Bug Fixed-point in LabView?


I am currently working on a project in LabView and met FPGA something weird that looks like a bug, but it could also just be me not well understand how fixed-point numbers are supposed to work. I work in LabView 2010 Pro (32-bit 10.0f2) on Win 7 Ent. 64 bit.

The problem I have is that when I try to add some numbers together I get unexpected answers, which obviously makes the code very unreliable.

For example: If you try to add 0.50002 with - 0.49998, you expect something around zero, but instead, you get like 3.99999... Am I missing something here? I saw this on several numbers around zero. Fixed points do not work around zero?

I should add that I just tried it on the local computer and not on the FPGA, but I don't see that would make a difference. The two entries are updated to 'Adapt to the entered data' and the output is "Adapt to the source", the Add function is also set to "Adapt to the source" and I tried the envelope and saturate... I also tried with fixed bits rather than adapt to the data entered on the inputs etc with no luck.

Can anyone shred some light on this please? Is it just me or is it really a bug or just as fixed-point numbers are supposed to work?

Thank you!

/ Joel

It is not a bug.  If you develop the indicator displays the result, you will see that there is an exponent of-5 at the end (e-5 = 10 ^-5).    So that is to say the answer is 4 * 10 ^-5 or 0.00004, which is pretty close to 0.

Tags: NI Software

Similar Questions

  • Understand the values fixed point of a NOR-9203

    I have a cRIO-9073 with a NI 9203 module. In LabVIEW FPGA, the card values are returned as a fixed point. LabVIEW chosen fixed point properties are (+/-, 21-4) that apparently offers a range of (- 3.125E - 2, 3.125E - 2, 2.98E - 8), and the current card in the range 4-20 MA with a resolution of only 16 - bit. My understanding of fixed point values is only slightly, then someone can explain to me why a 21-bit value is required for the 16 - bit acquisition data?

    Hi Thoric,

    Sorry, it took a while to get a response to you.   The necessary extra bits with fixed point representation are for the overflow feature.  There is more information in the link below.


    I hope that this answer to your question, but if you need more help please let me know.

    Kind regards

    Thomas Clark

  • cRIO wire becomes double, but I use fixed-point

    Hello world

    I've been looking for examples try to understand why a thread that I use is of type double. I thought it was because the number of bits was greater than 64 bits (LabVIEW does not have 64-bit fixed point). Can someone give an idea of why this phenomenon happens?  Look at the pictures below to see what is the wire which causes conflicts. Constants have the value 1 because I tried to use the minimum number of bits and the maturity of the operations were ultimately a superior 64-bit word length. If you can direct me to a link where a person might have a problem like it's ok. Thanks in advance.

    When LabVIEW does not know what digital use, default usually floating-point representation.  This happens in cases like this where shift registers are used and there is a mismatch as fixed-point.  You can work around this by configuring the output data type of multiply that nodes.  Right click on the node are multiplying, select 'Properties', then select the representation fixed-point you want in the tab "output Configuration.   This will enforce the data type.

  • fixed point conversion

    How do I know how LabVIEW manages type conversions and type cast to point fixed.

    In particular, I have two problems:

    (1) assume I want to convert an integer signed 32-bit to a point fixed <+ ,20,20="">with the function "-fixed-point."

    Of course this conversion involves a waste of bits. LabVIEW will keep the most significant bits, or those less important?

    The same problem occurs if I want to convert a fixed point of data (for example <+ ,20,20="">) to another fixed point data with fewer bits (for example <+ ,10,10="">).

    (2) I have a data represented with a point fixed <+ ,20,10="">and I want to cast to a point fixed <+ ,15,5="">by removing 5 whole bits and keeping the sign.

    For example, I have the number - represented 1.5 in <+ ,20,10="">and I want to convert it to <+ ,15,5="">(to be noted that the two representations can express the number correctly).

    How can I do such casting?

    Thank you

    Hi Tom,

    That's a fair criticism, there are probably a few tribal knowledge implicit in the documentation. I'll give some briefing notes that might help a little:

    • LabVIEW makes a distinction between the functions of conversion (aka "balls") and casts. Conversion always functions are trying to preserve the numeric value, with a handful of out-of-range values with different rounding and overflow modes. Conversions integer/integer wrap, while all other combinations round to closest and saturated by default. If the destination is FXP, you can configure the modes.
    • Coercion points have the same behaviors as their corresponding conversion function.
    • Type cast is not supported on the FPGA
    • The fixed point to whole and entire Cast to fixed-point slur was created with use FPGA to mind cases. They are not pure bit casts, because they support extension of sign for an arithmetic scaling, as well as ways of handling overflow. I hope that the new documentation is a bit clearer on their behavior.
    • Number of reinterpret is a little pure cast to numbers FXP or integer, available on digital > palette to comma fixed only under target FPGA. It simplifies the behavior by limiting the length of the input word and output is the same.
  • "bugs" fixed in Captivate 4?

