10V Octo LT1021 Reverence

[blackdog]

Hi Alex,

What are you doing, should I now go back to my drawing board? 

Thanks!
That I didn't think about this configuration myself 
I will keep it in mind.

This weekend I had decided to got for the minimum supply voltage of 12V. (battery backup...)
This will not work with the LT1021 stacked.
The upper LT1021 requires a minimum of 3.5V to work properly.
This I have already investigated years ago, which is also the reason that the first schematic also indicated the supply line of 13.5V.

So many decisions I will still have to make...
But keep them coming!

Kind regards,
Bram

[blackdog]

Hi,

Today I want to show how I started measuring 14x the LT1021 in a metal housing.
I can't show all measurement data because I don't have access to it now.

But first let's show some respect for the power supply I've used for almost 5 years of continuous operation to test LT1021 references in two groups.
The first group has resulted in my 10V LAB reference is now still use.

I built this power supply over 5 years ago to gain some experience with pre-regulators to limit the dissipation.
I had to, because I had chosen to build the power supply in one of the old Cisco Pix firewalls housings i had laying around.
The beauty of this housing is that it is very solidly built of thick iron.
As you can see here, I can build three more of these power supplies! 
The red push button has a double function, a short pulse turns the output on or off.
If you hold the button down longer, the power supply switches between a maximum of 30mA or 200mA output current.


The reference in this power supply is a VRE310a and the quality of this reference is more than sufficient for this power supply. 
I have searched for pictures of the inside of this power supply, but some are erased.
I will check later if I can find anything in one of my backups.

The properties of this power supply are very good and it is also very stable.
However, I am not satisfied with this power supply.
The switching pre-controller radiates into the metal of the box and the box radiates back into the sensitive electronics.
I can't win because it's all too close to each other...

The residue of the Pre-Switcher is approximately 1.5mV under full load.
And these specifications don't fit in with my company profile. 

So ... do you want to see the schematic of this power supply?
This is the power supply part together with some switching electronics.



And this is the linear controller, it has two difference amplifiers, one for current measurement
and the other for voltage measurement over the Power MOSFet and the difference signal then goes to the pre-regulator.



And now back to the LT1021 test setup!
The power supply is hidden here, so I can't turn it off by accident or change the settings.
It is also visible here that at 15V supply voltage the consumption of the 14x LT1021 references is 10.8mA.
The KeySight E3612A is not used in this setup.



Yes! breadboard, that has proven to be good enough for this test setup.
I used two ways to measure the LT1021 voltage so I didn't have a problem with the breadboard paracitic properties
The 1K resistors at the top connect all LT1021 to each other, so that I also have the average voltage of these 14 units.


The first values are noted...



This is one of the waysi used to measure the outputvoltage of a LT1021,
after installing the measuring clamps I leave the whole thing alone for a while until everything is at the same temperature again for a stable readout.



The list is getting longer...



Three months of drift data.
Be aware of the temperature, humidity and also the drift of the TEK 4050 DMM.



And now some data from a year later, there is more, but i can't show it right now, the notebook is on the other side of the world...



This it it for know!

Kind regards,
Bram

[GigaJoe]

I'm not hijacking the topic 

But if anyone would like to play with N - umerous  LT1021 , I may recommended this:   Just tell seller 5 or 10 version would you like.
https://www.aliexpress.com/item/-/32856068339.html

Yes it plastic ,  so an additional brainstorming needed , like use a beer-can, or champagne for humidity seal . Is it just having fun, isn't it ?

[serg-el]

blackdog, резисторы R5, R8 ..... R33 выбраны неверно. Их сопротивление должно быть гораздо больше. Например 200 кОм.
Они служат для зарядки С20 и отмены тока утечки С19.
И достаточно одного резистора.
Тогда ваша схема будет работать правильно, и частота среза фильтра будет 0.048 0.38 Гц.
Смотрите AN42 от Linear.

[GigaJoe]

Speaking about filter,  ideally it per each chip output, or after summary opamp.  200K need to make potential extremely low on capacitor, therefore decreasing leaking .

Speaking of "serg-el" native language posting, usually it looks awkward  and arrogant. I may imagine how it looks, when any post will full of holographics messages ...  Dude "serg-el" - please use a translator at least ...

[branadic]

blackdog, resistors R5, R8 ..... R33 selected incorrectly. Their resistance should be much greater. For example, 200 kOhm.
They serve to charge the C20 and cancel the leakage current of the C19.
And one resistor is enough.
Then your circuit will work correctly, and the cutoff frequency of the filter will be 0.048 0.38 Hz.
See AN42 from Linear.

-branadic-

[exe]

But if anyone would like to play with N - umerous  LT1021 , I may recommended this:   Just tell seller 5 or 10 version would you like.
https://www.aliexpress.com/item/-/32856068339.html

Price suggests it's a fake. I think LT/AD don't even have devices in this price range.

