RE: areaDetector

[Dalesio, L. <da...@bn...>]

That discussion is being had today - very timely. I think you will not see this problem repeated.


-----Original Message-----
From: Mark Rivers [mailto:ri...@ca...]
Sent: Mon 1/23/2012 10:39 PM
To: Greg White; Dalesio, Leo
Cc: James Rowland; EPICS PVData Developers
Subject: RE: areaDetector
 
Greg and Bob,

I'll read this more thoroughly when I have a chance.

But the one issue I see right away is the lack of support for unsigned integer data types.

There are lots of detectors that can produce both signed and unsigned data types. For example, a Roper CCD can produce 16-bit unsigned images when it is sending raw data, or 16-bit signed data if real-time background subtraction is being performed. The existing areaDetector drivers, plugins, and viewers fully support this. Similarly most video-type cameras (e.g. Prosilica) produce unsigned 8-bit images, which NTImage will not directly support. I understand that one could promote a unsigned 8-bit image to signed 16-bit, but one certainly does not want to do this for performance reasons.

The document contains this line:


NOTE: There is no specific support for unsigned types in pvData. * See definition in RESOLUTION of 4-Oct<https://sourceforge.net/tracker/?func=detail&aid=3426964&group_id=249167&atid=2483530>.


I followed that link, but it does not explain the decision not to support unsigned integer types.  I think the lack of support for unsigned types is a serious problem in EPICS V3 CA, and it would be a shame to continue that into V4.


Mark


________________________________________
From: Greg White [gr...@sl...]
Sent: Monday, January 23, 2012 8:34 PM
To: Dalesio, Leo
Cc: Mark Rivers; James Rowland; EPICS PVData Developers
Subject: Re: areaDetector

[Adding epics v4 working group]

Hi Mark, yes please, it would be excellent if you could review it. The document Bob is referring to is
called the EPICS V4 Normative Types [1]. This is a draft for a specification of basic functional data types
for use among EPICS V4 systems, such as a standard image type, a standard histogram data type, a basic
"table of data" etc. In this way, we address the interoperability problem for high level scientific data
in a high performance control system.

See NTArray (5.3) and NTImage (6.7). These respectively are the standard types we're proposing for
generic arrays (not very interesting) and for images (which build on the array type).

Do please bear in mind that this doc is only pre-public draft. Although it's at the unfinished draft
stage, we hope that we have enough for you to be able to comment on our first cut at the NTimage. James
Rowland of Diamond drafted what we have now for NTImage.

For context, the EPICS V4 home page is at [2]. In particular, these build on pvData, which is a typed
data exchange system which allows one to exchange only the changed parts of given structured data
objects between two piers.

Thanks again
Greg

[1] http://epics-pvdata.sourceforge.net/alpha/normativeTypes/normativeTypes.html
[2] http://epics-pvdata.sourceforge.net/

On 24 Jan 2012, at 01:23, Dalesio, Leo wrote:

>
> Yes, it is online. I am asking Greg to send you the document in this email.
> I hope it makes the work you have done better - but having a server available directly to the data.
> Bob
>
> -----Original Message-----
> From: Mark Rivers [mailto:ri...@ca...]
> Sent: Mon 1/23/2012 6:44 PM
> To: Dalesio, Leo; gr...@sl...
> Subject: RE: areaDetector
>
> I think I'm pretty happy with NDArray. I'm not sure how much of that class you are thinking about preserving/supporting? Is there a draft document of what you are thinking about?
>
> Mark
>
>
> From: Dalesio, Leo [mailto:da...@bn...]
> Sent: Monday, January 23, 2012 4:04 PM
> To: Mark Rivers; gr...@sl...
> Subject: areaDetector
>
>
>
> Hi Mark,
> We are formalizing a definition of Normative Types for EPICS version 4. We are strongly supporting what you have done for images and NDArray and thought that it would be a good idea to decide if you are happy with it before we formalize it into the version 4 services we are planning to build on top of it.
> Any advice, support, criticism is welcome.
> Thanks,
> Bob
> ps. Feel free to ask any questions.
>
>