was mad enough to hack one into an existing design?

Not an answer, but a quick hint - I've seen this thing in TiN's article - 'Project 792X - 10V "FX" reference'(https://xdevs.com/article/792x/), look for "RC network R7C1 and R11C3"
There is no deep observations regarding particular component values, put probably it was mentioned somewhere else or deduced from opamp specs.

[Magnificent Bastard]

Quick question.
So I re-read this entire thread (and a few others) looking for answers to these questions and can't seem to find anything other than vague subjective handwaving. So I'll ask a couple of questions and hopefully not trigger WW3.

LTZ1000 vs LTZ1000A. I'm under the impression that the standard LTZ1000 is actually likely to age / drift slower if it's run at a low enough temperature (ie with a 12.5k/1k divider). Are there really any real-world advantages to the A model other than power consumption (says he who has just taken delivery of some A units)?

Chopper amps as buffers. There are a number of posts dotted through several threads indicating that if you follow an LTZ1000 with a chopper it might be advantageous to stick an RC network in between to insulate the zener against the chopper input current pulses. My gut reaction is that sounds entirely reasonable and I'm struggling to see a downside. Has anyone done it? What sort of magnitude RC network might one look at if someone was mad enough to hack one into an existing design?

• WWIII has already begun-- it is just not nuclear (yet).
• The 'A' version is the same die, but with a die attach method that is thermally insulating.  This helps a lot if you are running your reference on battery power-- it will save around 20% energy vs. the non-'A' part if run at the same die temperature.  LT (now ADI) applications engineers have said that the 'A' version has less hysteresis through power cycles.  Since the non-'A' version can be run at 10^oC lower die temperature, the 20% energy savings is dubious.  The leads on the 'A' version will be cooler-- so that is a big plus in managing thermal EMFs between the Kovar leads and the copper board traces-- (the idea is that you want to keep all of these junctions balanced and at the same temperature).  Because of the die-attach method used in the 'A' version, it is advisable to have a high-temp (125^oC) burn-in period where the entire burn-in period is 15-minutes ON and 15-minutes OFF.  This exercises the die-attach so that spurious jumps are suppressed.  This burn-in process must be followed by a conditioning process (like the Pickering Patent).  Because it uses a (more or less) ordinary die attach method, the non-'A' version does not require this special burn-in procedure-- but it still needs a 90-day (or so) burn-in (always ON) to 'settle' before calibration.
• All you have to do is study the physics behind a lead-acid battery "desulphator" to understand that current pulses can indeed affect crystal lattices.  It's a real stretch, but I suppose it is at least possible that the current pulses on the inputs of a chopper amp can affect the LTZ (if over a year or so, and the effect might only be sub-ppm).  Assuming that this is true, a simple RC on each input will stop all of that (presumably).  This would be a non-microphonic cap to ground on the (+) input (with a resistor between the collector of Q1 and the (+) input), and a non-microphonic cap across the (-) input and the output of the amp-- with a resistor from the (-) input to the 120R resistor.  This way, any current pulses from the inputs of the chopper (due to charge injection from it's internal switching action) will be absorbed by the capacitors.  If you want to (unnecessarily) use a chopper amp for the LTZ current/heater control, the OPA187 looks good (with the above caveats).  A chopper for the 7V->10V boost circuit (or for a 1:1 buffer) is mandatory-- an OPA189 looks good or an LTC2057 (which has better input common-mode headroom).  I don't like resistor based boost circuits because you are at the mercy of the ratio drift of the resistors.  Some kind of filtered PWM technique seems much better IMHO-- but it would still need to be chopper based to eliminate amplifier drift from the equation.

[Dr. Frank]

• The 'A' version is the same die, but with a die attach method that is thermally insulating.  This helps a lot if you are running your reference on battery power-- it will save around 20% energy vs. the non-'A' part if run at the same die temperature.  LT (now ADI) applications engineers have said that the 'A' version has less hysteresis through power cycles.  Since the non-'A' version can be run at 10^oC lower die temperature, the 20% energy savings is dubious.  The leads on the 'A' version will be cooler-- so that is a big plus in managing thermal EMFs between the Kovar leads and the copper board traces-- (the idea is that you want to keep all of these junctions balanced and at the same temperature).  Because of the die-attach method used in the 'A' version, it is advisable to have a high-temp (125^oC) burn-in period where the entire burn-in period is 15-minutes ON and 15-minutes OFF.  This exercises the die-attach so that spurious jumps are suppressed.  This burn-in process must be followed by a conditioning process (like the Pickering Patent).  Because it uses a (more or less) ordinary die attach method, the non-'A' version does not require this special burn-in procedure-- but it still needs a 90-day (or so) burn-in (always ON) to 'settle' before calibration.

All my LTZ1000 run at about -measured- 50...54°C (12k/1k), and each circuit consumes in total about 25mA.
The PCB and the device are  inside an insulated box, so that assembly heats up to about 10°C above ambient.
At 22°C R.T., that makes about 20°C difference to heat up for the LTZ1000. From the datasheet diagram, that's about 100mW or 20mA heating power @ 240 Ohm heater resistance, which gives exactly these 25mA, when including OpAmps and reference supply.
I did not -yet- check for LTZ1000A, but the same calculation yields: 12.5k/1k => 62°C, about 30°C difference (due to higher thermal insulation) => 80mW, 18mA heater power, which would give 2mA less current consumption only (23mA in total).  Maybe somebody else can confirm that value for the LTZ1000A.
The only advantage of the A version would be a warm-up time of <1sec, compared to 20..30sec for the non-A version.

I still do not recommend such a 'burn-in', as it will only introduce big hysteresis, as measured and proven by Pickering, hp in their drift - AN18A, and by my own measurements.
Even if one applies the Pickering method, it's really difficult to get to the virgin state, if ever possible.
I would estimate, that the theoretical relaxation process by burn-in is minor compared to the hysteresis effect, so better never excessively heat the LTZ1000 / A device, neither by soldering (no short leads either), nor by 'burn-in'.

Anyhow, nobody knows -up to now- how LT tests the finalized LTZ1000/A devices.
As it's specified for -55/+125°C operating temperature, maybe they test it 100% for both temperatures at the final tester, although the datasheet implies testing parameters at 25°C only.
That might leave the finished device in an unknown hysteretic state, and some devices will stay there indefinitely, others may drift towards their virgin state.

So the initial drift of several ppm/ x months may arise from such a final test over temperature, but not being the 'real' annual drift.

I have encountered initial drifts between zero and -2.5ppm / yr. for all of my seven LTZ1000 / 50°C references, either after assembly, or after erratically over-heating and re-conditioning.
After that time, all seem to approach the typical -0.8ppm/year, or less.

So I again recommend to just leave them alone, as burn-in may do more harm than will really / practically / measurably improve the long-term drift.

Frank

[Andreas]

Hello,

All my LTZ1000 run at about -measured- 50...54°C (12k/1k), and each circuit consumes in total about 25mA.
The PCB and the device are  inside an insulated box, so that assembly heats up to about 10°C above ambient.
At 22°C R.T., that makes about 20°C difference to heat up for the LTZ1000. From the datasheet diagram, that's about 100mW or 20mA heating power @ 240 Ohm heater resistance, which gives exactly these 25mA, when including OpAmps and reference supply.
I did not -yet- check for LTZ1000A, but the same calculation yields: 12.5k/1k => 62°C, about 30°C difference (due to higher thermal insulation) => 80mW, 18mA heater power, which would give 2mA less current consumption only (23mA in total).  Maybe somebody else can confirm that value for the LTZ1000A.

the current consumption of my LTC2057 buffered devices is listed here:
(20-24 mA including a LT1763 voltage regulator which should be well below 1 mA)

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg874560/#msg874560

For ageing: I think it depends more on the individual device or datecode than on the "A" or non A version.
But since we do not have enough samples and a JJA at hand we will never know.

with best regards

Andreas

[branadic]

...as measured and proven by Pickering, hp in their drift - AN18A,

Would you please link the publications/pdfs you are refering to? Thanks. 

-branadic-

[MiDi]

Anyhow, nobody knows -up to know- how LT tests the finalized LTZ1000/A devices.
As it's specified for -55/+125°C operating temperature, maybe they test it 100% for both temperatures at the final tester, although the datasheet implies testing parameters at 25°C only.

Note 2: All testing is done at 25°C. Pulse testing is used for LTZ1000A to
minimize temperature rise during testing. LTZ1000 and LTZ1000A devices
are QA tested at –55°C and 125°C.

Whatever QA testing means 

[Andreas]

Whatever QA testing means 

Usually: you take from each lot a number of samples and test them.