SPT3.1 CP4Q_FMTA The maximum value?

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

SPT3.1 CP4Q_FMTA The maximum value?

Jump to solution
1,295 Views
Shi_test
Contributor III

I find that the valid range for CP4Q_FMTA is between -32,768 and +28621(6FFF).

When i use spt-tools in matlab to decompression data,if the source data in 6FFF-7FFF will be decompressed to 7F FF.

When i use SPT in S32R45 to decompression data,if the source data in 6FFF-7FFF will be decompressed to 80 00.

So the valid data of CP4Q_FMTA is between -32768 and 28671.I's that right?

Shi_test_0-1701235974372.png

 

0 Kudos
Reply
1 Solution
1,151 Views
yege
NXP Employee
NXP Employee

如果我没理解错,你将连续的4个数作为输入来压缩,在跳变处会导致具体幅值差别,建议你做如下测试:

生成4通道复数数组,幅值范围从0x8_0000到0x7_FFFF (24-bit signed number), 通道间加入随机相位变化,将此数据作为输入,对比解压缩结果。

另外, MATLAB SPT TOOLBOX中包含了很多PDMA使用样例,可以参考。

yege_0-1702611183089.png

压缩算法采用了“Block floating point” 方法,具体细节确实没有在RM中说明。通常来说,选择合适的压缩算法,建议利用MATLAB TOOLBOX工具,结合实际采样的雷达信号,对比不同压缩模式下SNR和动态范围损失对整个系统的影响,并结合SRAM大小,综合考虑选择出合适的模式。

 

View solution in original post

0 Kudos
Reply
6 Replies
1,183 Views
yege
NXP Employee
NXP Employee

Hi Shi,

This was not correctly observed, the issue was probably caused by improper test methods or bugs in test software.

1. the input complex data for compression must have a similar Magnitude, otherwise, significant differences between inputs and outputs will be observed.

2. SPT hardware and MATLAB Toolbox give bit-exact outputs, please check your code and PDMA configurations if differences are seen. 

0 Kudos
Reply
1,169 Views
Shi_test
Contributor III

是这样的,您可以查看图片最左侧的输入数据,当前我是使用了递增的数据进行压缩后再解压的过程,所以不会存在跳变很大的情况。

当前我有两个问题:

第一个,是硬件解压出的结果,在6FFF到7FFF之间被认为成了8000。但SPT TOOL是将其认成了7FFF,这一点SPT TOOL我认为没问题。下列列出了我的SPT压缩核函数与SPT解压核函数(二合一)与原始数据(ADCdataout.dat)。

第二个,是FMTA的压缩算法正常来说真的是将6FFF到7FFF之间的数据压缩成了7FFF吗?

最后我想请问一下,我看到有很多压缩算法,但是却没有相对应的压缩过程说明,以便我确定适合的压缩方式,能否提供压缩过程说明呢?

0 Kudos
Reply
1,152 Views
yege
NXP Employee
NXP Employee

如果我没理解错,你将连续的4个数作为输入来压缩,在跳变处会导致具体幅值差别,建议你做如下测试:

生成4通道复数数组,幅值范围从0x8_0000到0x7_FFFF (24-bit signed number), 通道间加入随机相位变化,将此数据作为输入,对比解压缩结果。

另外, MATLAB SPT TOOLBOX中包含了很多PDMA使用样例,可以参考。

yege_0-1702611183089.png

压缩算法采用了“Block floating point” 方法,具体细节确实没有在RM中说明。通常来说,选择合适的压缩算法,建议利用MATLAB TOOLBOX工具,结合实际采样的雷达信号,对比不同压缩模式下SNR和动态范围损失对整个系统的影响,并结合SRAM大小,综合考虑选择出合适的模式。

 

0 Kudos
Reply
1,018 Views
Shi_test
Contributor III
我理解了,是因为我输出的应该按照24bit输出,而我是按照16bit输出的。谢谢
0 Kudos
Reply
1,266 Views
petervlna
NXP TechSupport
NXP TechSupport

Hello,

I have asked SPT expert for feedback.

He will comment here once he has time.

Best regards,

Peter

0 Kudos
Reply
1,266 Views
petervlna
NXP TechSupport
NXP TechSupport

Hello,

I have asked SPT expert for feedback.

He will comment here once he has time.

Best regards,

Peter

0 Kudos
Reply