ATFX DataFinder

[PeterMtu]

Hello,

currently I have following problem with ATFX files (ASAM ODS) and the DataFinder. The ATFX data contains several matrices for different sampling rates. The DataFinder can only create an index for those files that contain only one matrix. If I would use DIAdem and ASAM Data Service, I will have a similar problem when the import mode is set to "Default". When the import mode is set to "SubMatrix" then I can load ATFX data. So I think there is a similar problem with the DataFinder.

Is there a way to tell the DataFinder to use the import mode "SubMatrix" when creating indices for ATFX files? I am using the "ATFX Example" plugin.

Btw: why is the ATFX plugin always disabled (and the checkbox greyed) but it is possible to install the "ATFX example" plugin? What is the difference?

Thanks and regards,

Peter

[AndreasK]

Hello Peter,

• In refreence to the attached document submatrixes are normally not intended to be loaded as DIAdem group but they are defined to be part of the group. So a bunch of submatrixes are intended to be used to establish one group with channels.
  We added the submatrix load mode because a lot of customers began to use the submatries as groups or just wanted to load the different sample frequences without establish one big value matrix.

• Datafinder only works on Datamodels that relate to the TDM Base model (root - group - channel).
  
• ATFX is much more powerful and can work on any ASAM ODS model. The ATFX example is an very simple example that does the following model mapping:
  - aotest -> tdm_root
  ... loop over all aosubtests and ignore them
  - aomeasurement -> tdm_channelgroup
  - aomeasurementquantity -> tdm_channel
  

So now we can get to your use case.

The atf example is an so called DataPluginPlugin.

Its work is to be an model mapper.

So in your case it should map submatrix to tdm_channelgroup

and localcolumn to tdm_channel. I am not completely sure if this

is possible but I wanted to give you some background and

add some additional information later.

Greetings from sunny Aachen

Andreas

The Datafiner

[AndreasK]

Hallo Peter,

in the actual ASAM spezification it is possible to have one or none independent measurementquantity. If there is none one measurementquantity is only allowed to b represented in a single submatrix. If this wouldmatch your use case there is no longer a difference in loading with default or submatrix mode and you could use your files in Datafinder like in all other places.

 g1

 | - sm1 5kHz

 | |     | - time1

 | |     | - values5

 | | - sm1 10kHz

 | |     | - time2

 | |     | - values10

 | | - sm1 20kHz

 | |     | - time3

 | |     | - values20

None of the measurmentquantities is marked as independent. In kind of ASAM ODS channel names on one measuremnt has to be unique. So its not alowed to name all of the time channels time. Choosing this solution will disable the ability to build up the value matrix. The channels will be loaded as they are with different length. I just mention this solution because most time people using the submatrix mode normally doing this becaus the stored different timimg systems into a single measurment and they are often not ntersted in loding a value matrix or didn't even know about the value matrix.

!!!! the independent channel is not the X-channl of a submatrix !!!

It is the sorting quantity to establish the value matrix.

So if it is not intended to work with a value matrix it is fine to mark none of the channels as independent.

Greetings from sunny Aachen

Andreas

P.S.: I am still looking for a solution to establish a dataplugin that is able to use the submatrix loading mechanism but currentl I believe the only way is a DIAdem load event

[PeterMtu]

Hello Andreas,

sorry I marked your last message as a solution, but it was just done by accident.

I checked the ASAM ODS specification 5.1.1 (hoping it is still valid). According to chapter 4.4.6 it is said for the attribute  "name" of local columns: "Name of the instance, only unique within its predecessor in a hierarchy". From my point of view, columns of different submatrices are not predecessors. The predecessors are the submatrices. For the flag "independent" it is said "not more than one local column per submatrix may be independent".

For us it is necessary to have these channels to build a large virtual matrix with holes. In chapter 4.4.4 it is proposed to use a measurement quantity "time" to merge to submatrices. This would allow different names for the time columns. Will DIAdem use the measurement quantity information to link matrices?

Thank you for your explanations. It helps me to understand how DIAdem is working.

Greeting from Munich,

Peter

[AndreasK · Accepted solution]

🙂

Seem like we getting to the center of it.

 Can you provide an example file?

I will attach the file that corresponds to the documentation file.

• each submatrix is allowed to have one independent.
• the local column should be named like its measurement quantity
• if there are multible submatrices containing a local column being marked as independent all independent localcolumns have to be connected to a single measurement quantity.

If all those things are given DIAdem will build up the value matrix if the ATFX file is loaded in the default mode.

Greetings

Andreas

[AndreasK]

Hello Peter,

Some additional information about the names.

DIAdem does not really care about the name uniqueness.

I just mentioned this because if the names are not choosen unique it may cause trouble in contact with an ASAM ODS Server

because some of the API methods are name based.

The localcolumns are special. They are very close related to the measurementquantity. localcolumn and related

measurementquantityname should be equal.DIAdem will also notcare about this name relation but in work with an

ASAM ODS server it causes errors too.

Because of this connection localcolumn names are only as free as the measurementquantity name.

So this matches the entry of the ASAM Documentation given by you.

Actually the ASAM ODS checker does not check the bulk contest. So it will not complain about the independent stuff but

it should complain about the uniqueness of names.

Greetings

Andreas

[PeterMtu]

Hi Andreas,

thank you for your help and explanations. Currently there are multiple instances of measurement quantity with the name "time", because there is a one-to-one relationship between columns and measurement quantities. In the future there will be only one measurement quantity for all independent columns. This is reasonable and with your explanation I could understand the specification better.

The NI support proposed my to create TDM files. So now I am writing an ATFX-TDM converter. So in the future there will be one binary file with bulk data and two description files. One ATFX file (we need this anyway) and one TDM file. So maybe I will have problems with TDM files in the future. 😉

It was also interesting what you have wrote about the ASAM API. I just knew that they exist but I was not aware that column names are used as identifiers. But in our target sites all column names would be unique - except the independent time columns.

Thanks and regards,
Peter

[PeterMtu]

Hi Andreas,

thank you to your help. I can now create ATFX files that are indexed by the DataFinder. The data is looking reasonable - so the problem is really solved. I think it is not necessary anymore to upload some sample files.

Danke,

Peter