Author Topic: REF70 new from TI  (Read 7152 times)

0 Members and 1 Guest are viewing this topic.

Offline fcbTopic starter

  • Super Contributor
  • ***
  • Posts: 2117
  • Country: gb
  • Test instrument designer/G1YWC
    • Electron Plus
REF70 new from TI
« on: April 22, 2021, 11:26:44 am »
Just came across this:

https://www.ti.com/product/REF70
https://electron.plus Power Analysers, VI Signature Testers, Voltage References, Picoammeters, Curve Tracers.
 
The following users thanked this post: edavid, Andreas, ch_scr, MegaVolt, CDN_Torsten, Marco1971, Grandchuck, KT88

Offline Grandchuck

  • Frequent Contributor
  • **
  • Posts: 625
  • Country: us
Re: REF70 new from TI
« Reply #1 on: April 22, 2021, 11:38:44 am »
It is in a ceramic package.  The typical drift is 28 ppm from 0 to 1000 hours. The REF70 will experience the highest drift in the
initial 1000 hr.  Subsequent deviation is typically lower than 13 ppm for the next 1000 hrs.
 

Offline kleiner Rainer

  • Regular Contributor
  • *
  • Posts: 108
  • Country: de
  • Rainer DG1SMD JN48ts
Re: REF70 new from TI
« Reply #2 on: April 22, 2021, 01:15:55 pm »
There is also a new application note on long term drift:

https://www.ti.com/lit/pdf/sbaa436

Greetings,

Rainer
 
The following users thanked this post: MegaVolt, SigurdR

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: REF70 new from TI
« Reply #3 on: April 22, 2021, 06:13:16 pm »
It is in a ceramic package.

Hello,

it is pin-compatible to the LTC6655. And the data is also copied from the LT datasheet except for the supply voltage range which is better for the REF7025.
The large question is: how sensitive is the chip in the SMD package to FR4 board stress due to humidity.
With other LS8-packages (LT1236, LTC6655) I have always had not so good results as with CERDIP 8 packages. (AD586LQ)

with best regards

Andreas
 

Offline ch_scr

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: de
Re: REF70 new from TI
« Reply #4 on: April 22, 2021, 07:50:28 pm »
It is in a ceramic package.

Hello,

it is pin-compatible to the LTC6655. And the data is also copied from the LT datasheet except for the supply voltage range which is better for the REF7025.
The large question is: how sensitive is the chip in the SMD package to FR4 board stress due to humidity.
With other LS8-packages (LT1236, LTC6655) I have always had not so good results as with CERDIP 8 packages. (AD586LQ)

with best regards

Andreas
Hmm, shooting from the hip but: flex pcb interposer between LS8 and rest of the FR4 PCB?
 
The following users thanked this post: CDN_Torsten

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1128
  • Country: nz
Re: REF70 new from TI
« Reply #5 on: April 22, 2021, 10:38:59 pm »
Quote
With other LS8-packages ... I have always had not so good results as with CERDIP...

LT recommend routing slots around the references, but I guess there’s still some stress to be had in the FR4 right underneath the chip.

I wonder if a ceramic SOIC would be a good compromise? That would be surface mount for reasonable assembly, with legs to take up some of the strain and reduce stress in the package body.
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: REF70 new from TI
« Reply #6 on: April 23, 2021, 04:39:13 am »
I wonder if a ceramic SOIC would be a good compromise?
I fear: not really.
costwise you would need a lead frame from Kovar instead of copper (for the plastic case) and a glass seal. (so no advantage against CERDIP).

For reducing the board stress something like gull wings would be proably a better choice for a SMD package.

with best regards

Andreas
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1128
  • Country: nz
Re: REF70 new from TI
« Reply #7 on: April 23, 2021, 07:05:50 am »
Sorry, I wasn’t clear. I meant a to compare a hypothetical CERSOIC to the LCC. Clearly the CERDIP has already been done.

As you say, carefully forming the CERDIP legs to gull-wing is probably a good way to go.

Edit: the thought of a flex PCB interposer is interesting. I guess the question would be how the polyimide compares to FR4 in terms of matching the LCC package. Maybe it could be slit in creative ways.

Anyhow, good to see someone else offering the LCC packaging.
« Last Edit: April 23, 2021, 07:09:47 am by jbb »
 

Offline ch_scr

  • Frequent Contributor
  • **
  • Posts: 809
  • Country: de
Re: REF70 new from TI
« Reply #8 on: April 23, 2021, 08:25:04 am »
Sorry, I wasn’t clear. I meant a to compare a hypothetical CERSOIC to the LCC. Clearly the CERDIP has already been done.

As you say, carefully forming the CERDIP legs to gull-wing is probably a good way to go.

Edit: the thought of a flex PCB interposer is interesting. I guess the question would be how the polyimide compares to FR4 in terms of matching the LCC package. Maybe it could be slit in creative ways.

Anyhow, good to see someone else offering the LCC packaging.
A quick search shows that polyimide does take up water as well. But I would hazard a guess that the polyimide flex pcb is well a lot more flexible.
The polyimide has to have a degree of "stretchiness" to conform to the forces bending it - my guess would be that as a consequence of that forces on the LCC package would be a lot lower.
Another point where geometry comes into play is: how to arrange the soldered contact between rigid and flex so that the stresses are not only decoupled by stretch but reduced by geometry.
Imagine a flex pcb with LCC + say a decoupling cap. It is now in one case soldered "all around" with contacts on 4 sides. Stresses would only be reduced by stretching of flex.
But when having all contacts to one sides of the flex, stresses from rigid to flex are completely decoupled in one direction and concentrated in the area of the solder joints in the other direction.
Further shenanigans could be played by having a "bend" in the flex pcb, but mechanical assembly would be more cumbersome.
« Last Edit: April 23, 2021, 08:30:12 am by ch_scr »
 

Offline 2N3055

  • Super Contributor
  • ***
  • Posts: 6451
  • Country: hr
Re: REF70 new from TI
« Reply #9 on: April 23, 2021, 10:33:01 am »
There is also a new application note on long term drift:

https://www.ti.com/lit/pdf/sbaa436

Greetings,

Rainer

That accelerated aging method of estimating long term drift was disputed long time ago.

https://www.analog.com/media/en/reference-design-documentation/design-notes/dn229f.pdf

Accelerated aging is a real thing and helps references to settle down quicker.  But it cannot be used to project long term drift from those measurements.
 

Offline kleiner Rainer

  • Regular Contributor
  • *
  • Posts: 108
  • Country: de
  • Rainer DG1SMD JN48ts
Re: REF70 new from TI
« Reply #10 on: April 23, 2021, 10:54:43 am »
I would put a reference in LCC package on a hybrid circuit - no problems with humidity, and the temperature coefficient should be close enough to alleviate stress in the solder joints. Thermal conductivity is also much bettter, you could even print a heating resistor onto the hybrid, together with a SMD NTC to close the loop. Hybrid circuit manufacturing, anyone?

Greetings,

Rainer
 
The following users thanked this post: Andreas

Offline dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: REF70 new from TI
« Reply #11 on: April 23, 2021, 05:08:57 pm »
What i don't understand in the TI paper on LTD: They find that the square root prediction overestimates aging. Why didn't they try to fit an exponential decay? In my own experiments with LTFLU references i found that exponential decay with a time constant of about 7000 hours described the observations very well. The idea behind this would be that the process isn't a random process but a decay of some initial disturbance(s). Like in radioactive decay the disturbance with the longest half life dominates after some time. Another observation i made: "Random" deviations from the exponential decay get smaller during the aging process as well. For me another indication that LTD is less than the square root estimate. TIs logs also look a bit bumpy between 3000 and 4000 h and a bit better afterwards. To be honest, their data look bad considering the use of a 8.5 meter and of a 0.01 °C oil bath.

Regards, Dieter
 
The following users thanked this post: 2N3055

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: REF70 new from TI
« Reply #12 on: April 23, 2021, 08:58:57 pm »
The curve for the LTD look in deed rather noisy, with quite a bit off correlation. Quite some reference noise would be expected for the reference chips, but the correlated part would be more like a problem with the meter or the switches . So there may be a hidden fail of some kind (e.g. using a wrong settting at the meter or EMI problems or cable thermal effects). The test for 85 C looks good though.

The square root law is only one possible case for long term drift. There are other commonly used law to describe aging effects. An exponential decay is one example. There is also the so called streched exponential often used to describe glasses. The epoxy is to a large part in glassy structure.
It really depends on the system how aging proceeds and there can be several contributions superimposed. So predictions are diffcult.

Besides just aging, there can also be an effect of humidity, that can be relatively slow at low temperature. This is especially important for plastic cases.
 

Offline KT88

  • Frequent Contributor
  • **
  • Posts: 322
  • Country: de
Re: REF70 new from TI
« Reply #13 on: April 23, 2021, 09:11:19 pm »
Did you notice the differences distribution and trend of the LDT in comparison to the LTC6655?
The drift has mostly settled after 1kHr with an eyballed median of about +15ppm after 2kHr. The REF70 is still moving after 2kHrs with a median of about -35ppm...

Looks like TI's copy machine has some issues :-DD

Cheers

Andreas
 

Offline tszaboo

  • Super Contributor
  • ***
  • Posts: 7307
  • Country: nl
  • Current job: ATEX product design
Re: REF70 new from TI
« Reply #14 on: April 23, 2021, 10:06:41 pm »
Cheap LTC6655, cool. Although I still dont understand the pricing of precision components. If I buy a crystal oscillator in a 2 pin hermetic package, its 10 cents. If I want VPG to place a resistor in the same package its 20 EUR.

I know, supply, demand... But still, I think it shouldn't be that difficult to place them into hermetic packages.
 

Offline exe

  • Supporter
  • ****
  • Posts: 2559
  • Country: nl
  • self-educated hobbyist
Re: REF70 new from TI
« Reply #15 on: April 23, 2021, 10:12:37 pm »
Why would they copy it? What market is it for? Maxim released max6226 in a similar package, but with incompatible pinout. Not as good as ltc6655, but twice less the price (in single qty), and only a fraction of power consumption. But I haven't seen any reviews of it. I have two of them, but I'm still to put them into service.

Anyway, I kinda like ceramic references and welcome alternatives to ltc6655 which was hard to get. Had to hunt it for months until I got them. Also expensive. Ref70 seems to be a bit cheaper. But I also noticed TI clones not just cheaper, but also have worse specs, often hidden in plots.

PS This IC was discussed a bit here with one guy having hands-on experience with it: https://www.eevblog.com/forum/metrology/ref70-family-voltage-reference/ .
 

Offline KT88

  • Frequent Contributor
  • **
  • Posts: 322
  • Country: de
Re: REF70 new from TI
« Reply #16 on: April 23, 2021, 10:36:07 pm »
Quote
Cheap LTC6655, cool. Although I still dont understand the pricing of precision components. If I buy a crystal oscillator in a 2 pin hermetic package, its 10 cents. If I want VPG to place a resistor in the same package its 20 EUR.
The level of accuracy requires extensive testing -in case of the LTC6655 at least two, if not three temperatures. Test time is an expensive commodity...

Cheers

Andreas
« Last Edit: April 28, 2021, 09:22:06 pm by KT88 »
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: REF70 new from TI
« Reply #17 on: April 24, 2021, 08:23:30 am »
There are plenty of high high resolution SD ADCs offered that need a suitable reference. So there is some need for good 2.5 V or similar references.
For low frequency application one may even like low lower noise, if it would be available.

The long term drift on the Ref70 does no look very good.  Even with the ceramic LCC case it looks like the board may have an effect too. So it may not be just the fault of the chip itself.
The hysteresis part also does not look that good. Still testing -40 to +125 is kind of extreme.

Given the way they describe the noise measurement system, the noise data may be a bit optimistic: 2 filters with 10 Hz cross over in sereis give a cross over a little less than 10 Hz. Still the higher frequency noise looks very good and the lower frequency noise is thus likely not as good as with the LTC6655.
 
The following users thanked this post: exe

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 9410
  • Country: gb
Re: REF70 new from TI
« Reply #18 on: April 24, 2021, 09:39:42 am »
I wonder how long, if ever, it will be before they put a decent burred Zener reference in one of those packages. If it's TI then they've at least got the REF102 available.

I know it's supply and demand but surely somebody must need decent long term stable references, especially as they phase out Cerdip packages.
Best Regards, Chris
 
The following users thanked this post: CDN_Torsten

Offline fcbTopic starter

  • Super Contributor
  • ***
  • Posts: 2117
  • Country: gb
  • Test instrument designer/G1YWC
    • Electron Plus
Re: REF70 new from TI
« Reply #19 on: April 24, 2021, 10:23:03 am »
Do any of the reference vendors sell unpackaged die?

Need an excuse to buy a 2nd hand Kulicke & Soffa.
https://electron.plus Power Analysers, VI Signature Testers, Voltage References, Picoammeters, Curve Tracers.
 

Offline KT88

  • Frequent Contributor
  • **
  • Posts: 322
  • Country: de
Re: REF70 new from TI
« Reply #20 on: April 24, 2021, 11:48:52 am »
They do: https://www.analog.com/en/products/adr225.html?doc=adr225-kgd-chips.pdf.
Search for *voltage reference "known good die"*
TI and others will show up as well.
It won't be a bargain though....

Cheers

Andreas
« Last Edit: April 24, 2021, 11:52:19 am by KT88 »
 
The following users thanked this post: fcb

Offline MiDi

  • Frequent Contributor
  • **
  • Posts: 599
  • Country: ua
Re: REF70 new from TI
« Reply #21 on: April 24, 2021, 12:46:37 pm »
I wonder how long, if ever, it will be before they put a decent burred Zener reference in one of those packages.

LT1236LS8?
 
The following users thanked this post: fcb, Gyro

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 9410
  • Country: gb
Re: REF70 new from TI
« Reply #22 on: April 24, 2021, 01:30:19 pm »
Good catch. The datasheet seems to show a fair bit of thermal hysteresis, maybe that's the penalty of the small ceramic package - although I don't see hysteresis figures specifically stated in the cerdip AD587 datasheet.


BTW, the AD587 is also available in die form. Edit: Make that 'was' (Datasheet Rev. D)

« Last Edit: April 24, 2021, 01:42:08 pm by Gyro »
Best Regards, Chris
 

Offline exe

  • Supporter
  • ****
  • Posts: 2559
  • Country: nl
  • self-educated hobbyist
Re: REF70 new from TI
« Reply #23 on: April 28, 2021, 07:29:06 pm »
One thing I noticed about ref70. Looking at the diagram on https://www.ti.com/product/REF70 , I see it has something called "digital core". I wonder what that is. Can they do some sort of software correction of temperature drift and ageing? Chopping? That could also explain weird 10Hz oscillations reported in another thread (unless that was a measurement error). I wish I could send one to Noopy for decapping and die shots.
 

Offline fcbTopic starter

  • Super Contributor
  • ***
  • Posts: 2117
  • Country: gb
  • Test instrument designer/G1YWC
    • Electron Plus
Re: REF70 new from TI
« Reply #24 on: April 28, 2021, 07:35:35 pm »
One thing I noticed about ref70. Looking at the diagram on https://www.ti.com/product/REF70 , I see it has something called "digital core". I wonder what that is. Can they do some sort of software correction of temperature drift and ageing? Chopping? That could also explain weird 10Hz oscillations reported in another thread (unless that was a measurement error). I wish I could send one to Noopy for decapping and die shots.
Yeah. I wondered what that was too.  That's one of the reasons I got interested in this part.

I think they'd be on a sticky-wicket if they tried to compensate out ageing, but you can compensate out thermal drift (we've just released a low-cost voltage reference that does this).
https://electron.plus Power Analysers, VI Signature Testers, Voltage References, Picoammeters, Curve Tracers.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf