Re: areaDetector driver pixel types

[Phil Atkin <phil.atkin@pixelanalytics.co.uk>]

Mark,

That all makes sense, thank you.  I'll check it out tomorrow.

Out of interest, if I use SHORT, how will the scaling/contrast be set by default?  So as to show 0 as black and 65535 as white?

Phil

On 7 Jan 2016, at 18:46, Mark Rivers <rivers@cars.uchicago.edu> wrote:

Hi Phil,

 

Ø  Now, I find that the ImageJ viewer plugin is showing "2000x1500 pixels; 32-bit; 12MB".  Also, the display is now permanently black.

 

I can think of 2 issues that could cause the display to be black

 

-          You have not set EPICS_CA_MAX_ARRAY_BYTES large enough for 12000000 32-bit values, i.e. greater than 48000000 (I’ll explain 32-bits below).  In this case you should be seeing error messages in the ImageJ EPICS_AD_Viewer Plugin window and in the ImageJ Log window.

-          You just need to adjust the brightness and contrast in ImageJ.  Use Image/Adjust/Brightness/Contrast or the Control+Shift+C keyboard shortcut.

 

The fact that ImageJ says the data is 32-bit is due to a limitation of EPICS Channel Access.  You set FTVL=USHORT in your waveform record.  However, Channel Access does not directly support unsigned 16-bit or 32-bit integers, it promotes them to the next larger data type (LONG and DOUBLE respectively).  That is why ImageJ is receiving 32-bit (LONG) data.

 

It works fine to use FTVL=USHORT except that it is inefficient, you are passing twice as much data over Channel Access as is necessary.

 

You can fix this by changing it to FTVL=SHORT.  This seems incorrect, because the data are really unsigned, but you are “casting” them to signed.  That works in this case because ImageJ always treats 16-bit data as unsigned.  This in turn is very counterintuitive because ImageJ is written in Java which does not even support unsigned integers!  But trust me, it works!

 

I just tested this with the simDetector set to collect UInt16 data.  I set the NDStdArrays.template to use either FTVL=USHORT and FTVL=SHORT.  If it was USHORT then the ImageJ data is 32-bits, and the values are correct (in the range 0 to 64K).  If it was SHORT the ImageJ data is 16-bit, but still has the correct range of 0 to 64K.

 

Mark

 

 

 

From: tech-talk-bounces@aps.anl.gov [mailto:tech-talk-bounces@aps.anl.gov] On Behalf Of Phil Atkin
Sent: Thursday, January 07, 2016 11:45 AM
To: tech-talk@aps.anl.gov
Subject: Re: areaDetector driver pixel types

 

Apologies - I made a mistake in my earlier post (below).

Instead of scaling the 16-bit data I generate, the 'something' is simply doing a c-style cast to an 8-bit form on the way to the display.  So 255 displays as peak white; 256 displays as black.

Phil

On 07/01/2016 16:45, Phil Atkin wrote:

Hi,

My camera only generates uint16 pixels, so I have removed the configuration setting and constructor argument.  In the constructor, I set the NDDataType parameter to NDUInt16.

When I debug my code, it seems as though the buffer is being allocated as expected as an unsigned 16-bit buffer and all is well.  However, I then notice to my surprise that the ImageJ viewer plugin is showing "2000x1500 pixels; 8-bit; 2.9MB.  What's more, something is clearly scaling my data from the 16 bits I generate (0..65535) to the 8 bit range of the display (0..255).  Also, I discover that the startup file contains:

dbLoadRecords("NDStdArrays.template", "P=$(PREFIX),R=image1:,PORT=Image1,ADDR=0,TIMEOUT=1,NDARRAY_PORT=$(PORT),TYPE=Int8,FTVL=UCHAR,NELEMENTS=12000000")

That looks wrong, so I change it to the alternative given for 16-bit data in the ADExample script I'm working from:

dbLoadRecords("NDStdArrays.template", "P=$(PREFIX),R=image1:,PORT=Image1,ADDR=0,TIMEOUT=1,NDARRAY_PORT=$(PORT),TYPE=Int16,FTVL=USHORT,NELEMENTS=12000000")

Now, I find that the ImageJ viewer plugin is showing "2000x1500 pixels; 32-bit; 12MB".  Also, the display is now permanently black.

I'm confused; can anyone explain, please?  Thanks,

Phil

 

 

--
<image002.jpg>Pixel Analytics is a limited company registered in England. Company number: 7747526; Registered office: 93A New Road, Haslingfield, Cambridge CB23 1LP

 

--
<image002.jpg>Pixel Analytics is a limited company registered in England. Company number: 7747526; Registered office: 93A New Road, Haslingfield, Cambridge CB23 1LP