GE UR COMTRADE Export: Oscillography Files From a D60 or L90

A GE UR relay writes four different things about a fault, and only one of them is a waveform. The one that most often reaches the engineer analysing the event is not that one. On top of that, the waveform record is usually stored at a quarter of the resolution the relay actually measured, because of a setting nobody revisits after commissioning.

This page covers what the UR family writes, how to identify and retrieve each artifact, and the two configuration details that decide whether the record in your hands can answer the question you have. The specifics quoted are from the GE D60 Line Distance Relay instruction manual and apply across the UR platform, but settings do move between firmware revisions, so confirm against the manual for the relay and version in front of you.

The UR writes to a 1997 working draft, not the 1999 standard

Worth knowing before anything else, because it explains a class of import failures that look like file corruption. The D60 manual, in the file transfer appendix, states that oscillography and data logger files are formatted using the COMTRADE file format per IEEE PC37.111 Draft 7c, dated 02 September 1997.

PC37.111 with a P prefix is a project draft. Draft 7c predates the ratified C37.111-1999 revision that essentially every modern tool is written against. GE is not doing anything irregular here, it documented the dialect it writes and it has stayed consistent with it, but the file you receive is a pre standard dialect of COMTRADE rather than a 1999 file.

The practical consequence is narrow but real. Header fields can differ in ways a strict 1999 validator objects to, and a reader that rejects rather than tolerates will refuse a file whose sample data is entirely sound. If a UR record fails to open in one tool and opens in another, this is usually why, and it is not evidence that the relay produced a bad capture. It is also a reason to be careful with a converter that rewrites the header, since anything that guesses at the missing 1999 semantics is guessing.

Four artifacts, and which one is the waveform

The UR keeps these as separate records with separate retention and separate retrieval. A request for “the fault record” can return any of them.

Note the retention difference. The manual states that only the last ten fault reports are available for retrieval, and that the number in the filename is a counter of how many fault reports have ever occurred, rolling over at 65535. A request for a fault report that has aged out returns a null file rather than an error. Oscillography retention is governed separately, by the number of records setting described below.

The nnn in the oscillography filenames is the trigger number. One quirk to be aware of when scripting against this: the manual itself prints four n characters in the CFG name and three in the DAT name. That inconsistency is in the published document, so if a retrieval script fails on one file of the pair and succeeds on the other, check the digit padding before assuming the record is incomplete.

The stored sample rate is not the measured sample rate

This is the detail that most often decides whether a UR record can support the conclusion you need, and it is set long before the fault.

The UR samples at 64 samples per cycle, always. The manual is explicit that this internal rate is fixed and that the oscillography setting has no effect on the fundamental calculations of the device. What the AC INPUT WAVEFORMS setting controls is the rate at which those samples are stored into the record, and it offers off, 8, 16, 32 or 64 samples per cycle. On the D60 the documented default is 16.

The trade is storage. There is a fixed amount of oscillography memory, so a lower stored rate buys more cycles per record. The manual’s own example table makes the scale of it clear: with two CT and VT modules, 64 digitals and 16 analogs, eight records hold about 93.5 cycles each at 16 samples per cycle, about 57.6 cycles at 32, and about 32.3 cycles at 64.

Whether 16 samples per cycle is enough depends entirely on the question. It is comfortably enough to time an inception and a clearance, to read which phases were involved, and to measure fundamental magnitudes. It gets thin when the question is about waveform shape. Sixteen samples per cycle puts the Nyquist limit at the eighth harmonic before any practical margin, so the higher order content that characterises a sharply saturating CT is not reliably represented, and the detail of a saturated current waveform is coarser than the feature you are trying to see.

The useful consequence: if a site is investigating a recurring instrument transformer or waveform quality problem, raising the stored rate on that relay before the next event costs record length and nothing else. The relay was already measuring at 64.

The setting that erases every record

One line in the manual deserves more attention than its typography gives it. When the NUMBER OF RECORDS oscillography setting is altered, all oscillography records are cleared.

The failure mode writes itself. An event occurs, someone connects to the relay, notices the record buffer is configured shorter than they would like, adjusts it so the next capture is longer, and destroys the record of the event they connected to retrieve. Pull the files first, change the configuration second. This is the same class of hazard as the Oscillography Clear Data register, which wipes the files, zeroes the trigger and available record counters, and stamps an Oscillography Last Cleared Date. The difference is that clear data announces what it is, and number of records does not.

