Ultra Precision Reference LTZ1000

[Dr. Frank]

...
True-- we are saving only about 600uA in a circuit that uses about 40mA-45mA (including the heater circuit), ..

Which kind a version do you mean?
I use LTZ1000, with external thermal isolation, at 45°C, 12k/1k, LT1013, powered with 12Vdc, standard circuit, plus modifications by Andreas.

The circuit consumes around 50mA initially, but 22mA only when stabilized.
So to which solution relates your 45mA current consumption?

Frank

[Andreas]

Hello,

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

I have 20-24 mA at room temperature. Even with 3*LTC2057 OpAmps which consume nearly 1 mA each.
So with 12*NiMH AA cells you will have easily 2-3 days of operation even if the cells are somewhat older.

With best regards

Andreas

[Kleinstein]

Even in the compensated circuit, there is something like a 5 mA zener current that is needed to get the low noise. So even than it does not make a big difference if the OPs take 600 µA or 300 µA. One might get away with lower zener current it is only for the good long therm stability and not the LF noise.

If you really needs to go for low noise and low power, there would be the rather crazy way of using a battery (e.g. 7.2 V nominal) as an intermediate reference to provide low noise and use some kind of filtering / adjustment to get the long term stability of the LTZ1000 (that might than run at lower current). So the battery (with companion circuit) as kind of low pass filter in the mHz range.

[Andreas]

Hello Ken,

do you have a S&H cirquit which generates no charge injection (noise) above that level?

with best regards

Andreas

[Kleinstein]

For just a reference charge injection is not that critical, as this will be largely a constant offset. So the critical part would be only the drift in charge injection. The H11F1 is also relatively slow (and speed may show drift) - this could add quite some dead time and thus require a low frequency.

Even with pulsing the zener, one would not get away with all the 1/f noise of the reference. It still is the same chip and thus long lived metastable states can still survive the pulsing, and new ones can get populated.  I am not even sure that this would give a real advantage in 1/f noise.

[zlymex]

......
AFAIK, the 1/f noise of a Zener operated in avalanche mode [like the LTZ1K is], will be inversely proportional to the square root of the Zener current.  So, going from 5mA to 20mA would (theoretically) reduce the 1/f noise by 1/2.  I got that from my ancient Motorola Zener Diode Handbook.

That is true. However, there is one thing I don't understand, because I don't know the physics behind the zener LF noise.
I thought the LTZ1k is the best as far as the LF noise is concerned, specifying at 1.2uVpp typical at 5mA zener current.
But, there is this 2DW233(of a particular maker), showing much small LF noise than LTZ1k at 5mA, typically 0.4uVpp   This is equivalent to a LTZ1k operating at 45mA. Anyone can explain the possible reason?
( https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg938465/?topicseen#msg938465 )

[zlymex]

....... 
Is there any way to speak with a scientist or engineer at the factory where those Zeners were made to get some additional information on them?

I tried but no result. I cannot find anything of related publications neither. They didn't mention that 2DW233 in their website in particular(such as special process or ultra-low noise). Those low noise 2DW233 is selling at similar price(all very cheap) as other manufactures with severe noise. The low noise feature of this 2DW233 is only recognized by amateurs like me.

[JS]

....... 
Is there any way to speak with a scientist or engineer at the factory where those Zeners were made to get some additional information on them?

I tried but no result. I cannot find anything of related publications neither. They didn't mention that 2DW233 in their website in particular(such as special process or ultra-low noise). Those low noise 2DW233 is selling at similar price(all very cheap) as other manufactures with severe noise. The low noise feature of this 2DW233 is only recognized by amateurs like me.

Maybe it's just a typo 

JS

[zlymex]

....... 
Is there any way to speak with a scientist or engineer at the factory where those Zeners were made to get some additional information on them?

I tried but no result. I cannot find anything of related publications neither. They didn't mention that 2DW233 in their website in particular(such as special process or ultra-low noise). Those low noise 2DW233 is selling at similar price(all very cheap) as other manufactures with severe noise. The low noise feature of this 2DW233 is only recognized by amateurs like me.

Maybe it's just a typo 

JS

If you meant the noise, no.
There is no specification for the noise in the datasheet. The ultra-low noise feature of 2DW233 was found out by actual measurement of many people at different places and on different times.

[JS]

....... 
Is there any way to speak with a scientist or engineer at the factory where those Zeners were made to get some additional information on them?

I tried but no result. I cannot find anything of related publications neither. They didn't mention that 2DW233 in their website in particular(such as special process or ultra-low noise). Those low noise 2DW233 is selling at similar price(all very cheap) as other manufactures with severe noise. The low noise feature of this 2DW233 is only recognized by amateurs like me.

Maybe it's just a typo 

JS

If you meant the noise, no.
There is no specification for the noise in the datasheet. The ultra-low noise feature of 2DW233 was found out by actual measurement of many people at different places and on different times.

