Author Topic: Ultra Precision Reference LTZ1000  (Read 958190 times)

0 Members and 2 Guests are viewing this topic.

Offline saturation

  • Super Contributor
  • ***
  • Posts: 4788
  • Country: us
  • Doveryai, no proveryai
    • NIST
Re: Ultra Precision Reference LTZ1000
« Reply #150 on: May 16, 2013, 11:10:41 am »
I wouldn't be surprised if your 732B actually is far less than 2ppm/year as its a much aged reference.  The real trick is measuring that because at least either the reference or your DMM will need calibration against a known standard in the calibration cycle to check the rate of drift of both.  From the factory the 732B is 0.8ppm in 3 months.  The real hard part in making a metrology setup is tracking the data over time to insure the stability is true.


It's not totally "easy" to make a voltage transfer standard that holds 2ppm per year-- but is can be done by the average engineer using average (and well known) approaches to the problem.  But-- (short of going out and buying a million-dollar Josephson Junction Array)-- to get below 2ppm per year takes a lot of study, hard work, patience, and help from your friends.

I already own a Fluke 732B, which (according to it's published specifications) will hold 2ppm (or less) per year.  So, I want something better.  I want *less* than 1ppm per year, and my goal is (somehow) to achieve 0.1ppm/year-- a lofty, but worthwhile goal (for me at least).  For most people, the Fluke 732B is "Good Enough"-- but not for me.
Best Wishes,

 Saturation
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 837
  • Country: de
  • pushing silicon at work
Re: Ultra Precision Reference LTZ1000
« Reply #151 on: May 16, 2013, 04:59:29 pm »
The principle of calibration (how does a reference or DMM agree with a better one?) and the quest to distribute the SI representation is clear to me.

My cheapo-no-magic reference aka "Vbs" ( Volt babysitter) was compared to a volt-nuts 4 other references (including a LTZ in a 3458 which was used for the 24h, 31h, 1 Week and noise experiments) that haven't been calibrated relative to a cal-lab for a while but their behaviour is closely compared to each other regularly. (If one of them wants to go for a ride voltagically speaking, you wouldn't realise if it is alone, but it will visibly differ from its siblings!)

My Vbs  behaves closely like that cluster, how it differs in the long term I will see, got it back yesterdays from the first experiments. Now the Vbs is on its own for say one or three months...

What I meant with my last question is that it is increasingly getting harder to tell who is the reason for a deviation, the noise and drift of the Vbs is no longer easy to discern from that of the 3458A.

Also, in October iirc I will get a 34401A at work back from the Böblingen Cal lab of Agilent. This has not the greatest long-term-stability, but immediately after arriving it will be reasonable close to what they measured against their traceable references. Knowing the instruments error as noted by the lab, taking some readings from the babysitter volt abusing the 34401As gratuitous extra digits coming out of GPIB will give a cloud of datapoints for you know what... the VAgilent vs. Vbs and later Vbs against Vvolt-nut and also Vbs vs VQuarks:blah:

But I have to tell, the main reason for the LTZ reference is the predecessor, a REF102-based 10V, did not comply with my minimum requirement: Measure it at work, stuff it in the backpack, take it home and back to work the next day, measure again, delta only at the last digit. This one does so, after travelling much more. Now I can trust it for testing the DMM at home.

Greetings,
Hendrik
I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline quarks

  • Frequent Contributor
  • **
  • Posts: 834
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #152 on: May 19, 2013, 05:36:15 pm »
...any capacitor in the signal path should be high quality metal-film, or NP0 [aka: C0G]...

So far I have planned to use WIMA MKS for all caps. But also I will try to get NPO caps. Is there a recommended brand/type to go for? 
 

Offline saturation

  • Super Contributor
  • ***
  • Posts: 4788
  • Country: us
  • Doveryai, no proveryai
    • NIST
Re: Ultra Precision Reference LTZ1000
« Reply #153 on: May 20, 2013, 07:57:19 pm »
Several years ago I explored doing a more robust reference but found the old Krohn Hite portable calibrators, like the MV106, incidentally also purchased by Dave about the same time.

