Ultra Precision Reference LTZ1000

[montemcguire]

I have to admit that I am using X7R capacitors to get them really close to the pins.
(one is 0603 in the middle of the LTZ-pins the other 0805)
But of course film capacitors (e.g. SMD ECHU type) are also possible.

with best regards

Andreas

You can now get 10nF C0G capacitors in an 0603 case. Many are 25V parts, but TDK has a 100V part in that size as well. So, you don't have to suffer the vagaries of X7R, unless you really want the 'voltage squish' and real component losses of X7R. Time marches on, and parts keep getting smaller!

Best Regards

[Insatman]

You can now get 10nF C0G capacitors in an 0603 case. Many are 25V parts, but TDK has a 100V part in that size as well. So, you don't have to suffer the vagaries of X7R, unless you really want the 'voltage squish' and real component losses of X7R. Time marches on, and parts keep getting smaller!

Best Regards

Jeez...1206 parts are hard enough to solder...0603 is a nightmare.

[Kleinstein]

Changing L3 to a resistor would ruin the low output impedance. The trick when makting the OPs output more tolerant to capacitive drive it to have the DC feedback from the output (thus behind L3 or resistor) and the fast feedback via a capacitor directly from the OP.  Just the extra cap and L3 instead of the usual resistor could also work reasonable.

Some here may not like the one rather thick line going to the LTZ1000, because of possible thermal gradients. I am not sure it matters.

[MisterDiodes]

Build the circuit how you need to, but keep in mind LT/AD does not recommend adding those extra EMI caps, and in our tests normally they are not required, or desired - Especially SMT ceramics always run the risk of adding mechanical noise.  The solution is to look at what your EMI source really is, and see if that can't be corrected.

I suggest you might want to test your circuit in your enclosure FIRST to see if they are actually needed at your location, with your test setup.  Add the extra caps -only- as required.  If your circuit operates normally without them, you're good.

The noise spec you're looking for as defined by LT is if you look at many 10 second periods of your output, you should see an -average- PEAK noise of < 1.2uV averaged across all periods of observation.   It is normal to see an occasional peak noise pulse of perhaps over 2uV on occasion, but that is considered perfectly normal and well within bounds.

As Kleinstein noted, normally all traces to the LTZ should have somewhat equal thermal impedance, otherwise you may have created a thermal EMF at the device leads even with the cap over the LTZ.  If the thermal EMF is stable you might be OK but the uneven traces might cause problems if any airflow in that area.

[Insatman]

Changing L3 to a resistor would ruin the low output impedance. The trick when makting the OPs output more tolerant to capacitive drive it to have the DC feedback from the output (thus behind L3 or resistor) and the fast feedback via a capacitor directly from the OP.  Just the extra cap and L3 instead of the usual resistor could also work reasonable.

Some here may not like the one rather thick line going to the LTZ1000, because of possible thermal gradients. I am not sure it matters.

I modified the layout slightly to make all the traces the same width...but not the same length.  Note that I intend to cover the LTZ1000 with a small cube of hollowed out styrofoam which also cover about 1/4" of the PCB around the device.  the bottom of the board will also have about 1/8" of foam in that same area.   This both prevents air currents from reaching the device and lowers the input power required by the heater to maintain temperature.  Furthermore, the finished board will be enclosed in a cast aluminum enclosure lined with thin foam on the inside.  So no air currents are expected.

[Insatman]

Build the circuit how you need to, but keep in mind LT/AD does not recommend adding those extra EMI caps, and in our tests normally they are not required, or desired - Especially SMT ceramics always run the risk of adding mechanical noise.  The solution is to look at what your EMI source really is, and see if that can't be corrected.

I suggest you might want to test your circuit in your enclosure FIRST to see if they are actually needed at your location, with your test setup.  Add the extra caps -only- as required.  If your circuit operates normally without them, you're good.

The noise spec you're looking for as defined by LT is if you look at many 10 second periods of your output, you should see an -average- PEAK noise of < 1.2uV averaged across all periods of observation.   It is normal to see an occasional peak noise pulse of perhaps over 2uV on occasion, but that is considered perfectly normal and well within bounds.

As Kleinstein noted, normally all traces to the LTZ should have somewhat equal thermal impedance, otherwise you may have created a thermal EMF at the device leads even with the cap over the LTZ.  If the thermal EMF is stable you might be OK but the uneven traces might cause problems if any airflow in that area.

I added those caps due to Andreas recommendation...someone I have come to respect in this area of expertise.   In parallel with this effort I am also building a 10K:1 amplifier to measure the noise you mention in your specification. see https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg1471764/#msg1471764. 

I can add the EMI caps if necessary.   Attached are the latest schematics and PCB layouts.   Note I also added some additional SMD caps on the input and output connectors...why not...I don't have to use them.

[chuckb]

I believe the latest guidance for C2 is 22nF. The Linear Tech datasheet calls out the incorrect value for this capacitor on the positive 7V regulator schematic. The capacitor value on the schematic for the negative 7V regulator is correct.

For safety during development, I like to have a 5V zener in series with the collector of Q1 to limit the maximum heater voltage to 10V. This can prevent permanent damage to the LTZ1000A chip if there is a problem during the development process. The Zener is not needed if you use the non A version of the chip.

Good luck!

[Insatman]

I believe the latest guidance for C2 is 22nF. The Linear Tech datasheet calls out the incorrect value for this capacitor on the positive 7V regulator schematic. The capacitor value on the schematic for the negative 7V regulator is correct.

For safety during development, I like to have a 5V zener in series with the collector of Q1 to limit the maximum heater voltage to 10V. This can prevent permanent damage to the LTZ1000A chip if there is a problem during the development process. The Zener is not needed if you use the non A version of the chip.

Good luck!

Thanks for the update on C2.  My PCB can accommodate that value easily.  I will update my schematic.

[Kleinstein]

The last diagram shown  an odd connection at the lower end of R10.


The ferrite at the output of the 10 V amplifier would add to DC output resistance. Normally feedback should be all the way from the output - could be down to the connector. In addition some local feedback via a capacitor might help a little with more tolerance to output capacitance. Even if an OP data-sheet says that the OP is stable with 1 nF or maybe 10 nF at the output, this only means it is not oscillating. Added capacitance usually still degrades the output and makes is more prone to ringing / oscillation. For an external output it would be a good idea to have to usual isolation with resistor to have a more capacitive tolerant output.

Those EMF caps Andreas suggested were part of more changes and just adding the caps, but without the other changes will reduce stability and likely make the circuit more sensitive to capacitive load.

[Insatman]

The last diagram shown  an odd connection at the lower end of R10.


The ferrite at the output of the 10 V amplifier would add to DC output resistance. Normally feedback should be all the way from the output - could be down to the connector. In addition some local feedback via a capacitor might help a little with more tolerance to output capacitance. Even if an OP data-sheet says that the OP is stable with 1 nF or maybe 10 nF at the output, this only means it is not oscillating. Added capacitance usually still degrades the output and makes is more prone to ringing / oscillation. For an external output it would be a good idea to have to usual isolation with resistor to have a more capacitive tolerant output.

Those EMF caps Andreas suggested were part of more changes and just adding the caps, but without the other changes will reduce stability and likely make the circuit more sensitive to capacitive load.

You caught an error in the schematic with R10...thanks.

As for the ferrites I use they have a thick wire going through them.  The resistance is very very low, probably less than a PCB trace of equal length.

As for the caps on the op-amp output, there is a lot of debate around this.   I have always used 10nF on my voltage references.   In this thread it has been suggested to add them using a series resistor for op-amp stability.  I am reluctant to add resistance.  So...I decided to put the pads in the PCB and decide down the road to add the capacitance or not based on experience. 