[GigaJoe]

it brutally tortured, torched, maltreatment, sadistically abused ...
   

well 20-30 years ago , probably they cost $4-6 as a brand new , not now ...

[exe]

it brutally tortured, torched, maltreatment, sadistically abused ...

Okay, but what about performance? Also, are you sure you bought them from the supplier you suggested? If both are fine I'll bring my apology for not trusting your link.

[GigaJoe]

I dont measure it for precision drifting or so, it a nice reference for this price , possible some good some bad, I measure a few "D" for TC , that gives me a range from 4-12 ppm/C . i'm satisfied.

for trusted or not ... it always gamble on ali.   Extremely cheap gamble if you find a good one, or just money back, if you confirm a fake failure.
Even seller may have no clue it good or not, due to many thousand items he has at store.

If you buy from this link it can be the same batch as my or a different one ..who knows ...

[iMo]

Let me kindly ask the experts here: those ebay cheapo LT1021-5:
In the datasheet I can see the 5V version requires a diode connected to the trimming circuit to compensate the TC of the TRIM input.
Am I correct? I want to trim the 5V, but the diode stuff is pretty annoying, isn't it? 

[exe]

Am I correct?

I think you are. But why you want to trim it? Often it is much easier to do calibration in software.

[iMo]

Am I correct?

I think you are. But why you want to trim it?

Sure, it could be done that way..
Another short Q:
The "C" versions are better initially trimmed (and with narrower TRIM range), but with higher tempco (why??) than the "B" version. So when targeting the lowest tempco are the "B" versions way to go?

[exe]

The "C" versions are better initially trimmed (and with narrower TRIM range), but with higher tempco (why??) than the "B" version. So when targeting the lowest tempco are the "B" versions way to go?

Wow, never seen this before. It's definitely not usual. My guess is they trim at only one operation point, and do not adjust compensation. Bandgap voltage references need a complicated temperature compensation, having just matched bjts is not enough for lot tempco (see https://en.wikipedia.org/wiki/Bandgap_voltage_reference). But that's just my theory.

[iMo]

Here is a nice app note on LTs VREFs saying what is going on with 1021 B/C/D

B versions are "Very Low Drift"
C versions are "Very Tight Initial Tolerance"
D versions are "Low Cost High Performance"

There is not such thing like "Very Low Drift & Tight Initial Tolerance" version..

It could be the tight initial tolerance fab trimming requires some additional stuff on the chip die, thus the "low cost" tempco with the C parts.. A pity..

[iMo]

Here is the die-shot of the LT1021C chip with 15 trimming fuses - when you download the picture and zoom in you may see some of the fuses (see below) are blown - it could be the not trimmed B versions engage less resistors in the trimming network, thus the lower tempco.

[Alex Nikitin]

The "C" versions are better initially trimmed (and with narrower TRIM range), but with higher tempco (why??) than the "B" version. So when targeting the lowest tempco are the "B" versions way to go?

Wow, never seen this before. It's definitely not usual. My guess is they trim at only one operation point, and do not adjust compensation. Bandgap voltage references need a complicated temperature compensation, having just matched bjts is not enough for lot tempco (see https://en.wikipedia.org/wiki/Bandgap_voltage_reference). But that's just my theory.

The LT1021 is not a bandgap, it is a buried zener.

Cheers

Alex

[Andreas]

There is not such thing like "Very Low Drift & Tight Initial Tolerance" version..

Hello,

you are wrong.

The "Very low drift & Tight Initial Tolerance" version is named LT1236A
Unfortunately it is not available in a metal can version which would be best.

compare this chip photo (LT1236)
https://www.eevblog.com/forum/metrology/t-c-hysteresis-measurements-on-brand-new-lt1027dcls8-5-voltage-reference/msg978086/#msg978086

with the LT1021
https://zeptobars.com/en/read/Linear-LT1021-precision-reference-zener

with best regards

Andreas

[branadic]

The "Very low drift & Tight Initial Tolerance" version is named LT1236A

That's an interesting finding.  Didn't knew that or forgot about that.

-branadic-

[Andreas]

Hello,

you never wondered why most of the specs are identical?
Even the pictures for adjustment (with the diode for 5V) and 1/f noise?

with best regards

Andreas

[iMo]

There is not such thing like "Very Low Drift & Tight Initial Tolerance" version.. of the LT1021 ..

.. as the people here are discussing the 1021.. we bought cheap on ali

[branadic]

Hello,

you never wondered why most of the specs are identical?
Even the pictures for adjustment (with the diode for 5V) and 1/f noise?

with best regards

Andreas

To be honest, no. Seems like you are a walking datasheet catalog.

BTW: Here was a better die photo of LT1236LS8.
https://www.eevblog.com/forum/metrology/decapping-the-lt1236ls8/msg906703/#msg906703

-branadic-