Its rated for ~ 2ppm/yr, through a range of operating temps.  Since it was made over 30+ years ago, the references are well aged, and the unit robustly made.  But kept at room temperature over just short a year now, the accuracy is far better, <<= 1ppm.  You can find these on eBay for under $100, with some watchful waiting, or choose variations of that voltage references made by KH from that era; they have a similar look but differ in their dial-a-volt capacities.



« Last Edit: May 20, 2013, 07:59:22 pm by saturation »
Best Wishes,

 Saturation
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 837
  • Country: de
  • pushing silicon at work
Re: Ultra Precision Reference LTZ1000
« Reply #154 on: May 22, 2013, 04:07:30 am »
@Dilligentminds.com

Looks like a reference diode Kindergarten to me :) Nice to see such a project coming up even if it differs from the topic LTZ1000A...

Does it differ from the 4400, and if, by how much ?

In my opinion its too many resistors for trimming, especially the trimpot. If it is not something special to cancel out individual drifts and tolerances, I would go for one single hi-spec feedback resistor at the op and the resistor to gound would be a slightly-too-low hi-spec value in series with a few lower-specced low-resistance types calculated for slightly too much resistance of the series circuit, and short each of the low-value resistors on the PCB.  This way I can use a knife to "activate" the additional low-resistors one by one - and solder bridge if i activated one too much. 

@Whoever is making a own circuit:
Learn from my fail (no test point for unbuffered voltage) and make a test point for the raw voltage before the amplifier !
This way you can either measure dc:dc ratio or simply the ~2.8V difference (with LTZ, ~3.1V here)  between your trusted Vref and what the amplifier does.

I will either accept what my buffer does to my reference or do a opampekcomy (decide later after watching longer-term drift), may set up a external chopper amp circuit which will be used with a existing KVD to make other voltages.

Still, I am confident that nailing together the AN86 circuit with standard means without black magic (teflon or Al2O3 board, stress relief slots, ultralow temperature, super secret thermal cycling and aging, VishayPG VHx Resistors) is sufficiently close to optimum that you you have a hard time to prove the difference to references made with black magic by usual means like 8.5 digit multimeters. 

I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline grenert

  • Frequent Contributor
  • **
  • Posts: 446
Re: Ultra Precision Reference LTZ1000
« Reply #155 on: May 23, 2013, 01:29:54 pm »
taking some readings from the babysitter volt abusing the 34401As gratuitous extra digits coming out of GPIB
Sorry for the off-topic (but in the spirit of volt-nuttery in this thread):
Does the 34401A put out more than 6.5 digits via GPIB?  I know the 3457A can do this, but had not heard the 34401A does the same.
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1873
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #156 on: May 24, 2013, 04:34:33 pm »
taking some readings from the babysitter volt abusing the 34401As gratuitous extra digits coming out of GPIB
Sorry for the off-topic (but in the spirit of volt-nuttery in this thread):
Does the 34401A put out more than 6.5 digits via GPIB?  I know the 3457A can do this, but had not heard the 34401A does the same.

Hello,

yes, definitely, 7 1/2 digits over the bus in each mode, if I remember right..
If you use the internal statistics, additional digits are also available.
I did not check recently, how stable the additional digit is, but have in mind that it's quite useful (for the LTZ1000 and on 4WOhm measurements), at least when additional averaging is applied.

Update: Just've checked by programming:
The GPIB output format is: x.xxxxxxx0E+yy, i.e. 8 digits, the ninth is zero.
The 7th digit is relatively stable.
It's better to average on 8 digits than on 6,5, because the average converges quicker.

The Min/Max function already delivers stable 7,5 digits for the average also, which I have used for measuring with higher resolution without GPIB.

Frank
« Last Edit: May 24, 2013, 05:08:13 pm by Dr. Frank »
 

alm

  • Guest
Re: Ultra Precision Reference LTZ1000
« Reply #157 on: May 24, 2013, 04:40:18 pm »
You can get the extra resolution from averaging anyhow, so if it's down in the noise, you might as well average the 6.5 digit value.
 

alm

  • Guest
Re: Ultra Precision Reference LTZ1000
« Reply #158 on: May 24, 2013, 06:21:13 pm »
I believe the recent(ish) Keithley 200x meters give more digits over GPIB. No idea how significant they are.
 

