The invention relates to a test method for electronic data processing equipment,
in which a memory for holding program and data information
is provided. This is a reliable detection of impending
Errors of memory guaranteed
such computer systems
Errors in the storage of data or program code occur.
Such errors can
already at a new device
occur if the memory modules at least partially defects
exhibit. However, it may also be that over time
of aging phenomena the memory chips more and more defects
exhibit. Furthermore, it happens that certain manufacturing batches
of memory chips are faulty.
Automation technology is widely used in all areas of industry
and also kept in the private sector. Especially in the industrial environment
a memory error can have fatal consequences if e.g. a safety-related
Automation system stops working reliably. In production plants
Production loss also serious consequences, at least financial.
Memory module errors are especially those errors difficult
recognizable, which occur only sporadically. Such non-deterministic
can hardly be detected with a punctual troubleshooting. The reason for this is usually
the operation of the EDP or automation system addressing
the memory is not static, but at least partially dynamic
is so unpredictable which memory areas during the
Operating constantly be used.
It can happen that defective memory cells only sometimes addressed
and therefore a memory error only sometimes occurs.
known methods, for example, the entire memory area
completely checked when switching on or after a reset of the device. Later in the normal
Operation then only the so-called active memory area,
that is the area of the memory in which during the
Operating data is written, checked. For this purpose, e.g. so-called checksum tests
here is that a detected error in the active memory area
usually also leads to an operating error of the device, as just in the active memory area
the processing of the current automation solution takes place. such
Error detections are therefore usually very late or even
to a failure of the device
still to prevent.
The invention therefore has the object of providing a test method for EDP and
Specify automation systems, by means of which memory error as possible
can be recognized
causing a failure of the device
Invention is based on the consideration,
that a test of those memory areas, which does not use
become, valuable conclusions
on the functionality
of the entire memory including the active memory areas
Likelihood can be an unused one with a recognized error
Memory area also on possible
Errors of the used memory area are closed as both
Memory areas are physically realized in the same hardware.
therefore a test procedure for
with at least one memory area, where one currently not
used part of the memory area written test data written
Test data checked and with
Target data are compared to the functionality
to close the memory area.
Inventive test method
the operation of the device
not because it is on the unused part of the memory area
For example, the test data can be bit patterns,
which specifically in the unused part of the memory area
to be written. Such test patterns may be suitably predetermined in advance
to be at a later
Comparison of the written test data with the expected score
good evidence of a possible
present or imminent source of error.
currently unused part of the memory area can parts of memory area
include, which during
the operation of the device
not be used or which may only be used intermittently during operation of the device
the test data will be kept running during
the operation of the device
written to the currently unused part of the memory area.
Hereby, the unused part of the memory area is constantly tested by preferably writing an always different test bit pattern into the area and constantly checking whether the written content meets the expectations. The current writing of the test data should mean that not only unique test data in the unused part of the memory area are written, but that this happens several times during operation.
to bring the comparison to light,
that there is a memory error, that is the written test data
does not match the expected writing result (target data),
so can the device
Nevertheless, continue to be easily operated as the test method of the invention
refers to the unused part of the memory area and thus
the processing of program code and data is not affected.
give such recognized memory problems in the unused part
the memory area hints that sooner or later also in the active area
the memory error could occur.
Such probability statements are a valuable basis
for predictive maintenance.
According to an advantageous embodiment, the memory area comprises a
active and a passive memory area, where the active memory area
that part of the memory area which is provided for this purpose
is, sometime during
of the operation to record and read data. Farther
includes the currently unused part of the memory area
at least part of the passive memory area.
In the invention, the test method is applied to an unused one
Memory area, the passive memory area. This will check if this
unused part of the memory is working properly or whether
there faulty memory cells are present.
Active and the passive memory area are in the same
Hardware realized, with a distinction in active and passive
memory usage is defined by a program. If so
found faulty memory cells in the passive memory area
It may also indicate possible problems in the active memory area
be closed because, as already mentioned, both memory areas
physically not different. For example, aging processes
that some memory cells over time become memory errors
tend. A detection of a memory error in the passive memory area
may therefore be an indication of a possible memory error as well
in the active area, which for
the program execution
is used. Will now be at a memory area in the passive memory area
defective memory cells found by means of the test method according to the invention,
For example, the entire storage area can be renewed,
which can also affect the active memory area, to prevent.
a further preferred embodiment
The currently unused part of the memory area comprises at least
a part of the active memory area, which currently not used
Program runtime often happens
before that, although an active memory area for the execution of the program
is provided, but not at all times, all parts of the active memory area
to be used. The test method according to the invention can now also
such parts of the active memory area are used which
not be used at a current time and therefore
the test method according to the invention
can be used.
