欢迎访问ic37.com |
会员登录 免费注册
发布采购

PDSP16510AMA 参数 Datasheet PDF下载

PDSP16510AMA图片预览
型号: PDSP16510AMA
PDF下载: 下载PDF文件 查看货源
内容描述: 单机FFT处理器 [Stand Alone FFT Processor]
分类和应用:
文件页数/大小: 23 页 / 97 K
品牌: MITEL [ MITEL NETWORKS CORPORATION ]
 浏览型号PDSP16510AMA的Datasheet PDF文件第8页浏览型号PDSP16510AMA的Datasheet PDF文件第9页浏览型号PDSP16510AMA的Datasheet PDF文件第10页浏览型号PDSP16510AMA的Datasheet PDF文件第11页浏览型号PDSP16510AMA的Datasheet PDF文件第13页浏览型号PDSP16510AMA的Datasheet PDF文件第14页浏览型号PDSP16510AMA的Datasheet PDF文件第15页浏览型号PDSP16510AMA的Datasheet PDF文件第16页  
PDSP16510A MA  
with a complex transform of the same size. The 4 x 64 point  
complex mode becomes an 8 x 64 real mode, but the change  
from 16 x 16 complex transforms to 32 x 16 real transforms is  
not supported.  
When a real transform is performed the algorithm pro-  
duces complex results for each of the incoming data blocks,  
but each result only represents the first half of the frequency  
domain data. This does not cause any loss of information  
since the two halves are mirror images of each other. As with  
complex transforms, it is necessary for a different system  
configuration to be used when 1024 point transforms are  
required. These are considered later, and the following only  
applies to 256 or 64 point transforms.  
Configuration  
Clock Periods  
16 X 16PT  
4 X 64PT  
256PT  
COMP  
COMP  
COMP  
COMP  
REAL  
REAL  
420  
624  
816  
1024PT  
3907  
816  
8 X 64PT  
2 X 256PT  
1032  
4699  
2 X 1024PT REAL  
Table 4. Computation Times in Clock Periods  
In a single device system, performing non overlapped  
transformsondatafromaSINGLE source, onlytheRealinput  
pins are used, and the Imaginary inputs are redundant except  
when configuring the device. By setting Control Register Bits  
8:6 to 101, however, it is possible for a single device to accept  
data from two independent sources using the real and imagi-  
nary inputs. Maximum sampling rates will then only be half  
those possible when a single source is used, if no incoming  
data is to remain un-processed. With two sources a transform  
must be completed in the time to load parallel blocks, other-  
wise incoming data will be lost. With one source a transform  
neednotbefinisheduntiltwodatablockshavebeenacquired.  
In this dual input mode results from data on the real inputs  
always precede those from the imaginary inputs.  
If block overlapping is needed, it is always necessary to load  
pairs of data blocks simultaneously, using both the real and  
imaginary inputs. With dual sources of data this presents no  
problem, and Control Bits 8:6 should be set to 110 or 111 for  
the relevant amount of overlapping. If data is from a single  
source an external FIFO is needed to provide a simple delay  
forablockofdata. Decodes001through100fromControlBits  
8:6 must be used to select the required overlap.  
The output of the FIFO must provide data for the real inputs.  
Continuous inputs can still be accepted, and each block will  
initiallyoccurontheimaginaryinputs, andthenoccuragainon  
the real inputs as an output from the FIFO. The data output  
sequence will consist of the results from a pair of inputs,  
followed by the results obtained after the required overlap.  
Thus with 50% overlapping the sequence is 1 & 2 followed by  
1.5 & 2.5 followed by 3 & 4 followed by 3.5 & 4.5 etc., where  
1 2 3 4 are the sequential inputs to the external FIFO, 1.5 is the  
overlapbetween1and2, and2.5istheoverlapbetween2and  
3.  
When eight simultaneous 64 point transforms are per-  
formed, the sampling rates given in Table 5 assume that data  
is from a common source. The data outputs will be in the  
correct sequence from 1 to 8, corresponding to inputs 1  
through 8 in normal order from a single source. When data is  
from two sources the sampling rates will be halved, and the  
output sequence will be 1A 1B 2A 2B 3A 3B 4A 4B, where A  
and B are the dual simultaneous sources on the real and  
imaginary inputs respectively. If data block overlapping is  
used in either of the above cases, the eight outputs will be  
followed by results from the same basic eight blocks but time  
displaced to give the required overlap. If more than two  
sources are to be handled the user must provide appropriate  
buffering and multiplexing, and the sampling rates must be  
proportionally reduced.  
When two 1024 point transforms are performed with a  
single device, on data from a single source, the input buffer  
must be arranged to acquire two blocks before initialising a  
transfer to the device. In order to improve the maximum  
sampling rates possible, data should be read simultaneously  
from each half of the buffer, and loaded into the real and  
imaginary inputs. This halves the transfer time from the buffer  
to the device, but requires the device to expect dual inputs.  
Thus if block overlapping is not needed Control Register Bits  
8:6 should be set to 101.  
This fast transfer mode is supported by a special option  
on the PDSP16540 Bucket Buffer. It will acquire two 1024  
point non overlapping blocks using the sampling clock, and  
thentransfertheresultstotheFFTprocessoratthefullsystem  
clock rate. Figure 8 shows the system arrangement. It does  
not support block overlapping.  
With1024pointtransformsallblockoverlapsarehandled  
bythebufferlogic, andnotbytheinternalRAM, butthedevice  
must still be programmed to expect the required overlap if the  
external buffer makes use of the in-active LFLG edge to mark  
the overlap point. To achieve the performance given in Table  
5with50%overlaps, thebuffermustprovidesufficientstorage  
for at least 2.5 data blocks. With 75% overlaps it must provide  
storage for 2.75 blocks. This extra storage allows transfers  
between devices to be only needed when a complete new  
block has been acquired for 50% overlaps, and when half a  
new block has been acquired for 75% overlaps.If storage is  
restrictedtotwodatablocks,onlyhalfthesamplingratesgiven  
will be possible. Transfers between devices must then occur  
when a half or a quarter of a new block has been acquired.  
Since the minimum time between transfers must be no less  
than the transform time itself, the sampling rates must be  
proportionally reduced to prevent loss of data.  
Table 5 Sampliing Rates in MHz Obtained with a Single Device and various overlaps.  
12