Knowing a new record exists without downloading anything

Three registers describe the state of oscillography storage, and they answer different questions.

• Oscillography Number of Triggers. Increments by one on every capture. The newest file number equals its current value. Poll it and a change means new data is available. This is the register to watch if you are automating collection.
• Oscillography Number of Records. The configured maximum, meaning how many files can be held and, indirectly, how many cycles fit in each.
• Oscillography Available Records. How many files are actually stored and still readable right now. This is the one that tells you whether the record you want still exists.

A gap between the trigger number you were given and what is available is the normal consequence of automatic overwrite mode, in which a new record replaces the oldest. It is not a fault. It is a clock on how long you have to collect after an event, and on a busy feeder with a small record count that clock can be short.

Getting the files off the relay

Retrieval is by name. The manual documents Modbus file transfer in detail, including that the relay tracks read position per connection so files may be read simultaneously on multiple Modbus connections, and states that all the files available via Modbus may also be retrieved using the standard file transfer mechanisms in other protocols, naming TFTP and MMS explicitly. In day to day work this is normally wrapped by the GE EnerVista UR setup software rather than driven by hand.

The data logger accepts a range. Appending a start time, or a start and an end time, as Julian dates in numeric text, meaning seconds since 1 January 1970, limits which records come back. That matters because the whole logger can be a large transfer over a slow substation link when what you needed was an hour either side of an event.

Reading the record once you have it

Two things about a UR record surprise people opening one for the first time.

Channel names describe hardware, not function. The documented convention is slot letter, terminal number, then I or V and the phase, so F2-IB is B phase current on terminal 2 of the CT or VT module in slot F. The fourth current input in a bank is IG and the fourth voltage input is VX. Nothing in the name says which circuit that current belongs to, so matching channels to the single line diagram is a step you do once per relay, using the module layout, and it is worth writing down.

Every CT and VT module channel is in the file. All eight are stored whether or not the fault involved them, so a record routinely contains channels that are simply irrelevant to the event. The analog channels selected in the oscillography settings are metering values recorded as additional traces, and are separate from those waveform channels.

Once the file is open, the record is just COMTRADE and the vendor question stops mattering. The order of operations for working through it is set out in how to read a fault record, and what is a COMTRADE file covers what the CFG and DAT files hold. A UR record and an SEL capture of the same event open side by side in the same COMTRADE viewer, which is what makes a multi ended or multi vendor event analysable at all.

What to ask for

If you are requesting a UR record from site and you intend to analyse the waveform rather than the sequence of operation, ask for the oscillography files, both the CFG and the DAT, by trigger number, and say oscillography rather than fault record so the HTML fault report does not arrive instead. Ask what the AC input waveforms setting is, because that tells you the resolution before you open anything. If the sequence of operation is also in question, the event recorder text file is a separate request and a small file.

And if the event is still fresh and someone is about to reconfigure oscillography on that relay, ask them to collect first.

Frequently asked questions

How do I get a COMTRADE file off a GE D60?

By file transfer, requesting OSCnnnn.CFG and OSCnnn.DAT for a binary record or OSCAnnnn.CFG and OSCAnnn.DAT for text, where the number is the oscillography trigger number. The manual documents this over Modbus file transfer and notes the same files are reachable through TFTP and MMS. The EnerVista UR setup software is the usual front end for it.

Which COMTRADE revision does a GE UR relay write?

The D60 manual specifies IEEE PC37.111 Draft 7c, dated 02 September 1997, for both oscillography and data logger files, in text or binary form. That is a working draft rather than the ratified 1999 revision, which is why a strict 1999 reader can object to a file whose data is fine.

Why is my GE UR record only 16 samples per cycle?

Because the AC input waveforms setting selects the stored rate, and 16 is the documented default on the D60. The relay measures at 64 samples per cycle regardless. The options are 8, 16, 32 and 64, and a lower rate buys more cycles per record out of fixed storage.

What is the difference between oscillography and the data logger?

Oscillography is a triggered waveform capture around an event, at up to the full sampling rate, held for a configured number of records. The data logger is a continuous trend of selected values over a long period. Both come out as COMTRADE, and only oscillography is a fault waveform.

Does changing an oscillography setting delete my records?

Changing the number of records setting clears all oscillography records, which the manual states directly. Writing 1 to the oscillography clear data register also wipes the files, zeroes the trigger and available record counters, and sets the last cleared date. Retrieve before you reconfigure.