New TI REF54EVM Evaluation Module Leaked

[laichh]

Look at what I've found.

https://www.ti.com/lit/ug/snau289/snau289.pdf

Description
The REF54EVM is a precision voltage reference evaluation module that demonstrates the performance of high precision series reference device (REF54) from Texas Instruments (TI). The REF54 is a family of high precision, low drift, low current consumption series voltage reference devices. The REF54 family offers low temperature drift coefficient (0.5 ppm/°C), low flicker noise (0.16 ppm p-p with 100uF capacitor on NR pin) and high accuracy (±0.02%), while consuming 260 μA current.

Features
• Enable and disable the output
• Capacitor at NR pin is configurable with 0 ohm resistor to optimize the noise performance

Applications
• Semiconductor test equipment
• Precision data acquisition systems
• Precision weight scales
• Ultrasound scanner
• X-ray systems
• Industrial instrumentation
• PLC analog I/O modules
• Field transmitters
• Power monitoring
• Battery management system

Introduction
The REF54EVM is a series voltage reference evaluation module that demonstrates the REF54 performance in SOIC package. The REF54 with low long-term drift (30 ppm), excellent load and line regulation helps meet strict performance requirements of high precision applications. This device also comes with a enable pin (EN) that allows the device to be set in shutdown mode to improve power efficiency. This user's guide describes the characteristics, operation, and recommended use cases of the REF54EVM. This document provides examples and instructions on how to use the REF54EVM board. Throughout this document, the terms evaluation board, evaluation module, and EVM are synonymous with the REF54EVM. This document also includes a schematic, reference printed circuit board (PCB) layouts, and a complete bill of materials (BOM).

[Andreas]

Is it anywhere available?
Not at DigiKey or Mouser.

with best regards

Andreas

[RandallMcRee]

Looks like there is a webinar announcing this product on Dec 14.

Seems like TI's answer to the ADR45xxD which has similar specs (0.8/C TC, 0.02% Vout, low noise)

[DavidKo]

If I catch it correctly, than the mentioned specs will be for 2.5V reference (one used in breadboard). I expect that at 5V the noise will be double - common in TI datasheets and specs.

[Andreas]

Yes,
Noise is given as ppm p-p in this case so uVpp/V output voltage.
with best regards
Andreas

[Andreas]

Seems like TI's answer to the ADR45xxD which has similar specs (0.8/C TC, 0.02% Vout, low noise)

Hello,
was not aware of this package.
When I worked with ADR4550B (in ADC18) the ageing drift was rather high. And also thermal hysteresis.
Do you have experience with the ADR4550D ?

with best regards
Andreas

[iMo]

While looking at the ADR4550 (and friends) output voltage vs temperature graph in the DS - not an optimal for a precision reference.. In 0-60degC a pretty downwards slope..

[RandallMcRee]

Seems like TI's answer to the ADR45xxD which has similar specs (0.8/C TC, 0.02% Vout, low noise)

Hello,
was not aware of this package.
When I worked with ADR4550B (in ADC18) the ageing drift was rather high. And also thermal hysteresis.
Do you have experience with the ADR4550D ?

with best regards
Andreas

I do have some experience, although no graphs to share. Currently, I am creating a daughter board for the AD5791 EVM using a mix of ADR4525D and ADR4550D chips. Using an LT5400/ADA4523-1 to gang up the 5v to 10v I see a daily difference <1ppm in my heated room (no enclosure currently!).

There ought to be a separate datasheet for the ADR45xxD, e.g. look at hysteresis, for example:

You can see that the D grade is quite good. Again, that has been my experience, as well. But, like much of AD's inventory, impossible to get more of, lately.

Also, the long term drift has been very low. In the datasheet, that is also broken out separately since the D grade is much better.

Thanks,
Randall

[floobydust]

REF54 seems to be a REF70 variant with added NR pin and deleted SENSE pin? The REF54EVM datasheet has cut'n'paste errors.
4.1 "The schematic for the REF70EVM is illustrated in Figure 3-3"  uh there is no fig. 3-3 and it's not the 70.

100uF for C_NR seems large, and it's a polymer tantalum Panasonic POSCAP 16TDC100MYF which has more leakage current above a tantalum-only part, not sure if that is a problem for the IC's new internal node. There's no thermal reliefs on the EVM board so working with that is not possible.

[ivo]

I did wonder why all the modern references started dropping the NR pin. A single good cap gets you to a much better starting point than having to implement a good filtering solution.

[ivo]

While looking at the ADR4550 (and friends) output voltage vs temperature graph in the DS - not an optimal for a precision reference.. In 0-60degC a pretty downwards slope..

