PCF2127 RTC RAM does not store data

quex · ‎08-09-2019

I have a problem with one PCF2127 RTC IC.

The clock works, I can set and read the time via I2C.

The other feature I'm using is the backed-up RAM.

But when I write the registers and read them back, I get only zeros.

The PCF2127 responds, but the values are zero.

It looks like this:

Writing 5 bytes starting at address 0:

Reading the 5 bytes back:

On another unit (the same device type, identical MCU application) it behaves correctly:

Writing 5 bytes starting at address 0:

Reading the 5 bytes back:

I also tried writing 5 bytes @ address 64, but with the same result.

The schematics of the circuit:

It seems to me that the PCF2127 is defective.

Is there any procedure that could be used to test/identify the problem or restore the correct function?

Has this behavior been observed before?

quex · ‎08-15-2019

Hi Tomáš,

thank you for your reply and sorry for the delay.

So far I had six devices in my hands, two were bought some months ago for prototypes and four belong to a new batch of 16. The 16 devices are soldered, but so far I have commissioned only 4 of them. All are from Digikey, so I do not expect them to be counterfeit. I observed the behavior only on one device.
The two from the first batch work, two from the new batch were sent to a customer without testing of this feature (An automatic test only checks whether the RTC communicates and if the time runs and is retained. An automatic check of the RAM was not implemented) one is faulty (that was found out after delivering the first two) and one which I tested after discovering the problem works as expected.
That leaves 12 more to be tested, not counting the two already delivered which would inevitably return later...
Image of the faulty device is attached. The PCBs got a layer of protective varnish, which makes the markings a little harder to read. And it gathered some dust during testing.

Best regards,

Michal

TomasVaverka · ‎08-19-2019

Hi Michal,

Thanks for your answers.

I am afraid that it will be necessary to perform a detailed analysis of faulty parts by our quality engineers in their lab. Our standard procedure for this is to submit a CQI (Customer Quality Incident) request through a distributor you bought the parts from (DigiKey).

Best regards,

Tomas

quex · ‎09-29-2019

Hello Tomas,

please send me information to the procedure of submitting the CQI.

Is NXP involved from the beginning or does the distributor contact NXP?

I will have to have the part de-soldered.

Best regards,

Michal

TomasVaverka · ‎08-12-2019

Hi Michal,

It sounds like a quality issue. I do have a few questions to let us start:

1. How many parts have you tested so far and on how many of them have you observed this behavior?

2. Are all tested parts from the same batch?

3. What is the complete marking found on the top side of tested parts?

Best regards,

Tomas

dave_harmonjr · ‎09-14-2020

Did this ever get resolved? I am experiencing the same issue with the exact same chips that we bought around the same time frame. Approximately 35% of our chips the RAM does not work at all, but every other feature works. Same exact boards, circuits, firmware etc...

BOWUx · ‎03-03-2021

Hello, we have 4 faulty tested pieces from 52 ICs from 2 batches,

which we bought at Farnell over 2 years ago.

The difference to the problem described here is, that we don't know, when the RAM get into this state (RTC works further very well).
And, when we discharge our connected GoldCap capacitor at VBAT (which unfortunately takes about 2 weeks in the closed device) we have full access to the memory again.

I adjusted the latter on the laboratory bench and unloaded the GoldCap. Then I can write to the RAM again and read back the exact content.

We have the following 2 batches:

Batch: 180817000009

48 pieces of PCF2127T

ACV098.

1 19th

knD18192

and

Batch: 180531000055

4pieces = PCF2127T

ACT659.

1 59

knD18072

Because, the order is more than 2 years old, we have received the answer from Farnell, that NXP will not make an CQC failure analysis.

Was the RAM of the PCF2127 not usable in all of these from the first start of device and permanently?

Does somebody has any idea?

Best regards

tkhansen · ‎02-17-2021

We experience something similar, so I'm interested in knowing if you reached a solution or conclusion?

dave_harmonjr · ‎02-17-2021

In my case I had a bad batch of IC's and simply replacing them with a new batch fixed it.

tkhansen · ‎02-17-2021

Thanks for getting back so quickly. Can you remember the bath codes (week, year etc.)?

In out case these boards where produced last year and we didn't catch anything in production but have received complaints from costumers so it could be we have used components from the same batch.

dave_harmonjr · ‎02-17-2021

This is what is written on one of the faulty IC's.

"PCF2127T

AFF 173

1 05

knD19202"

We had probably a 80+% fail rate on all the RTC's we got from this batch.

After we bought a new batch 100% have worked.

tkhansen · ‎02-17-2021

Thank you very much, I'll reach out to our production to hear if they can give me the batch codes for the failing parts.

BOWUx · ‎02-10-2021

I have the same question, 
because we have the same problem. 

Was there always a workaround, 
or are the ICs defective and have to be replaced?

NXPRalf · ‎06-25-2021

At the moment i do not have a problem with the Ram, but i have many problems with the acuracy of the frequency.
The data sheet says +-3 ppm. But our chips of the year 2019 have more than 19 ppm. If i understood our customer, they have several days per year time differences (> 100 ppm).

The batch code is:
ACU4771.1 05
knE18142