Offline cyr

  • Frequent Contributor
  • **
  • Posts: 252
  • Country: se
Re: Ultra Precision Reference LTZ1000
« Reply #159 on: May 24, 2013, 10:39:35 pm »
The 2015 does, not sure what the actual quantization steps are but seems like an extra digit or so of usable resolution. The 2001 on the other hand gives the exact same value as the display, at least by default.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2545
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #160 on: May 27, 2013, 07:40:20 am »
Hello together,

On ordinary Zeners (1N829A) there exists a "zero tempco current" where the tempco is nearly 0 ppm/K.

On 2 samples of LTZ1000A with the standard 120 Ohms resistor, which gives around 4 mA zener current, I have measured a tempco of the zener element of around 50 ppm/K. So to come to the 0.05 ppm/K the heater stability has to be better than 1/1000 K.

Has anyone tried to find out the optimum current/resistor value for the zener of the LTZ1000(A) where the tempco gets zero?
That what I tried is to change the 70K Resistors. Lowering them to 50K will increase the tempco of the reference slightly.

With best regards

Andreas
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2094
Re: Ultra Precision Reference LTZ1000
« Reply #161 on: May 27, 2013, 08:32:04 pm »
On ordinary Zeners (1N829A) there exists a "zero tempco current" where the tempco is nearly 0 ppm/K.

It isn't an ordinary zener it is a zener + forward biased diode combination and the temperature coefficient of the diode forward voltage depends on the diode current. As far as I know zener avalanche breakdown voltage temperature coefficient isn't dependant on avalanche current.

The LTZ1000A nominally 7.2v 'zener' is the combination of the zener and transistor Q1 Vbe so is is the same and you can control Q1 emitter current by changing the collector resistor. If dropping from 70k to 50k made it worse I assume you were going the wrong way.

The LTZ1000 datasheet shows a circuit to trim the temperature coefficient when you don't bother to use the heater, you would think the same approach could also improve the temperature coefficient when you do.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2545
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #162 on: May 27, 2013, 09:31:24 pm »
Quote
I would also encourage you to use (2) LTC2057 op-amps to control your LTZ1000(A)-- these op-amps have almost zero drift over time, and they have much lower wide-band and DC-10Hz noise than the LT1013 used in the original applications note. 
Thats good news for me the LTC2057 is noted on my wish-list. Up to now there are only samples available in LTC web shop. But I think they will be soon available on DigiKey too.

Quote
I strongly suspect that the ~7V buried Zener in the LT1034 is the exact same one that is used in the LTZ1000(A),
I cannot believe this: If I compare the "voltage change over current" diagrams of LTZ1000 and LT1034 at 1.25 mA the change of LT1034 is much larger (100mV) against the LTZ1000 (20mV). The LT1034 is developped as "micropower reference".

Quote
The LTZ1000 datasheet shows a circuit to trim the temperature coefficient when you don't bother to use the heater, you would think the same approach could also improve the temperature coefficient when you do.

Hello Rufus,

my intention is to use the sweet spot by changing the 120 Ohms resistor to either enhance the stability or be able to use cheaper resistors for the temperature setpoint. I know the cirquit from the datasheet which makes only sense if you need a low noise source. The additional resistor adds its own tempco to the cirquit and increases the output voltage by up to 1V.

So is there anyone who tried changing the 120 Ohms resistor and checked tempco before and after?
Or does anyone shurely know the maximum current limit for the LTZ zener without degradation?

with best regards

Andreas
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2094
Re: Ultra Precision Reference LTZ1000
« Reply #163 on: May 27, 2013, 11:07:00 pm »
my intention is to use the sweet spot by changing the 120 Ohms resistor to either enhance the stability or be able to use cheaper resistors for the temperature setpoint. I know the cirquit from the datasheet which makes only sense if you need a low noise source. The additional resistor adds its own tempco to the cirquit and increases the output voltage by up to 1V.

In typical LTZ1000 circuits the zener current is very dependant on temperature (about -16uA/C) combined with the slope resistance of 20 to 60 ohms that gives a relatively huge -ve zener voltage temp co of around -640uv/C at 40 ohms. Yet that temp co + the avalanche temp co + Q1 Vbe temp co add up to almost nothing. That has to be by design. Who knows, they might even laser trim some deliberately added slope resistance to make them sum to nothing with 100uA Q1 emitter current.

So yes it seems adjusting the 120R will adjust the effect of the slope resistance temp co. Adjusting some added slope resistance (like the 200 ohm variable resistor in the datasheet circuit) will also adjust the slope resistance temp co. Adjusting Q1 collector resistor will adjust the Q1 Vbe temp co.

It is strange that the datasheet adjustment circuit seems to indicate you can null the overall temp co just by adding slope resistance without other component changes.

Which is best to adjust I don't know. I would imagine shipped parts are close to their sweet spot with 120R and 70k resistors, by design and/or trimming.

Won't there already be some 'voltnut' talk about stuff like this? I'm just looking at the circuit and making it up.

 
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2094
Re: Ultra Precision Reference LTZ1000
« Reply #164 on: May 28, 2013, 02:14:01 am »
Snip another wall of text with lots of knowledge and not enough understanding.

We are not trying to adjust an LTZ1000 Zener current to find a 'sweet spot' because there isn't one. Show me a reference that indicates zener avalanche voltage temperature coefficient depends on avalanche current.

We are trying to cancel the temperature coefficient of the zener avalanche voltage with the temperature coefficient of the Q1 transistor Vbe. Q1 Vbe temperature coefficient will drop by about 200uV/C per decade increase of collector current so changing the collector resistance is a valid way to adjust the temperature coefficient matching.

Additionally the recommended circuit configuration of the LTZ1000 gives the zener diode current the same temperature coefficient as the Q1 Vbe. That current temperature coefficient produces a zener voltage temperature coefficient proportional to the zener slope resistance / the 120R. Using the mid range datasheet slope resistance of 40 ohms means the effect of the Q1 Vbe temperature coefficient is multiplied by 1.3.

So yes you can change the balance of temperature coefficients by adjusting the 120R with the likely undesirable side effect of changing the zener current. You can also change the zener slope resistance in one direction by adding some resistance on top of the zener. As I said I would not be surprised if they add and laser trim some resistance on top of the zener to get the temperature coefficient match in the first place (has anyone cut the lid off one to look?).

Lol at sitting for hours with an LTZ1000 in an oven with Q1 collector base shorted to determine a perfect operating point which the circuit doesn't actually use. What winds me up is you are supposed to be the volt nut and I am just an electronics/software engineer who has never even seen an LTZ1000.
 

Offline Rufus

  • Super Contributor
  • ***
  • Posts: 2094
Re: Ultra Precision Reference LTZ1000
« Reply #165 on: May 28, 2013, 09:17:01 am »
Well, we will just have to disagree.  You keep talking about "avalanche mode" and I am talking about the transition point between "tunneling mode" and "avalanche mode";

I keep talking about avalanche mode because the zener in the LTZ1000 is required to operate with a +ve temperature coefficient of around 2.3mV/C which means the zener is predominantly if not completely in avalanche mode. The transition point (if it has one) is irrelevant it doesn't operate anywhere near the transition point.

We are trying to cancel the temperature coefficient of the zener avalanche voltage with the temperature coefficient of the Q1 transistor Vbe. Q1 Vbe temperature coefficient will drop by about 200uV/C per decade increase of collector current so changing the collector resistance is a valid way to adjust the temperature coefficient matching.

Good luck with that.  It didn't work in my simulations, nor did it work on the bench with any of the LTZ1000 circuits I tried.  But, maybe I did it all wrong-- why don't you show me the Right Way?

