Flash memory (EEPROM) reliability in protection relays 1

[Bung]

Does anybody have any thoughts / data or anthing on the reliability of components in modern day digital technology power system protection relays?

I'm particularly interested in flash memory (EEPROM) reliability and electrolytic capacitor reliability.

We have already seen EEPROM problems in two different manufacturers relays (who shall remain nameless - it is the OEM component that is the problem, not the relay itself). Is any particular EEPROM technology better or worse than any other? What affects their reliability - electric fields, mag fields, heat, temperature cycling, number and frequency of re-programs, etc.
Bung
Life is non-linear...

 

[busbar]

These links do not directly address the matter, but do discuss a noteworthy false operation attributed to an EEPROM.

http://www.ameslab.gov/esha/Lessons_Learned/Blue_Alerts/980204b.htm

http://www.nrc.gov/reading-rm/doc-collections/gen-comm/info-notices/1997/in97082.html

 

[Bung]

Thanks, Busbar. We lost 30MW of load when an EEPROM failure caused a busbar strip operation on an 11kV bus recently. Hence my interest in the matter! Bung
Life is non-linear...

 

[Lewish]

The EEPROM wouldn't happen to be from MicroChip would it? We have been having failures from these, and MicroChip refuses to acknowledge the problem. The EEPROM goes blank about 9 months after being programmed. Reprogram them, and they last another 9 months. Then they go blank again.

We have been replacing them with a different brand, and those seem to work fine.

The articles busbar reference are for EPROMs, not EEPROMs.

 

[busbar]

Lewish -- So noted. My apologies for the mistake.

 

[designchain]

So is it a flash EPROM or an EEPROM? They're different technologies - flash is a single transistor with a floating gate; EEPROMs have two transistors per cell and tend to be more reliable (or used to be, anyway. Flash has generally caught up).

Reliability of flash (and EEPROM, for that matter) is affected by temperature, noisy power supplies, number of programming cycles, manufacturing quality, and other factors. If the power rails to the memory aren't clean (how clean are they during switching of this relay?), large enough (usually negative) spikes could cause problems by causing current flow in the substrate.

Eletrolytics caps tend to dry out, affecting their performance. This is accelerated by operation at high temp.

 

[Bung]

Flash EPROM or EEPROM - basically I have NFI. But I suspect it will be flash memory.

The "relay" is essentially just a permanently powered up computer taking in current and voltage analogue signals, doing some fancy footwork inside DSPs, and then issuing a "trip" command if necessary via some small traditional style electromagnetic relays.

Clean power rails could be a challenge in the electrically hostile environment of a substation control room - we just have to rely on the relays being designed with that in mind. But I have seen a relay die from electrostatic discharge when I plugged my laptop into it - and the laptop survived.

Manufacturers tend to get sniffy if you mess with the insides of the failed relays before they have a chance to get their hands on them. Take it out of the case, and they will try to blame the user for doing some damage, and they are not at fault- their relays never fail!. And anyway, we aren't digital electronics experts - I won't necessarily know a flash memory chip from any other kind of chip even if it jumped up and bit me.
Bung
Life is non-linear...

 

[busbar]

Bung — Interesting point; that most often is probably the case. FWIW, one producer informally admits never having charged a nickel for returned, damaged-relay repairs, although there service guys admit that some come in absolutely incinerated, typically from lightning.

 

[johnspark]

Hi everyone, I experienced electrolytic drying out in a relay of one manufacturer back in the 1990s with a relay that came out just before eproms.
Back in those days the company I worked with had time to try to do basic fault finding inside the transistor logic.

In this particular instance the manufacturer admitted this was a fault that they had seen in these relays only once we had pointed it out to them.

cheers

 

[buzzp]

One item to look for when looking for protective relays is yo buy one with CE certification. The tests should include ESD testing as well as immunity from disturbances. Also, it is a good idea to leave all power off before hooking or unhooking any connections (even low voltage).

Lewish, What you stated is disturbing to me since we use EEPROM chips from microchip. I have seen a blanked E^2 in one field return. Would you care to elaborate on a part number? (without checking, ours is 93C46, I believe and 93C66-might have 93 screwed up in PN)

 

[busbar]

buzz — That is good advice. I wonder if that is what happened here.

http://67.115.161.42/pwr/nerc.dawg-96.15brkr.htm

(!)

 

[Bung]

Problem with protection relays is that you can't just turn them off just to plug in your laptop. Nor is it convenient to switch off the laptop when going from one relay to the next (with all the coporate guff on mine, it takes around 45 seconds to boot - a very long time when you are just waiting). I'm afraid that if the equipment dies in such real life practical usage, I just won't buy any more of it. (and I have had one relay die, on the bench luckily, when I plugged my laptop into its comms port).

As for CE or any other certification, it really only means that a lmited number of devices got through a limited number of tests once. Relays live their entire working lives being assailed by every transient nasty a substation can throw at them. The testing only gives a modicum of assurance that they will survive in the longer term.
Bung
Life is non-linear...

 

[dpc]

I have to agree with Bung. Any protective relay that must be de-energized in order to plug into its serial port has serious design problems.

The EEPROM failures are a little scary. I would hope that the relays would have some type of memory testing that would allow it to shut itself down and alarm without mis-operating.

 

[Lewish]

Hi BUZZP, the part number on our parts is: 93C46B/P89R, and the date code is 0032.
What date code do you have?