|
Hi all, I am facing problems in simulating interrupts for ADSP2184 in VisualDSP++. I have written service routine, & i am forcing interrupt by using raise() function. raise function is working properly, but program is not taking interrupt routine. What i am supposed to do ? Does anybody have code for the same ? Thanks in advance Kiran __________________________________ |
|
|
Help required
Started by ●June 4, 2003
Reply by ●June 4, 20032003-06-04
|
Folks, I am trying to implement a low pass FIR filter in Visual DSP for the 2181 processor (Am very new to signal processing). I used the signal processing toolbox in Matlab to generate coefficients of the required filter. (its a 844 order filter) and then exported these coefficients to a file and then read in the coefficients in Visual DSP. The coefficients I obtained are in floating point format and some in fixed point format. Can u tell me how to convert these to the 1.15 point representation (the 16-bit representation with 1 sign bit and the others 2^(-1), 2^(-2) ..2^(-15)). Do I do this conversion in Visual DSP or in Matlab itself? I have an 800-sample input signal which is a sum of two sinusoids. I need to filter out one of the signals. Thanks in advance, Hans |
|
|
Reply by ●June 5, 20032003-06-05
|
If your coefficients are in decimal, (range + - 1) just put an "r"
after each number like this: .00564856r .00045751r -.12545654r The assembler converts them to 1.15. It took me hours to find that out. Dave David Tiefenbrunn N1WWY Essential Telecommunications Corp. - http://www.essentialtel.com David Tiefenbrunn - http://users.abac.com/dandatief -----Original Message----- From: [mailto:] Sent: Wednesday, June 04, 2003 7:49 AM To: Subject: [adsp] coversion between data formats Folks, I am trying to implement a low pass FIR filter in Visual DSP for the 2181 processor [snip] The coefficients I obtained are in floating point format and some in fixed point format. Can u tell me how to convert these to the 1.15 point representation |
|
|
Reply by ●June 6, 20032003-06-06
|
Hi, As part of an assignment using ADSP2181 in Visual DSP, I am required to create 2 sinusoids of different frequencies using a lookup table. Heres the exact words "Write a macro that creates a sinusoid of amplitude A, lookup table cycle c, with DAG pointer and address modifier as i/p parameters." I generated the sinusoids using the "sin" function, but the lecturer prefers creating it from a lookup table. What does lookup table mean? and how do we generate a signal using that. Any hints anyone? Thanks, Hans |
|
|
Reply by ●June 6, 20032003-06-06
|
Hi Dave, Thanks for replying to my query. Yes, my coeffs are in decmal. I tried the sugestion with something like " hz[0] = 0.2134r;", but the compiler comlains abt the syntax. Clearly, it was expecting an ';' instead of the'r' at the end. Do I have to set any other compiler options for this to work? Without the 'r', if I just input the decimal (double), like say the value '0.5', it gets stored in the dm as 3F00 0000. What format is this? In 1.15 format, I would expect 4000 0000. Thanks, Hans Quoting Dave Tiefenbrunn <>: > If your coefficients are in decimal, (range + - 1) just put an "r" after each > number like this: > .00564856r > .00045751r > -.12545654r > > The assembler converts them to 1.15. It took me hours to find that out. > > Dave > > David Tiefenbrunn N1WWY > Essential Telecommunications Corp. - http://www.essentialtel.com > David Tiefenbrunn - http://users.abac.com/dandatief > > -----Original Message----- > From: [mailto:] > Sent: Wednesday, June 04, 2003 7:49 AM > To: > Subject: [adsp] coversion between data formats > Folks, > I am trying to implement a low pass FIR filter in Visual DSP for the 2181 > > processor > [snip] > > The coefficients I obtained are in floating > point format and some in fixed point format. Can u tell me how to convert > these > to the 1.15 point representation > |
|
|
Reply by ●June 6, 20032003-06-06
|
Hi..its me again... :) Another itsy bitsy doubt..i noticed that when a "double arra[]" was stored in the data memory, each element of the array took up 4 bytes, but when stored in the program memory, each element took up 6 bytes. Heres the actual snippet of the test code and the corresponding memory dump: double dm arr1[3] = {0.5, -0.5, 0.25}; arr1 (in data memory) [000006] 3F00 0000 BF00 [000009] 0000 3E80 0000 double pm arr2[3] = {0.5, -0.5, 0.25}; arr2 (in program memory) [0037BB] 3F0000 000000 [0037BD] BF0000 000000 [0037BF] 3E8000 000000 Is there a reason for this diffrence? Thanks, Hans Quoting : > Hi Dave, > Thanks for replying to my query. Yes, my coeffs are in decmal. I tried the > sugestion with something like " hz[0] = 0.2134r;", but the compiler comlains > > abt the syntax. Clearly, it was expecting an ';' instead of the'r' at the > end. > Do I have to set any other compiler options for this to work? > Without the 'r', if I just input the decimal (double), like say the > value '0.5', it gets stored in the dm as 3F00 0000. What format is this? In > 1.15 format, I would expect 4000 0000. > Thanks, > Hans > > Quoting Dave Tiefenbrunn <>: > > > If your coefficients are in decimal, (range + - 1) just put an "r" after > each > > number like this: > > .00564856r > > .00045751r > > -.12545654r > > > > The assembler converts them to 1.15. It took me hours to find that out. > > > > Dave > > > > David Tiefenbrunn N1WWY > > Essential Telecommunications Corp. - http://www.essentialtel.com > > David Tiefenbrunn - http://users.abac.com/dandatief > > > > > > > > -----Original Message----- > > From: [mailto:] > > Sent: Wednesday, June 04, 2003 7:49 AM > > To: > > Subject: [adsp] coversion between data formats > > > > > > Folks, > > I am trying to implement a low pass FIR filter in Visual DSP for the > 2181 > > > > processor > > [snip] > > > > The coefficients I obtained are in floating > > point format and some in fixed point format. Can u tell me how to convert > > these > > to the 1.15 point representation > > > |
Reply by ●June 6, 20032003-06-06
|
>From: >To: Dave Tiefenbrunn <> >CC: >Subject: RE: [adsp] coversion between data formats >Date: Fri, 6 Jun 2003 20:44:45 +1200 > >Hi Dave, > Thanks for replying to my query. Yes, my coeffs are in decmal. I tried >the >sugestion with something like " hz[0] = 0.2134r;", but the compiler >comlains >abt the syntax. Clearly, it was expecting an ';' instead of the'r' at the >end. >Do I have to set any other compiler options for this to work? > Without the 'r', if I just input the decimal (double), like say the >value '0.5', it gets stored in the dm as 3F00 0000. What format is this? In >1.15 format, I would expect 4000 0000. >Thanks, > You will find something called EZForum for old 2181 EZ-Kit Lite on the ADI web site. That contains a utility, TODAT.com, written by Dwight Elvey that converts a file containing FP values into 1.15 format. Leon -- Leon Heller, G1HSM Tel: +44 1424 423947 Email: My web page: http://www.geocities.com/leon_heller _________________________________________________________________ Sign-up for a FREE BT Broadband connection today! http://www.msn.co.uk/specials/btbroadband |
Reply by ●June 6, 20032003-06-06
|
On Fri, 6 Jun 2003 wrote: > Hi, > As part of an assignment using ADSP2181 in Visual DSP, I am required to > create 2 sinusoids of different frequencies using a lookup table. Heres the > exact words "Write a macro that creates a sinusoid of amplitude A, lookup table > cycle c, with DAG pointer and address modifier as i/p parameters." > I generated the sinusoids using the "sin" function, but the lecturer prefers > creating it from a lookup table. > What does lookup table mean? and how do we generate a signal using that. Any > hints anyone? You missed a day in class eh? A lookup table is just a list of values in memory. You use the address as an input and the value you have stored there is the output. It's really simple. So for a sine table, say of 16 values you'd have offset value 0 sin(0) 1 sin(pi/16) 2 sin(2*pi/16) : : 15 sin(15*pi/16) Which only gives half a sine wave (can you see why?) Your prof wants you to build a table, and then use indirect addressing to look up values in the table. Use a circular DAG, and a full table, and you can generate a sinisoid really easily. Patience, persistence, truth, Dr. mike |
Reply by ●June 9, 20032003-06-09
|
Hans wrote: > Thanks for replying to my query. Yes, my coeffs are in decmal. I tried the >sugestion with something like " hz[0] = 0.2134r;", but the compiler comlains >abt the syntax. Sorry, I didn't include all the details. Define the variable like this: #define Filterlength0; .VAR/CIRC FilterCoeffents[FilterLength] = "Filter 100 taps.txt"; The coefficients are in the file "Filter 100 taps.txt" in the format: .00215542r .00045522r ... An alternative is: .VAR/CIRC FilterCoeffents[2] = .0001245r, -.0044452r; Dave David Tiefenbrunn N1WWY Essential Telecommunications Corp. - http://www.essentialtel.com David Tiefenbrunn - http://users.abac.com/dandatief |