Andeas stated a couple of posts ago that changing the 70k to 50k did change the temperature coefficient of the reference slightly. Vbe temperature coefficient is proportional to log of collector current so I am not surprised a 30% change in current only made a slight difference. You might also consider this from the Fluke 732B manual
Quote
Biasing the Refamp for Low Temperature Coefficient 4-8.
As mentioned earlier, the Reference Amplifier contains an NPN transistor and a zener diode in series. The TC (Temperature Coefficient) of the Reference Amplifier is the sum of the TC of the zener voltage and the transistor base-emitter voltage. The zener voltage TC is negative and the transistor TC is positive with a value dependent on its collector current. Each Reference Amplifier is pretested to determine the collector current at which the two TCs cancel out yielding an overall Reference Amplifier TC very close to zero. To generate this same collector current in the standard, a voltage of 2.976V is generated across thin film resistor Z1-R3 on the Reference Hybrid (HR1). This resistor is pretrimmed with a laser to the value that results in the correct collector current.

So yes you can change the balance of temperature coefficients by adjusting the 120R with the likely undesirable side effect of changing the zener current. You can also change the zener slope resistance in one direction by adding some resistance on top of the zener.

Well, as I said, good luck with your experiment.

What experiment? The LTZ1000 datasheet already shows resistance added on top of the zener as a method of adjusting overall temperature coefficient. What do you think the intentionally introduced -ve temperature coefficient of zener current is going to do with the zener slope and any added resistance?

Thanks for the link to the LTZ1000 die photos. I don't see any trimming so I guess they just rely excellent temperature control which probably indicates there is scope for improved temperature coefficient nulling.

Therein lies the difference between you and I-- I have been working in electronics for over 30+ years, and have built my fair share of electronics projects.  And yes-- I have built many voltage references using almost all of the available components.  I have built dozens of circuits based on the LTZ1000, many of them with novel improvements which I cannot go into here due to patent issues.

If you have built so many LTZ1000 based references especially with novel improvements you ought but apparently don't know how they work.
 

Offline nukie

  • Frequent Contributor
  • **
  • Posts: 777
  • Country: au
Re: Ultra Precision Reference LTZ1000
« Reply #166 on: May 28, 2013, 11:25:34 am »
I love how our friends in China are pushing the development of precision DIY references. Having easy access to affordable used/recycled precision parts sure helps alot. Unfortunately, it seems that used LTZ1000 is getting rare lately which drives up the prices. This is v3.0 of the LTZ1000 reference board I got(v2.0) from jj3055, use lots of high precision resistors. v3.0 improvement is low noise suitable for 10V buffer but it is not tested for long term stability. V2.0 is the 'community' standard.

http://item.taobao.com/item.htm?spm=0.0.0.0.jPEYqD&id=22720224575

More expensive version sports the AE metai foil oil filled resistors.

http://item.taobao.com/item.htm?spm=0.0.0.0.jPEYqD&id=17294769547

« Last Edit: May 28, 2013, 11:37:18 am by nukie »
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1873
  • Country: de
Ultra Precision Reference LTZ1000
« Reply #167 on: May 28, 2013, 09:10:13 pm »
Dear fellow-nuts & EEVbloggers,

I’ve followed this thread here and in other forums around the LTZ1000 with great interest, especially the discussion about gimmicks (*), additional to the original simple design from the datasheet.

But on the implementation of those gimmicks I am desperately missing well-founded engineering & metrology practices, as:

1) A qualitative explanation, what their purpose is / how they work / root cause of additional instabilities (theoretical model)
2) A quantitative analysis, how much they  mitigate possible instabilities (error analysis calculus)
3) Practical stability measurements on realized LTZ1000 references, especially on the effects of additional gimmicks
4) An estimation, if the additional gimmicks noteworthy improve the LTZ1000s basic circuitry stability (e.g. 1ppm/yr., 0,2ppm/K), or if they contribute in the range of a few tenths ppm only, which would make them useless or exaggerated.

Instead, most designs only copy those gimmicks found in references of Datron, Fluke or HP, obviously without understanding, and forgetting the most obvious aspects.

I have to admit, that me too, have only copied the basic circuitry from the LT datasheet, but the few features I have considered or added, have been analyzed deeply in terms of 1) – 4).
Also, before I started, I had a design goal concerning the stability of my box, which was met.

I also have designed additional circuitry to set up a complete reference system, i.e. to calibrate the complete DCV range on a calibrator like the Fluke 332B, or a HP34401A, from one LTZ source.