    Questions has a 'bug' fixed Adobe Captivate 4 specifically, when you import Powerpoint slides in Captivate 4 in some cases added Captivate bullet points to a table... It's not good... Also, there are a few problems with the graphics imported into Captivate.  They would be very grainy.  I tried this in Captivate 3 then I used the presenter.  Not added bullets and graphics went well in presenter.  But with the presenter, it's a big challenge for editing audio files.  Raison d'etre, presenter cost number in the order of the audio files, they seem to be random.  This means that you have to hunt and peck to find the right audio file to change to the appropriate slide.

    Also, I get a "Powerpoint has encounterd a serious error in the presenter add-in, you want to turn it off?  I continue to selecting No, but it's annoying.

    Hi Rich

    Adobe Presenter forum is on the link below.

    Click here to see

    See you soon... Rick

    Useful and practical links

    Captivate wish form/Bug report form

    Certified Adobe Captivate training

    SorcerStone blog

    Captivate eBooks

  • Notes crashing with El Capitan 10.11.6. Are updated recent bug fixes.

    Notes crashing with El Capitan 10.11.6. Are updated recent bug fixes. It always sends Apple incident reporting, so I am surprised that they have not yet set. This is a brand new Apple iMac computer, I went for the best to have a trouble-free operation (I hate LAGS). But the computer hangs out sometimes which is disappointing. No one knows what to do to keep my computer running efficiently? In addition, someone at - it a solution to the issue with Notes crashing?

    Try each of them in order. A test to see if the computer works properly:

    1. Reset the PRAM
    2. Reset the controller (SMC) system management
    3. Start the computer in Safe Mode, and then restart normally. It's slower than a standard home.
    4. Repair the disk by booting to the HD recovery immediately after that the chime press the command and R keys until the Utility Menu appears. Choose disk utility and click on the continue button. Select the withdrawal (usually Macintosh HD) volume entry in the list to one side.  Click first aid button in the toolbar. Wait the fact button appear. Disk utility and return in the Utility Menu. Restart the computer from the Apple Menu.
    5. Download and install the update of OS X El Capitan 10.11.6 Combo.
    6. Reinstall OS X by booting from the Recovery HD using the command and R keys. When the Utility Menu appears, select reinstall OS X and then click on the continue button.
  • I get the following message:-Firefox this new version of Firefox contains important security and bug fixes and page offers much improved making it

    I have received several messages inviting me to download the software updates that I use. I think they are attempts to get help with phishing or such. The last "software update available" amplifies - "this new version of Firefox contains an important security and bug fixes, and much improved page rendering times offer." Is generated by anyone at Mozilla, or automatically in Mozilla? Mozilla distributes these types of messages? That Firefox is updated automatically, then these messages should have no useful function except to help with the download of some intrusive camera, which I did not need. Are there plug-ins that I should throw?

    Upgrade to Firefox 35 is currently limited and not everyone gets the new update offered at this time.
    If this is the case for you, then you can manually update via help > topic.

    See also:

  • The fixed Point data

    It is possible to specify the length of Word and whole for the fixed point data format, and if so where this work?

    The module I use is a NI 9205 in a cRIO-9074 and seems to have a fixed point default data format of <+-, 26,="" 5="">, which I interpret as meaning that the data is signed, has a length of 26 Word and a length of whole number of 5. This gives a maximum of 16, a minimum of-16 and a size of minimum step of ~4.768 x 10 ^-7, that is to say 1/2 ^ 2.

    Could someone please confirm that this is the correct interpretation, and also explain how it is possible to obtain a length of 26 Word when the analog digital converter / base is only 16 bits.

    Best regards


  • fixed point math with shift on FPGA register

    It would be easy, but it's on FPGA.

    Essentially, I want to do what is shown, but with fixed Point data. I also need my release of the add function to be <+,32,0>, as she goes into a generation of sine wave VI as the frequency. Basically I want just the frequency to rise slowly to "step size" at certain pre-defined rate. Is it possible to do? There must be a way I can do number table boolean then table of Boolean number, do a few manipulations to the bitwise and convert to the point type fixed I want. But I can't understand it. I also read some fixed point on NI.com unsuccessfully to operate. Frankly, fixed-point math always escapes me.

    Any suggestions?

    Ah, what the trial and error method will do for you.

    It seems to work.

  • So... bug fix?

    So I know Jelly Bean is apparently right around the corner, but what happened to this bug fix that Motorola has promised us? It was what? 8 months since ICS was released on this phone filled with tons of bugs and they have never bothered to solve these problems, many people seemed to have. I know that Jelly Bean will set probably all these questions, but seriously. They knew that we take a ton of questions and not a single bug fix?


    So what bugs you referring exactly? There are a multitude of messages on this forum where ICS was not the issue, but third-party applications or hardware malfunctions. Motorola knows that most of the people (to judge by the responses on this site) had not problems that cannot be corrected with a set of troubleshooting steps. So please let us know what we can help you. And as far as I remember bike never promised a bug fix unless she was needed if I'm wrong someone please correct me! Thank you!

  • Table 2D-fixed points allocated to the amount of disk space!


    I'm trying to create a 2D table file that will be saved on a real-time target. And should be read by my application in real time.