[Edwin G. Pettis]

Insatman, I strongly suggest you look at the Vref circuits in HP (Keysight), et al, Linear Tech originated the LTZ1000 over 30 years ago and their published schematic has stood the test of time (sans typos).  A bunch of bypass capacitors are neither necessary or wise, if your circuit needs such band aids, it is because you are doing something wrong or trying to compensate for external EMI and it is a well know fact that you always clean up noise at the source first before trying to do it in your circuitry.

MisterDiodes manufactures LTZ based references for industrial use in environments none of you are likely to encounter, he knows what he is talking about and I concur completely with his advice.  You are certainly welcome to ignore it but be advised it is not the best course to pursue.  The 3458A has utilized the LTZ circuit since conception without all the excessive capacitors and it certainly appears to be working just fine.  I suggest you implement the original circuit as is, if you find that you have have EMI problems then take the proper steps to eliminate the interference externally before modifying the LTZ circuit.

The first good step is using a steel enclosure not aluminum.  Keep noisy equipment such as laptops as far away as possible, that includes any circuitry with clocks in them and their associated cabling (such as ADCs or DACs), LED lighting or CCFL lighting (or standard fluorescent).  Incandescent or Halogen work just fine here.  This advice is all time proven and verified by many independent sources over the years.

If you have any questions about the operation of the LTZ circuit, the best place to go for advice is the engineers at Linear Technology, they will be most happy to answer them, for some reason, that course of action does not seem to be very popular on this forum.

[MisterDiodes]

Yes, I've built a couple   

Insatman:  Build your box however you want, just understand -  What happens by adding an "EMI" cap directly across an analog PN junction on a precision substrate, is you can potentially develop a situation where you convert a small voltage spike into what can be a larger current spike on the LTZ (when you count various inductive effects in the leads, etc) - and you don't want to "ring the bell" on your LTZ chip, at least as little as possible.

You're output might look a little cleaner but you may not like effects on the substrate crystal lattice over time.  This can show up as an affected long term drift rate, which may or may not be noticeable if you build just a few Vrefs.

The LTZ zener will add it's own noise of course but we want to try not to add any -more- current noise.

As Edwin pointed out:  There is absolutely nothing inside the LTZ circuit that should be creating much noise, except for the LTZ Zener itself.  Especially when you use PWW resistors.

Best engineering practice is to mitigate EMI noise elsewhere or add shielding if possible - that's only why I suggested you try the circuit as per LTZ datasheet first.  It _should_ be fine.  You can always add an extra spike filter farther downstream, away from the LTZ directly.

Or just use the "EMI caps" if you find there is no other way for your application if your test setup is just too noisy.  Andreas was using the extra "EMI caps" because he said that's where he got best results for his test setup - your test setup might be different and not require them.

As I just posted on the "shielded chamber" thread, it's handy to have some of these around to test with your Vref (like you see in every good low-noise app note):

https://bascousa.com/1-2-gallon-metal-paint-can.html

Or even standard drawn steel electrical boxes come in handy too.

Aluminum boxes won't do much shielding at low freq H-field.





 

[Andreas]

Yes, I've built a couple   

If I would be evil: I would say "glad to be not your customer". But since I'm not, I have never said that.

Of course there is one strategy to ignore EMI. And live like 30 years ago without high speed internet, cable TV, mobiles, WIFI and do not forget to ban the keyless entry FOB for your car. And also prevent your neighbours from using such "devils things" if you are living in a urban region. So today we have the complete frequency range from about 150kHz to 5GHz.

Insatman, I strongly suggest you look at the Vref circuits in HP (Keysight), et al, Linear Tech

By the way the EMI capacitors are adopted from the DATRON reference cirquit. So no invention from me.
https://doc.xdevs.com/doc/Datron/4910_4911/4910%20Reference%20Assembly%20c20090208%20%5B1%5D.pdf

Of course a metal housing helps against direct radiation. And a tinnned steel housing somewhat against magnetic fields.
(thats why I normally use a double shielded strategy).
But are you really using a transformer with shield between primary and secondary winding
like better instruments (3458A) do with their shield + guard stragegy?
And your paint can does only help if there is no antenna (ground strap or signal strap) looking out of the CAN.

The easiest way to look if you are affected by EMI is to measure the heater voltage on your oscilloscope (with AC-coupling in a sensitive range). If you have randomly (typically negative large (up to several 10mV)) spikes on the heater output besides the normal regulation noise then its time to improve EMI measures. Of course you can also see the spikes on the VREF output, but with much smaller amplitude so you will need a appropriate LNA for the scope.
So it is always good to have a option besides shielding.

with best regards

Andreas

[Edwin G. Pettis]

"If I would be evil: I would say "glad to be not your customer". But since I'm not, I have never said that."  Really Andreas?  You know absolutely nothing about his designs and the fact that he sells to demanding industrial customers says a lot about it.  They are not sitting in a comfortable calibration lab, they are in industrial environments, anything but comfortable.  If you were his customer (which you probably couldn't afford) at least you would have a high quality reference in your hands.  Think about that.

I'm curious, why did you think we were talking about you in particular, we made no mention of anybody?

[MisterDiodes]

Yes, I've built a couple   

If I would be evil: I would say "glad to be not your customer". But since I'm not, I have never said that.

Of course there is one strategy to ignore EMI. And live like 30 years ago without high speed internet, cable TV, mobiles, WIFI and do not forget to ban the keyless entry FOB for your car. And also prevent your neighbours from using such "devils things" if you are living in a urban region. So today we have the complete frequency range from about 150kHz to 5GHz.

Andreas

<Sigh> I never knew this thread would turn unpleasant, and I can't remember why spending time to post seemed to be a good idea.  I have no concern or argument if anyone adds "EMI" caps or not...If it makes you happy, knock yourself out!  I just thought it would be helpful to know why you might want to check first to see if it's really required or not before adding extra parts.  That was my intention, and maybe that wasn't clear enough.

Datron?  Maybe if they had stuck to the datasheet, they might still be around! (Kidding)

5GHz?  That's all?? - Andreas, sometimes in on the production line your device will be in a very quiet place, no clocks over 10 or 100kHz allowed.  Other times your device will be sitting next to a 100Ghz- capable test jig with 6 high current servo drives and about 25 various solenoids, pressure regulators, etc. An EMI nightmare, but the standard circuit LTZ's shields are holding against the onslaught (without any extra parts) and the Vref is steady, test data is all good.   And one day the customer will call and complain he's got 1Hz noise on his test data, and they swore it seemed like it was the LTZ.   Do we start adding caps?  No, we first track down the problem...it turns out it's high freq all right, an 80GHz signal being cycled on and off at 1Hz on the test jig running near a measure signal.  The real solution was to re-route one of the measurement cables to the back side of a steel support leg (the tech had taken apart the machine and put the cables back in the wrong position) - end of problem.  Sometimes it's that simple, sometimes not.

If we couldn't have found the problem and the only solution was to alter the LTZ circuit then of course it's a good idea to do whatever needs to be done.  If the customer needs a customized solution and we need to alter the basic LTZ circuit for some reason, then of course we would do that - whatever it takes.  We have to keep the customer happy, because if they aren't - well we don't get paid.  That's a pretty tight feedback loop to let us know how we're doing.

Do we really use screened and shielded transformers?  Yup!  Its a great idea!  We have them made at one of the same facilities that makes transformers for Keysight, and a lot of other high end test equipment... It's amazing what is done on transformers for equipment like a 3458a that is never shown on the schematic.  You're right - we -are- bombarded by EMI more and more every day, and keeping the noise out of the precision systems is very important.

You can do the same "quiet" transformer power supply techniques at home by building your own or looking at "Medical" screened transformers (usually toroid).  These will typically be a single screen but really help reduce capacitive coupling between windings on a toroid.  For hobby use even the low cost Signal transformer standard Split Bobbin E-Frame types are really pretty good at power line noise reduction - they won't have screens but the split bobbin keeps interwinding capacitance pretty low (because the windings aren't on top of each other like on a toroid), and you can add your own steel box cover as needed (or as required).  Really not a bad solution and not too expensive either.

Peace out!

[zhtoor]

@MrDiodes

Dear Sir,

it has been a privilege to be getting advice from an experienced person like you,
and i don't think that i can thank you enough for that.

i am most grateful and best regards.

-zia

[MisterDiodes]

...Now I remembered why I posted... Thanks and same to you.

And really - the world is bad enough right now.  EEvblog should be about camaraderie and good, fun discussion - please let's just keep it at that level.

If you disagree with me that's fine, but I'll look forward to common ground another day.

[MisterDiodes]

...and I was hoping someone would get the "common ground" joke [rimshot]

I'll just show myself out now

[Cerebus]

...and I was hoping someone would get the "common ground" joke [rimshot]

I'll just show myself out now

Where did you bootstrap that 'joke' from?

Here's your coat, I've already got mine...

[Cerebus]

If you have any questions about the operation of the LTZ circuit, the best place to go for advice is the engineers at Linear Technology, they will be most happy to answer them, for some reason, that course of action does not seem to be very popular on this forum.

That's probably because the folks here are either amateurs, who don't have a relationship with Linear Technology Analog Devices, therefore probably don't have the first clue how to get in touch with one of their application engineers, or professionals working on hobby projects who feel some reluctance to exploit 'professional' resources for a private project.

[kj7e]

This so describes everyone here;

    You Might Be an Engineer if...

    you see a good design and still have to change it.
    you still own a slide rule and you know how to use it.
    your family haven't the foggiest idea what you do at work.
    in college you thought Spring Break was metal fatigue failure.
    you have modified your can-opener to be microprocessor driven.
    you are better with a Karnaugh map than you are with a street map.
    you think "cuddling" is simply an unproductive application of heat exchange
    you have owned a calculator with no equal key and know what RPN stands for.
    you make four sets of drawings (with seven revisions) before making a bird bath.
    you have trouble writing anything unless the paper has horizontal and vertical lines.
    your ideal evening consists of fast-forwarding through the latest sci-fi movie looking for technical inaccuracies.
    you think the value of a book is directly proportionate to the amount of tables, charts and graphs it contains.

Thanks to all who have offered their insight and experience, for one, its helped me out a bunch.  My LTZ1000 project is done and now off to others for testing and comparison.

[Insatman]

And really - the world is bad enough right now.  EEvblog should be about camaraderie and good, fun discussion - please let's just keep it at that level.

If you disagree with me that's fine, but I'll look forward to common ground another day.

Thanks for the advice and I agree with you statement concerning what EEvBlog should be about.   All too often on the internet people have thin skins and good discussion turns into something unproductive.

A step back...My goal in this design is to make a PCB with lots of options, giving me the best chance for an improved LTZ1000 reference.   My present reference does not seem to perform as well as I'd like.  It was built largely to the original published Linear Tech circuit complete with typos.   There are some differences like the 10nF on the scaling op-amp output.   see https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg1403374/#msg1403374 for the original post describing my first effort.

My reference will be battery powered using the hardware described in the above referenced post.  Nothing digital is in this box, unless you want to count what I describe next as digital.   The only large noise source is my poor choice of switch mode battery chargers which can be turned off prior to making measurements...although I tend to forget to do so when logging the voltages daily.   These switching based chargers need replacement or modification....they are the largest source of noise in my little E-lab....something to add to my already long list of things to do. 

I do have extensive experience with EMI noise, shielding and suppression.  In my career, I was on a team that designed and built several EMP and lighting generators for a variety of customers, both governments and private industry.   Imagine designing control and instrumentation electronics to be housed on an EMP generator?   

By using the EEvBlog "hive mind" I was trying to get everyone's input on the LTZ1000 circuit.  This way I could design a PCB that would ultimately yield a good design after some experimentation.  In that process, I would learn what works and what does't.  After all, isn't that one of the prime motivations for this hobby...learning?    At this point I think the present PCB design is versatile enough to serve my purpose.