I notice they have a typo on the vertical scale for the 4550D graph 



They all have a nice slope from 10-40, these are guestimated numbers:

grade/temp	10	40	Diff
B	5.000085	4.99985	-0.000235
C	4.99908	4.999025	-0.000055
D	4.99962	4.999538	-0.000082

The best SOIC'er wins!

Although the LCC8 is supposed to be regarded for removing humidity from the equation. Pretty sure I remember seeing some Andreas' posts on how annoying that can be to a finely tuned plastic reference over a year.

[iMo]

I did wonder why all the modern references started dropping the NR pin. A single good cap gets you to a much better starting point than having to implement a good filtering solution.

Perhaps there is not such a "good cap" available.. So they rather dropping the NR pin - otherwise the customers will complain their reference with a larger cap wired to the NR pin fluctuates much more than without the cap..

[iMo]

While looking at the ADR4550 (and friends) output voltage vs temperature graph in the DS - not an optimal for a precision reference.. In 0-60degC a pretty downwards slope..

I notice they have a typo on the vertical scale for the 4550D graph 

Keeping the 4550"D" at those aprox 45C would definitively help, it also repels the humidity off the package a little bit.

[Kleinstein]

The NR pin definitely makes sense with a reference that includes a buffer to directly drive an ADC.  Usually this is only for the higher frequency noise part, but this can still matter as some ADCs also react to higher frequency noise at there ref. input. Just a relatively large capacitor at the the low impedance reference output is not the most efficient way of filtering - though in parts also not as bad as one may think, as the output impedance is usually inductive and thus making it a 2nd order filter.
When using an external buffer / ADC driver anyway one can as well add filtering there and have the opition to choose the series resistor.

[eplpwr]

Although the LCC8 is supposed to be regarded for removing humidity from the equation. Pretty sure I remember seeing some Andreas' posts on how annoying that can be to a finely tuned plastic reference over a year.

In the ADR45xx datasheet they show the effect of humidity on the SOIC-8 references. What is missing - for comparison purposes - is how the ceramic "D" packages handle humidity. In the document "Analog Dialogue Volume 53 Number 3" there is an article called "Does My Voltage Reference Design Hold Water?" that actually gives the numbers for the LCC-8 packages as well (for ADR4525). The numbers are telling: 2.5 ppm/%RH for SOIC-8 vs 0.075 ppm/%RH for LCC-8. The article recommends hermetic packages for highest precision. The tempco is still a problem - maybe an oven can solve that?

Now, hermetic packages and an oven to handle the tempco - I think I've encountered that concept before, somewhere. 

[Andreas]

vs 0.075 ppm/%RH for LCC-8. The article recommends hermetic packages for highest precision.

I doubt this number if the package is soldered to a epoxy PCB. (even with slots).
Perhaps whole cirquit on a ceramic substrate.

With best regards

Andreas

[iMo]

Dead bug wiring should be none problem..

[Andreas]

Hello Imo,

is that picture from your ageing measurements of ADR1001 in the other thread?
(is that the reason why it is that good?)

with best regards

Andreas

[iMo]

Yes, therefore it is so good..

[laichh]

The REF54 is now in PREVIEW status on TI website:

https://www.ti.com/product/REF54

[RandallMcRee]

The REF54 is now in PREVIEW status on TI website:

https://www.ti.com/product/REF54

Yes, I noticed this, as well. And...REF54 is a big nothingburger! All of the supposedly state-of-the-art guesstimates from before are now either exactly matching ADR45xxD or are worse! For example, 0.5 ppm max TC is now 0.8ppm, same same.

It's another case of TI copying a successful ADI product, I guess.  Perhaps also a second source is profitable in a niche market like this?

[ivo]

Interesting, looking at the application note for the datasheet, they want at least a 1uF cap on the output ('must be connected') but warn against the cheapest route, X7R MLCC.

Perhaps also a second source is profitable in a niche market like this?

Having another OEM for a good part is never a bad thing

[Kleinstein]

The piezo effect of X7R and similar capacitors is well known. In a quiet environment they can still be OK and for the lower frequency part the reference and not the capacitor is relevant. So it depends on the application if an X7R at the output is acceptable. For many uses this should be OK, as high frequency noise may not be that relevant and other parts of the circuit may be sensitive to mechanical coupling too.

An iteresting point is getting low noise with a relatively low power consumpten  (like 40% of the ADR45xx).
Being pin compatible (with the optional capacitor at the NR pin) is a good thing.