Oh, that's nice then, thanks for the info!

JS

[chickenHeadKnob]

I tried but no result. I cannot find anything of related publications neither. They didn't mention that 2DW233 in their website in particular(such as special process or ultra-low noise). Those low noise 2DW233 is selling at similar price(all very cheap) as other manufactures with severe noise. The low noise feature of this 2DW233 is only recognized by amateurs like me.

I tried to buy these from the ali-express seller you mentioned but after 75 days and no shipment I canceled the payment. At least Aliexpress claims they refunded me - I haven't checked my credit card bill yet. It was the first time I tried aliexpress  so I don't know what the transaction failure rate is. It is apparently less reliable than ebay. I have never had any problems with chinese ebay sellers. Its all for the good though, I changed my mind on what approach to take and testing these would just be a waste of time now. There is some value in the pursuit references that are temp compensated, non-ovenized,  cheap  and maybe a little drifty  but ultra low noise. Say for longscale ADCs that can be disciplined once a day (or more often) to a stable master lab reference.

[zlymex]

I tried to buy these from the ali-express seller you mentioned but after 75 days and no shipment I canceled the payment. .........

I'm sorry to hear that. I didn't know AliExpress before I came across some people at an electric unicycle forum who often buy electronic parts from AliExpress with free international shipping(which is slow). I buy most of my electronic parts at taobao.com that backstaged by the same Alibaba but has a dedicated on-line chat system allowing sellers and buyers to talk. One difference(of taobao and aliexpress with eBay) is that they don't charge/tax e-stores on per item base therefore the average unit price tends to be very low and the number of items in a e-store is huge so is the number of daily packages. These often result in poor management and poor service especially when they are short handed.

Any way, I've bought 500+ pieces of 2DW233 just in case they stop producing it using the same technique(and I may not find the ultra-low noise zener anywhere in the world in that case).

[chickenHeadKnob]

I tried to buy these from the ali-express seller you mentioned but after 75 days and no shipment I canceled the payment. .........

I'm sorry to hear that. I didn't know AliExpress before I came across some people at an electric unicycle forum who often buy electronic parts from AliExpress with free international shipping(which is slow). I buy most of my electronic parts at taobao.com that backstaged by the same Alibaba but has a dedicated on-line chat system allowing sellers and buyers to talk. One difference(of taobao and aliexpress with eBay) is that they don't charge/tax e-stores on per item base therefore the average unit price tends to be very low and the number of items in a e-store is huge so is the number of daily packages. These often result in poor management and poor service especially when they are short handed.

Any way, I've bought 500+ pieces of 2DW233 just in case they stop producing it using the same technique(and I may not find the ultra-low noise zener anywhere in the world in that case).

No need to be sorry! If it wasn't for your interesting insight into the the Chinese semiconductor scene I wouldn't have known about these. My plan was to try with 80 parts, characterize them, and hope that was enough for a lifetime buy. Yes I realize it is too uncertain to source these parts for someone not in China, even the diamond brand factory may not exactly  know how to reproduce them from one batch to another. My plan now is to use LT1027 with a Walt Jung style noise reduction circuit.
best regards

[zlymex]

Using a copper block is a good idea. I've build a test circuit without one and the noise is not as low as expected.
2DW233 is a temperature compensated (two identical zeners back to back) on its own, and altering the applied current will find the zero-TC spot.

[acbern]

Does anybody have data re. the aging rate of the 2DW233? Is it comparable to the LTZ1000?

[Kjelt]

This is getting confusing, is it an idea to start a seperate topic on the 2DW233 ?

[zlymex]

Sure, discussion of 2DW233 is here https://www.eevblog.com/forum/metrology/ultra-low-noise-reference-2dw232-2dw233-2dw23x/

[plesa]

After difficulties with PWW resistors I finally finished all 4 LTZ1000 boards from TiN and started collecting data.
Almost all resistors are PWW made by Edwin except 4x 100k resistors which were out of tempco spec.
He has some manufacturing issues with potting.
Missing PWW were replaced by pre-selected Vishay UXB series resistors.
Initial measurement by 34411A/34410A is rock stable, so I used ratio measurement with 34401A and 34420A.

[plesa]

After difficulties with PWW resistors I finally finished all 4 LTZ1000 boards from TiN and started collecting data.
Almost all resistors are PWW made by Edwin except 4x 100k resistors which were out of tempco spec.
He has some manufacturing issues with potting.
Missing PWW were replaced by pre-selected Vishay UXB series resistors.
Initial measurement by 34411A/34410A is rock stable, so I used ratio measurement with 34401A and 34420A.

Very cool!  I'm looking forward to your measurement results!  Is that your temperature chamber on the right side of the photo?

-Ken

Only usable for measurement seems to be 34420A, 34401A in ratio mode is much more fluctuating (2 last digits) on same references. It is not temperature chamber, only the polystyrene box from pharmacy without any heater. But it seems to be good enough.

[TiN]

Neat setup. I see my earpick trick works for you too  . Will you have datalogging running as well?
Would like to see reference board pics in your setup, once you done with measurements.

And what would be the ultimate plan of using these modules? Long-term testing or some practical application, if that's ok to ask?

[d-smes]

Data sheet thermal resistance specification for the LTZ1000 (non-A version) doesn't seem right.  Looking at the "Die Temperature Rise vs Heater Power" graph in the data sheet, I see 85C rise with 0.5W of heater power which equates to 170 C/W or about twice the specified 80 C/W data sheet value.  The "A" version does looks about right: ~82C rise with 0.2W is close to the 400 C/W data sheet value.  Can anyone explain this discrepancy?   Am I supposed to add normal circuit configuration power dissipation of the transistors & zener to the heater power when doing this calculation?

I ask because I'm leaning towards a 4-layer board layout with considerable copper pours for the leads and inner layers to equalize lead temperature per data sheet recommendation.  I'm wondering what this will cost by way of heater power.  To answer this, I'm trying to determine how much heat escapes by way of the leads versus the case.  I thought comparing thermal resistances of the highly isolated "A" part versus non-A would yield insight.   Instead, I can't seem to make the numbers add up...

[Kleinstein]

Usually the thermal resistance for a TO99 is about 160 K/W. Also the similar size TO39 is in that range. So I guess the 80 K/W number is wrong, or in forced air or with rather short leads.

The thermal resistance of the wire can be estimated form the dimensions and the known material. My calculation give something like 900 K/W for the 8 full 15 mm leads, and about 300 W/K for 5 mm leads. So there is quite some power going through the leads, especially if short. With extra thermal insulation around the case the heat flow through the pins can be the major part.

[JS]

Data sheet thermal resistance specification for the LTZ1000 (non-A version) doesn't seem right.  Looking at the "Die Temperature Rise vs Heater Power" graph in the data sheet, I see 85C rise with 0.5W of heater power which equates to 170 C/W or about twice the specified 80 C/W data sheet value.  The "A" version does looks about right: ~82C rise with 0.2W is close to the 400 C/W data sheet value.  Can anyone explain this discrepancy?   Am I supposed to add normal circuit configuration power dissipation of the transistors & zener to the heater power when doing this calculation?

I ask because I'm leaning towards a 4-layer board layout with considerable copper pours for the leads and inner layers to equalize lead temperature per data sheet recommendation.  I'm wondering what this will cost by way of heater power.  To answer this, I'm trying to determine how much heat escapes by way of the leads versus the case.  I thought comparing thermal resistances of the highly isolated "A" part versus non-A would yield insight.   Instead, I can't seem to make the numbers add up...

  Wouldn't that add to the mechanical stress on the board? As some are using thinner PCBs and/or cut outs to allow for less mechanical stress, I guess adding 2 extra layers would be against that effort. Some soft thermal potting or pad may help equalize the lead temperature without adding mechanical stress.

JS

[plesa]

Neat setup. I see my earpick trick works for you too  . Will you have datalogging running as well?
Would like to see reference board pics in your setup, once you done with measurements.

And what would be the ultimate plan of using these modules? Long-term testing or some practical application, if that's ok to ask?

Yep, they are amazing.Earpick is in lab easy available
Data logging is in progress. I upgraded RPi to latest Raspbian Jessie kernel and it stopped working.
I would like to use them for 10V reference including 10k reference in better case.

[d-smes]

Usually the thermal resistance for a TO99 is about 160 K/W. Also the similar size TO39 is in that range. So I guess the 80 K/W number is wrong, or in forced air or with rather short leads.

The thermal resistance of the wire can be estimated form the dimensions and the known material. My calculation give something like 900 K/W for the 8 full 15 mm leads, and about 300 W/K for 5 mm leads. So there is quite some power going through the leads, especially if short. With extra thermal insulation around the case the heat flow through the pins can be the major part.

Thanks for the confirmation data sheet figure of 80 K/W is wrong.  I just looked at LT1012 data sheet where TO-5 has Rth-JA = 150 K/W and Rth-JC = 45 K/W which further attests to the error.

Once a plastic cap is placed over the part and insulation placed about leads to kill air currents, it looks like overall thermal resistance will be just as high as the "A" part and the majority of heat flow will be through the leads.   It makes sense now why the data sheet recommends equal size PC traces to equalize heat loss and maintain them at similar temperatures.

@JS  The wire leads will offer plenty of mechanical compliance in the x-y axis.  If z-axis (direction of leads) mechanical compliance is necessary, I could add a "S" bend to each lead before soldering into board.   But I doubt mechanical stress is even an issue for a TO-5 part.  I realize someone flexed their reference board and saw several uV of change, but it wasn't clear how much of that came from stress on resistors and op-amps versus the LTZ1000.