This forum is a valuable source of information and camaraderie for me in my retirement years.  This hobby helps me keep my mind from rotting away and allows me to explore areas of electronics I wasn't able to pursue earlier in life.  An example is the LTZ1000 circuit.   My previously built 3x LM199 voltage reference is more than good enough really for my lab....but...the LTZ1000 is supposed to be better and I want to build one that actually is as good as I can make it.   

 

[cellularmitosis]

I fear this may have simply been a joke which was lost in translation, or a simple difference of opinion with no ill-will attached.  Andreas, MisterDiodes, Edwin, Kleinstein Dr. Frank, TiN, and many others have been dedicated contributors to this forum, with data, designs, feedback, and advice.  I am very grateful to you all.   

[TiN]

I never knew this thread would turn unpleasant, and I can't remember why spending time to post seemed to be a good idea.  I have no concern or argument if anyone adds "EMI" caps or not...If it makes you happy, knock yourself out!  I just thought it would be helpful to know why you might want to check first to see if it's really required or not before adding extra parts.  That was my intention, and maybe that wasn't clear enough.

Don't let it get under the skin, your input is valued here. Nowadays more people seem to expect everything presented for them on the plate, with theory and instructions and explanations that even 5 year old can read. And when you give just a hint you get bashed for not breaking your NDAs and business agreements, with puzzled faces. After all can't please all, there will be always unhappy fella (too expensive, too difficult, too hard to buy parts, no schematics posted, no test result posted, whatever..) .

On my refs (KX with chopper amp and SMT parts, FX with traditional 1013 and THT parts) I saw no positive effects of any additional caps (when modified one of the boards, including additional resistors as Dr.Frank was quite convincing). So as result I did not implement them, keeping design simpler, by the books. I saw no spikes or jumps when good LTZ chip used and proper measurement setup, power source and wiring are used. And my homelab is not some EMI-shielded faraday cage, just a standard apartment with LED lights, monitors, desktop confuser w/o case and bunch of other crap turned on.

Using extra caps or other design changes because of Datron or another company used it before, without rationale why should not be a reason for design engineer to do same in own build. Datron refs even have voodoo slots and VPG networks too, but as we already know, plain 3458A ref works just fine without any of that, providing tempco and stability better than majority of the LTZ builds over these 90 pages.

And approach not to generate EMI instead of patching circuits around to suppress it is the industry standard for decades already. Take any RF equipment, and ask a question, why so many expensive custom made shields around everything, especially around active circuits? Heck, even standard consumer stuff like PC VGA cards, motherboards and peripherals, being as low cost as possible, still usually have ground planes on the edge with via stitching and shielded connectors to reduce emitted EM. Otherwise lot of that stuff would not pass certification and might have trouble selling . 

[Andreas]

Hello,

@MisterDiodes:
every time I recommend the EMI-capacitors you shoot agaist them.
Telling the fairy-tale that they will do harm to the poor LTZ (like choppers et al).
Creating unnecessary FUD for the new-comers.
As a professional I wouldn´d do that with any official prove. (which is still missing).
If it should have been a joke then it is hard to understand for a non native speaker.

@TIN:
If you have no problems in your environment then of course you are free not to implement any measures.
Perhaps the multi-layer design implements some parasitic (small) capacitors
which are sufficient for your environment. (would be worth to examine).

Others here in the forum (also in the KX thread) have problems
"with some small unexplained glitches"
https://www.eevblog.com/forum/metrology/kx-reference/msg1213344/#msg1213344
which could be either a problem of the meter or the reference.

Others have "impact at the end comes from touching" [the DMM]
https://www.eevblog.com/forum/metrology/kx-reference/msg1298271/#msg1298271
Which is also still unclear.

Does it really help those members telling they should do "better shielding" (how?)
Ups forgot to mention they need a special transformer too...

Better shielding + transformer are several 10$.
EMI-capacitors are some cents.

just my 2ct

with best regards

Andreas