[floobydust]

I think price-point of the REF54 will tell if it's a bit of competition for ADI, or just a "we're still in the game" product for TI. Website gives USD $6.95/1ku - which is a top price, and yet it's another plastic-package reference IC with marketing numbers.

All the acquisitions by TI and ADI have made a huge mess. I was reading of ADI closing the Linear Tech Hillview plant (wafer fab and packaging) where LTFLU1, LTZ1000 etc. all TO packaged parts were built- and offshoring all that, and outsourcing the packaging as well. To save not much money actually. What a heartbreaker. Hillview had a 35,862 square-foot clean room.

At a time the USA wants semi fab back in America 

[deepfryed]

The numbers are telling: 2.5 ppm/%RH for SOIC-8 vs 0.075 ppm/%RH for LCC-8. The article recommends hermetic packages for highest precision.

Can you dry the PDIP package and then coat it in a marine waterproof epoxy to get the same benefit ?

[miro123]

Can you dry the PDIP package and then coat it in a marine waterproof epoxy to get the same benefit ?

I think that you are referring to conformal coating. Humiseal  is leader in this area. I evaluated different conformal coating and potting in the past. My conclusion match the findings from AD. Conformal coating only delays reaction to humidity variance. Any material absorb water, some material dry faster. Adding conformal coating makes you mathematical model or resistor/voltage reference more complex as it was.
https://eprintspublications.npl.co.uk/3870/1/DEPC_MPR60.pdf
https://www.analog.com/en/analog-dialogue/articles/does-my-voltage-reference-design-hold-water.html
https://chasecorp.com/humiseal1/wp-content/uploads/sites/12/2018/10/1B51-TDS.pdf

[miro123]

Interesting, looking at the application note for the datasheet, they want at least a 1uF cap on the output ('must be connected') but warn against the cheapest route, X7R MLCC.

Perhaps also a second source is profitable in a niche market like this?

Having another OEM for a good part is never a bad thing

With the risk that I offend AD fanboys I will make the following statement.
 - @ 9-th of December 2023 ADR4525D - only exist on datasheets and Youtube intro videos
 - ADR4525C grade is available on insane price of 21..22Euro -23..24USD
Please correct me if I'm wrong
I'm lucky guy. I have chosen Ref7025QFHT. It is more down to earth device - broadly available and reasonable price of 12 Euro for single pcs. 
My point is, yes the second source is always good to have.

[Kosmic]

Can you dry the PDIP package and then coat it in a marine waterproof epoxy to get the same benefit ?

I think that you are referring to conformal coating. Humiseal  is leader in this area. I evaluated different conformal coating and potting in the past. My conclusion match the findings from AD. Conformal coating only delays reaction to humidity variance. Any material absorb water, some material dry faster. Adding conformal coating makes you mathematical model or resistor/voltage reference more complex as it was.
https://eprintspublications.npl.co.uk/3870/1/DEPC_MPR60.pdf
https://www.analog.com/en/analog-dialogue/articles/does-my-voltage-reference-design-hold-water.html
https://chasecorp.com/humiseal1/wp-content/uploads/sites/12/2018/10/1B51-TDS.pdf

I wonder if potting with a really tick layer could give better results ? From AD tests, 1B73 applied on the board seem useless.

[Kleinstein]

A thick layer of potting material or similar plastics would add additional possible stress from curing / thermal expansion. It would make things slower, but also more compicated with more processes to add drift processes and maybe slower setttling. Unless really water repellant I am afraid it would only make things worse.

There is still quite some long term drift for both the ref54xx and ref70xx. So I don't see them a replacement for a hermetic reference. The nice point is gettting away with low power and having a reference driver included. I already see little sense in using the ref54xx together with an external reference buffer. The main use for the ref54xx would be more less demanding ADCs / DACs that can woork with the internal driver directly. If the ADC needs a lower impedance, the ref70xx may be the better choice.

[iMo]

Keep away from plastic/epoxy packages when targeting something more demanding.
The only suitable packages are the metal hermetic one, then ceramic one (when not soldered directly onto epoxy pcb). There is not a sealant available which would make an epoxy package hermetic..

[miro123]

There is still quite some long term drift for both the ref54xx and ref70xx. So I don't see them a replacement for a hermetic reference.

Just for clarification, not all ADR45xx are in ceramic package and not all REF70xx are plastic. In case of ADR4525 only D grade is ceramic package - unfortunately only awailable on datasheets and AD advertisements. C grade in plastic is available for 22Euro/pcs
In case of REF7025 in ceramic package is available in most shops ~12Euro/pcs -  https://nl.mouser.com/ProductDetail/Texas-Instruments/REF7025QFKHT?qs=TuK3vfAjtkUpFxL4w6H61A%3D%3D

[miro123]

Hopefully I will no  hijack the threat with my questions.
 1. How semiconductor manufacture measure the output voltage thermal hysteresis
   - Temperature ramp-up rate - e.g. from 25->70C with ram rate of 10mK/min
   - time to stay at temperature - e.g. 10 min, 24H or many days stay @70C
   - ramp down time
  - time to stabilize at 25C before take measurements.
2. How the semiconductor manufacture measure the long term drift - The list of parameters is enormous.
3. Is there any ISO standard for testing 1 & 2

[floobydust]

Long-Term Drift in Voltage References TI App. Note  SBAA436 they use an oil bath to measure LTD, so it appears moisture is not included as a factor and I would say misleading for plastic-package parts.

REF7025 LCC package same price point as REF54 SOIC-8.  I guess the $ is for the lower quiescent current.

[Andreas]

Hello,

the sad truth is that you often cannot compare these values.
In some older data sheets hysteresis is given for 25 deg C +/- 25 deg. In others over the full commercial or industrial range.
As a rule of thumb the drift is proportional to the square of the temperature excursion.

Often hysteresis and ageing drift are only given for the best package (metal can) and the information is not updated if this package is no longer produced ...

Some specs for the hysteresis are given for the 2nd temperature cycle.
Or for the 2nd 1000 hrs regarding ageing.

So I never trust the datasheet values.

with best regards

Andreas

[eplpwr]

In my earlier post, I should have included the link to Analog Dialogue Vol 53 number 3. Here it is:

https://www.analog.com/media/en/analog-dialogue/volume-53/number-3/volume53-number3.pdf

There are no pictures of the board(s) they used, but they tried different HumiSeal coatings for the plastic Vrefs. In summary, the coating didn't do much.

BTW, somewhere in my stash I have 10pcs ADR4550DEZ, so they do exist. I remember putting in a back order with Mouser DigiKey, and receiving the chips long after I had forgotten about my order.  [Shipped 4th of june 2021]

Edit: Wrong distributor specified.

Edit2: DigiKey order history says I ordered (and they shipped) 2 pcs ADR4525DEZ on 2021-11-15, so those exists as well. Maybe I need some better inventory management skills ...

[floobydust]

I wonder if they are collectibles    Mouser does not list the ADR45xxD. Digi-Key shows no stock, no price, no lead time - request quotation only.
There were delays closing the Milpitas wafer fab(s)/hermetic packaging, it might have happened end 2021/beginning 2022 and you got that stock? But today it appears the parts are unobtainium.

Why is the new REF54 not offered in the LCC-8 package?

[Kleinstein]

Why is the new REF54 not offered in the LCC-8 package?

That is a good question. Still with the ref70xx the performance in the LCC-8 package is still not that much better than in a plastic case as one may hope for.
If power is not a big issue the ref70xx may be the alternative that is available in LCC-8.

[floobydust]

It must be marketing specmanship - advertise the Holy Grail of low TC, yet moisture and flex stress of the plastic package undermine the overall performance and let's all be in denial and not mention it in app notes or datasheets.

I found this on LCC CTE which I was wondering about, did not compare to old friend TO-46 package rolled out in 1962, it's over 60 years old.

[Andreas]

Hmm,

https://cleanroom.byu.edu/cte_materials

Kovar (the bottom of a TO-package) is 5.9
Silicon is 2.6
The big advantage of a TO-package is that the PCB stress is decoupled by the wires.

with best regards

Andreas

[floobydust]

I was thinking of thermal properties between the PCB and LCC, and any advantages of the LCC package beyond the hermetic seal.

LCC ceramic material I find Kyocera A440 CTE 7.1 and thermal conductivity 14.0 W/mK, Young's Modulus 310.
So it is quite rigid and an excellent conductor of heat. CTE is greater than Kovar by a little. FR-4 CTE is 13-17 and I have no idea how much stress is on the IC package after reflow, if it's significant. Ceramic is very stiff.

A reference with on-die heater i.e. ADR1399KEZ "The heater power for the LCC version is about 3× to 4× greater than the TO-46 version." They could probably use the LTZ1000A die attach technology if they weren't lazy.

So it seems a REF54 in LCC would only benefit from the hermetic seal against moisture, and see much less package flex and die stress from that?
I remember a product SOIC-8 op-amps that had huge offset shifts when the PCB was flexed, it was a nightmare in production to calibrate.