this currently unused
Part of the active memory area during the program runtime
can be used once, a detection of a
Memory error in this area a valuable indication
provide that a memory problem could be imminent, which would affect the operation of the
If you look at a detected error fast enough
responds, for example, refreshing the memory area
a failure of the device
be prevented before e.g. a loss of production or a hazard occurs.
or in addition
is it possible, for example,
to define test memory areas between active memory areas
and these as currently unused parts of the memory area
the test method according to the invention
to use. Here, the detection of an optionally
existing memory error in this currently unused part
the memory area between active memory areas a particular
good measure for eventual
imminent memory problems because of a closeness of this currently unused one
Memory part to the active memory area is present, usually in
physically, allowing an overriding of the memory error
to the active memory area is likely.
becomes a memory problem identified by the comparison
prefers an indication of this memory problem to a user
In the case of a number of computer system supported devices, the test method may preferably be applied in parallel to this number and the output hints collected and be evaluated together.
this way it is possible
Correlations between the detected memory problems of different
to determine, for example, using stochastic methods.
So, a prediction about
upcoming memory problems even more accurate.
Therefore, the evaluation of the output hints to a forecast
which indicates an imminent failure of the storage area at least
one of the devices
will be the writing of the test data and the comparison by a processor
and the processor is with the at least one memory area via a
Data bus connected.
In the case of a decentralized arrangement of a plurality of devices, the test method according to the invention can thus have a
Datenbus be made, which the respective devices with the
Processor connects. The data bus is a means of doing this
Distribute test data from the processor to the devices and the memory results
to read out for the
Comparison. So it does not have to be a test procedure separately in each device
but the test procedure is also physical
separate units applicable.
a particularly preferred embodiment
will the test data during
of operation varies, for example, one used as test data
Bit pattern, which is in the currently unused memory area
should be written, constantly changing.
Is it possible,
to predict the upcoming memory error even better, because under
undetected systematic memory errors caused by
a rigid test bit pattern can not be revealed, undetected
Following are two embodiments
closer to the invention
1 a computer-assisted device with RAM and ROM memory for carrying out the method according to the invention, and
2 an example of a memory area comprising a RAM and ROM memory with active and passive memory parts for carrying out the method according to the invention.
should also be representative of erasable ROMs such as
EPROMs, EEPROMs, etc.
In the 1 is a computer-aided device 1 represented, which is a central unit 3 comprising a Prozes sensor and a RAM memory 5 and a ROM memory 7 includes. The central unit 3 , the RAM memory 5 and the ROM memory 7 are here by means of a data bus 9 connected, which can also be designed redundant.
The inventive method is now carried out by means of the central unit 3 over the data bus 9 Test data, for example a test bit pattern, in the RAM memory 5 and / or in the ROM memory 7 to be written. The said memory can also be present several times. In RAM memory 5 it can be a data store and ROM 7 may include, for example, a program memory with the program code.
The central unit 3 again reads via the data bus 9 the memory results resulting from the writing of the test data to the memories 5 and 7 have resulted. The writing of the test data happens in currently unused memory areas of the memory 5 and or 7 ,
Now deviates the readout result from at least one of the memory 5 and or 7 from the original written test data, then a user note 11 which indicates a possible memory problem.
In essence, it is therefore checked here whether test data is error-free in the unused memory area of the memory 5 and or 7 were written. In the case of a negative result of this comparison is by means of the user information 11 a memory problem is displayed.
During operation of the device 11 the test method is preferably performed continuously, wherein the test data is continuously written, read and evaluated in the current unused memory area.
In the 2 is a data store 13 and a program memory 15 a computer-aided device 1 shown, the data memory 13 for example as RAM and the program memory 15 for example, as a ROM.
The data store 13 In this example, it has three active memory areas 17 and two passive memory areas 19 on. The program memory 15 includes four active memory parts 17 so like two passive memory parts 19 ,
For the passive memory areas 19 These may be memory areas that occur during operation of the device 1 not at all or around such memory areas that are not used at any time (in each processing cycle).
By means of the test method according to the invention, the test data now becomes at least one of the passive memory areas 19 at least one of the memories 13 and or 15 written, read out and evaluated.
A defect of one of the passive memory areas 19 , in which test data was written, is expressed by the fact that when reading this passive memory area another data pattern is read than previously written using the test data.
This discrepancy leaves on a memory problem of the memory 13 respectively. 15 shut down.
Since at least currently unused memory areas are used by means of the test method according to the invention, a detected memory problem usually does not (yet) lead to a failure of the device 1 Rather, it is an indication of a potential imminent memory problem. A timely initiated maintenance measure can therefore prevent bad consequences usually targeted.
Furthermore, in the case of an active memory area recognized as defective 17 this recognized by a recognized as error-free passive memory area 19 be replaced. This allows the device to remain in operation despite a memory error.