    To create a file as small as possible I have it recorded in fixed point format. (integer 16 bits 6 bits)

    The table size is 1024 X 1024 data points which is ~ 1 M data.

    If I save the file to the DBL format I expect to get 8 MB (each data point is represented by b bytes) of the file - and that's what I

    If I save a U16 format, I expect to get 2 MB file - and that's what I indeed.

    I waited get a 2 MB file with fixed point format-, but as you can see in the screenshot - I did not!

    No explanation why the checkpoint file is fixed so great?

    Thank you

    The FXP is 64-bit internally.

  • 18/09/2012 is no sign yet of ics bug fix, I wonder

    Well, we are in the middle of September, and still no bug don't fix for ics I wonder if the bike is jump a quite bug fix for jelly bean guess is it possible other thoughts on this

    I don't really have problems with mine in all but it seems to be of those who are and it was supposed to be a fix mid-August rumours anyway so I just thought hmmm I wonder if they just push jelly bean instead of a perhaps frozen bug fix bean fixes some of the bugs reported in the first place

  • host vi data only appears in fixed point format


    The FIFO DMA are configured as fixed point integer length 16 and with note 19. When they are acquired by DMA FIFO in FPGA entire watch and mantissa. But when acquired his watch only the integer in host and I lose mantissa. He get rounding upward? Is there a way I could get the accurate data both the mantissa and integer?



    The table of fixed points in your host vi application is defined as: 33 bit word length and 33 whole bit, in this way you will be non-integer values.

    You must change your configuration of fixed point. Right click on the table, select properties and change the parameters in the tab type data of select simply adapt to the source.

    Hope this helps,


  • Signature keys and code updates - bugs fixed and improvements

    Apologies if I put this thread in a wrong forum. Please feel free to move.

    I have a few questions about code signing and approval of the request and I would be really grateful to anyone who could please help me with that.

    1. I bought a single set of keys to sign my .cod files to deploy the portable device of code signing. Should I get an authorization from RIM or BB to deploy my application on a device in the following cases - i) if I want to use BB App store, ii) if I don't want to use BB App store and load the application manually from a PC, iii) if I want to use another installation for deployment outside BB OTA.

    2. for updates in my current application for example bug corrections or improvements, I needed to buy a new set of keys for code signing or those I bought previously would do the job (assuming that I sign my files from the same workstation COD).

    3. If the old code signing keys would be sufficient for the same application, but with some bug fixes and improvements, what all would I need to update either in my camera (if the application is deployed on the device) or on the BB App store.

    4. what happens to jar files? I also need to load jar files in my handheld device to run the application? If so, then I need to update the jar files in the event of bug fixes?

    Please help me on the above questions because my project is in a very critical stage now. I will be really grateful.

    Thanks in advance a ton.

    Your set of keys (files *.csi and sigtool.*) can be used for unlimited applications and for unlimited signatures.

    There is no limitation for you.

    Just make a backup of your key (*.csi and sigtool.* files) and restore JDE/bin folder when necessary.

    This means that all the files of new/updated cod to date could be signed with the same key set?

    You can use the keys that you have already bought for new and updated and you want cod files.

    There is no need to buy these keys again.

  • Equal Server ID &amp; ID Office versions include equal bug fixes?


    I use InDesign Server so much InDesign Desktop.

    Server 11.0 and 11.0 Office have a bug with the processing of scale functionality. In Desktop 11.4 this bug has been fixed.

    Can I be sure that this bug is corrected in 11.4 Server without installing it? Adobe apply these patches of engine in the new version of these two products?

    Thank you


    Hi Oleg,

    unless an employee badges Adobe stops, maybe noone here can answer this question.

    I can't tell if the server and Desktop are still tied with bug fixing.

    HM. As the opening of new versions of the document in earlier versions of InDesign service uses a server behind the stage installation, and if we assume, that this version of the server is the last version available for the clients of the server, maybe you can try to open a new document in the version with an older version to find out what the server. But this only works if the bug can be fixed by open and save a document (IDML) with CC 2015 v11.4.1.102.

    Kind regards

Maybe you are looking for

  • How to remove an old address recipient that fills automatically when I type a name?

    When I start typing a name in the address line, an old pop address in French – how can I permanently delete this address?

  • Why the download my screen is empty

    On all my other computers of firefox, I get a list of docs that I just downloaded, but not on this computer. How can I fix it?

  • Reference to the parallelism - improve update tree times

    Hello Experts! I have a big problem, and I do not know how to solve. I have a huge tree, nodes rely on 40-50. I have an array of object, each object is a node in the tree. I want to update the State of the tree with the objects, but it takes so long.

  • XP SP3 vs. Windows 7; network failures

    I have 2 laptops, xp sp3 and 7 Home Edition and the only options I have to share computer files on the network, I had already set up is by implementing another homegroup that does not allow the computer XP network to join. the fix on the support page

  • Windows Search

    Today I deleted from my computer, Windows Search as the magnifying glass has arrived without notice a couple of weeks, when I've updated windows.  He has been looking for index automatically, and even when I put it on snooze, my computer ran much mor