We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
Gents;
Not a direct Keil question. I have picked up a project that has 2 serial FRAM (FM25V20-GTR) in the design.
I have read the specs and data sheets. The specman-ship make them sound the greatest thing since the pentode valve.
I would like to know what real world experiences might point to gotchas that I need watch for in terms both of software and hardware.
The design has good reverse termination on the serial lines and the devices have good by-pass decoupling. So I am assuming good SPI program design should make the FRAM devices quite happy.
Yes, I will be using Keil tools in the project.
Any and all comments are welcomed.
Thank you,
Bradford
We encountered some kind of FRAM problem on 2011. It was not related to SW, nor HW. The problem was that Ramtron failed to deliver the FRAM we had used for a long time for our production. I believed that the root-cause was Ramtron changed his foundry supplier.
www.eetimes.com/.../FRAM-customers-mad-over-IBM-foundry-delays
So we were forced to buy any FRAM from the market, finally we used Fujitsu's equivalent FRAM.
(We use I2C FRAM, not SPI FRAM.)
Hi John;
Thanks for the feedback and the link. Quite interesting. I just googled Digikey and Mouser and Avnet. All have stock so Ramtron may have resolved their Fab issues.
.... serial FRAM .... I have read the specs and data sheets. The specman-ship make them sound the greatest thing since the pentode valve.
just like EEPROM. Last I used them was as a direct replacement for EEPROMS, I recall no SW or HW change (except, of course, the chip)
the only reason I see to use FRAM over 'regular' EEPROM would be 'wear', if 'wear' is not a concern, the EEPROM is just as good and a lot cheapre.
Erik
Hi Erik;
Thanks for the feedback. Especially about the no problems. Yes, wear will be a concern for this application so the original designer seems to have made the correct choice.
I am getting a warm fuzzy feeling the FRAM is what the specs seem to claim.
The IBM factory had big issues getting the production certified. Think TI had to step up their production to make sure there would be enough Ramtron memories available.
Hi Per;
Thanks for the comments. That's eactly what the link that John posted indicated. But I just looked around at some of my vendors and they seem to have stock. This is not a large project at this time so large quantities are not required.
I have had a datasheet on my desk for a few months for the 5v versions.
It looks to be directly compatible with standard E2 for reads and writes provided you use the correct number of address bytes. They don't seem to support bulk erase commands.
Writes can probably be sped up by changing your page size loop mask to the "full address range" and eliminating status register polling and / or associated delays.
The E2 chips I am using take 2-5 mS to write after the CS goes high from what I remember.
I should get my hands on one of these ... it is on the todo list.
Chad
Hi Chad;
Thanks for the feedback. This device (FM25V20-GTR) is a 3.3V device. It has life of 10E14 read/write cycles. I don't think I'll test that very much :-).
Unlike Serial Flash, the FM25V20 performs write operations at bus speed. No write delays are incurred. Data is written to the memory array immediately after it has been transferred to the device.
A quote from the specs. Yes, after a status check, CS must go back high prior to the next cycle but otherwise there is no delay other than the serial write time at 40MHz. There is no other write delay. There are no erase commands. Also, the device supports only Mode 0 and Mode 3 SPI protocols.
I understand your address concerns. The full 18 bit address selects an individual byte.
All writes to the memory array begin with a WREN op-code. The next op-code is the WRITE instruction. This op-code is followed by a three-byte address value, which specifies the 18-bit address of the first data byte of the write operation. Subsequent bytes are data and they are written sequentially. Addresses are incremented internally as long as the bus master continues to issue clocks.
Again from the spec sheet. Read works the same way except one dummy byte to prime the pump so to speak. I believe most serial memories support the same feature. 'Fast Read' is supported.
I will need to spend a bit more time with the data sheet before I write the world's greatest FRAM control functions.