I would like to present my design and my findings in the following days, also presenting some ideas, what perhaps is much more important than all those gimmicks (i.e. low hanging fruits)

Anyhow, I’m still interested in a profound discussion of those gimmicks, and would implement them in a future redesign, but only after given explanations 1) – 4).
 
Btw: The Chinese colleagues are doing a great job of showing the interior of the Holy Grails, and also in setting up LTZ boards in numbers, but sorry to emphasize that, I did not yet find any new/improved designs, or explanations there, either. (Hope I haven't overlooked that)

Anyhow, I would greatly appreciate, if some of them also would join the discussion here, or on Volt-Nuts.

And of course, I would also like to personally meet and exchange with Metrology-Nuts here in Germany.. 'babysitter' I already have met  :-+

Frank

(*) slots in the PCB, guarding rings, usage of “A” version, excess temperatures. metal foil resistors, zero offset OpAmps, ovenizing the complete LTZ circuitry, Burn-In on assembled PCB, cleaning in US bath, TempCo trimming .....
« Last Edit: May 28, 2013, 09:44:21 pm by Dr. Frank »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2545
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #168 on: May 28, 2013, 09:24:42 pm »
Which is best to adjust I don't know. I would imagine shipped parts are close to their sweet spot with 120R and 70k resistors, by design and/or trimming.

Won't there already be some 'voltnut' talk about stuff like this? I'm just looking at the circuit and making it up.
Definitely no trimming as you can see on the chip photo. Btw. the transistors Q1 and Q2 are built each out of 4 single transistors in a staggered configuration around the zener. This will shurely help to average out any temperature gradients on the chip.
On voltnut we talked about the additional 200 ohms resistor which is definitely no option for me. The TC of this resistor goes with a 1:7 ratio directly to the output voltage. The other resistors have a much lower influence on the output voltage.

Quote from: DiligentMinds.com
I have built dozens of circuits based on the LTZ1000, many of them with novel improvements which I cannot go into here due to patent issues.
Can you give us the affected patend numbers?

With best regards

Andreas
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2545
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #169 on: May 28, 2013, 09:41:29 pm »
Instead, most designs only copy those gimmicks found in references of Datron, Fluke or HP, obviously without understanding, and forgetting the most obvious aspects.

Hello Frank,

one often overlooked gimmic from the datron schematic is the 100nF capacitor from the base of the temperature sensing transistor to the emitter. Especially if there is any connection outside the shielded box (power supply) this capacitor will help to stabilize the output voltage. Otherwise any mains disturbance has a direct influence on the temperature setpoint of the reference and the output voltage will drift.

with best regards

Andreas
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1873
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #170 on: May 28, 2013, 09:55:32 pm »
That's great, Andreas!

my references suffer from short termed disturbances, as such, obviously. (but that does not affect the more important long term stability)

The temperature control knot, i.e. the base of Q2 obviously is very sensitive to external influences, so far I've explored the circuitry by myself.

The box is currently undergoing a refurbishment, so I will add that.

Thanks.

Frank
 

Offline quantumvolt

  • Frequent Contributor
  • **
  • Posts: 395
  • Country: th
Re: Ultra Precision Reference LTZ1000
« Reply #171 on: May 28, 2013, 10:13:06 pm »
This very interesting tread contains a lot of hot air  :-DD but not very many builds (except from the ready built unit shown above).

The tread imo also contains quite heated arguments  :box: arising from different views on what really constitutes necessary elements in actually building a working reference. Some of the things mentioned in this tread are: Enclosure, PCB, Thermal Gradients  and EMF.

It amazes me however that some posters use all this energy on discussing the merits or lack of such for details of building without first checking datasheets (& more) from manufacturer(s). {I might be wrong, but I cannot find any references to the text below}.

Here is a "copy and paste" from  http://cds.linear.com/docs/en/datasheet/1000afd.pdf  :
(Please forgive me the lack of decent formatting - blame "Control-C"   :palm: ).

---

LTZ1000/LTZ1000A
4
1000afd
a
pplica
T
ions
i
n
F
or
M
a
T
ion
LT Z
1000 and
LT Z
1000A are capable of providing ultimate
voltage
reference
performance. Temperature
drifts
of
better
than
0.03
ppm/°C and long-term stability on the order of
1?V per month can be achieved. Noise of about
0.15
ppm
can also be obtained. This performance is at the expense
of circuit complexity, since external influences can easily
cause output voltage shifts of more than 1ppm.
Thermocouple effects are one of the worst problems and
can give apparent drifts of many ppm/°C as well as cause
low frequency noise. The kovar input leads of the TO-5
package form thermocouples when connected to copper
PC boards. These thermocouples generate outputs of
35?V/°C. It is mandatory to keep the zener and transistor
leads at the same temperature, otherwise
1
ppm to
5
ppm
shifts in the output voltage can easily be expected from
these thermocouples.
Air currents blowing across the leads can also cause small
temperature variations, especially since the package is
heated. This will look like
1
ppm to
5
ppm of low frequency
noise occurring over a several minute period. For best
results, the device should be located in an enclosed area
and well shielded from air currents.
Certainly, any temperature gradient externally
generated
,
say
from a power supply, should not appear across the
critical circuitry. The leads to the transistor and zener
should be connected to equal size PC traces to equalize
the heat loss and maintain them at similar temperatures.
The bottom portion of the PC board should be shielded
against air currents as well.
Resistors, as well as having resistance temperature coef-
ficients, can generate thermocouple effects. Some types of
resistors can generate hundreds of microvolts of thermo-
couple voltage. These thermocouple effects in the resistor
can also interfere with the output voltage. Wire wound
resistors usually have the lowest thermocouple voltage,
while tin oxide type resistors have very high thermocouple
voltage. Film resistors, especially Vishay precision film
resistors, can have low thermocouple voltage.
Ordinary breadboarding techniques are not good enough
to give stable output voltage with the
LT Z
1000 family
devices. For breadboarding, it is suggested that a small
printed circuit board be made up using the reference, the
amplifier and wire wound resistors. Care must be taken to
ensure that heater current does not flow through the same
ground lead as the negative side of the reference
(
emitter
of Q1). Current changes in the heater could add to, or
subtract
from, the reference voltage causing errors with
temperature. Single point grounding using low resistance
wiring is suggested.

---
 

Offline nukie

  • Frequent Contributor
  • **
  • Posts: 777
  • Country: au
Re: Ultra Precision Reference LTZ1000
« Reply #172 on: May 28, 2013, 10:33:04 pm »
Hello Frank,
The Chinese community has more shared data and discussion on long term stability. Actually, it's a little hard to find schematics and implementation. You need to dig harder in the Chinese forums to find what you find missing.
 

Offline quantumvolt

  • Frequent Contributor
  • **
  • Posts: 395
  • Country: th
Re: Ultra Precision Reference LTZ1000
« Reply #173 on: May 28, 2013, 10:46:54 pm »

The temperature control knot, i.e. the base of Q2 obviously is very sensitive to external influences, so far I've explored the circuitry by myself.


Copied from the datasheet (link in earlier post):


It is mandatory to keep the zener and transistor leads at the same temperature, otherwise 1 ppm to 5 ppm shifts in the output voltage can easily be expected from these thermocouples.

The leads to the transistor and zener should be connected to equal size PC traces to equalize the heat loss and maintain them at similar temperatures.

..
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1873
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #174 on: May 29, 2013, 07:00:21 am »

The temperature control knot, i.e. the base of Q2 obviously is very sensitive to external influences, so far I've explored the circuitry by myself.


Copied from the datasheet (link in earlier post):


It is mandatory to keep the zener and transistor leads at the same temperature, otherwise 1 ppm to 5 ppm shifts in the output voltage can easily be expected from these thermocouples.

The leads to the transistor and zener should be connected to equal size PC traces to equalize the heat loss and maintain them at similar temperatures.

..

No, I meant electrical / electromagnetic disturbances, those thermoelectrical aspects I have considered in the design and the enclosure sufficiently, I think.

If you simply measure the U(BE) of Q2 with a high impedance DMM, the stabilization circuitry will be easily disturbed and the regulation runs wild.
Accordingly, this is an entry point for EMC disturbancies, explaining perhaps my observed instabilities.

Frank
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf