Is there any known issue with AHCT when operated in between 3.7V to 4 V Vcc?

[aju11]

I am working on a design which where STM32 controller sends in a serial data to SN74AHCT595 at 250kHz,

I am monitoring and comparing the data getting latched in the register (Qa to QH) and the data comming out of the serial register (pin 9) with a known fixed data(101010101010.... or 111111111...) stream sent into the register(Pin 14)

The shift register is powered by a super cap charged to 5V and the controller is powered by 3.3V LDO

Everythig works fine till the ACT595's Vcc is in the range of 5.00V to 4.03V and 3.70V to 3.3V.

Strangely,
When the ACT595 voltage is in between 3.70V to 4.03V the data latched in the register and the streem comming out of the register DOES NOT MATCH. it looks like some data bits get vanished.

Is this a known issue? Is it a noise margin issue? Can this happen if there is an issue in PCB artwork?
Even TI engineers are not able to comment on this!!

Need Help!!

[langwadt]

recommended supply voltage for AHCT is 4.5-5.5V ...

[aju11]

Yes. However,
1) The chop works fine below 3.7V down to 3.3V and
2) NXP chip 74HCT595 works without any issues from 5V down to 3.0V VCC

[Siwastaja]

Outside specified operating voltages, interesting and totally unexpected things can happen - such as what you see, the chip stops working when expected, but starts working again after you violate the specs even more.

You are doing this on purpose and applying an illegal voltage continuously, so you deserve what you get ( ), but sometimes this can hit you in kinda normal operating conditions - like powerup or powerdown!

For example, I have seen an issue where a limited-current supply cannot "boot" an IC (IIRC it was some kind of a voltage reference, don't remember which), simply because while the Vcc is ramping up, there is a tiny range of voltages where the IC consumes tens if not hundreds of times the specified maximum current, so the weak driver was basically "shorted"; and this wasn't an AC effect. This was about tens of milliamps for a small IC normally consuming hundreds of uA max, so I wasn't expecting it. No mention about it in the datasheet.

[TomS_]

Yes. However,

Yes. However, I can drive my car at 100KM/h and not crash, but other people dont seem to be able to do that? 

That is characterisation, and is something you would typically do when designing a product. You test different vendors products to see how they behave in a variety of situations. Perhaps in your situation, you should use the NXP part if it behaves in the more favorable way.

But still, dont expect any guarantees if you exceed the specifications laid out in the datasheet. You may just be lucky in this case, but chips from a different batch may behave entirely differently in the same situation.

[T3sl4co1l]

This is a whole mess of problems, and you have discovered why there is a canonical form that designs follow -- in this case: you don't screw with the supply voltage, you leave it where it needs to be.

The supercap should run a power converter, which delivers a constant 5V or whatever, and that in turn powers the logic. We expect the supercap voltage to vary widely.  This allows it to do so.  This also allows greater extraction of its total energy (by say 10-50%).

And why AHCT when HC would do fine?  Maybe with a level translator.

Tim