On Sun, 23 Mar 2014 17:58:45 +0000, Eric Jacobsen wrote:
> On Sun, 23 Mar 2014 00:50:38 -0500, Tim Wescott
> <tim@seemywebsite.really> wrote:
>
>>On Sat, 22 Mar 2014 19:59:54 -0700, fl wrote:
>>
>>> Hi,
>>>
>>> This question is about digital PLL design. The DPLL is in fact a
>>> software PLL according to some PLL's definition. It calculates the
>>> phase error with hardware. This phase error passes through a typical
>>> digital PI filter, arrives at a NCO. I want to design a similar PLL.
>>> Before implementation, I would like to simulate this PLL in Matlab.
>>> The questions are about the phase detector unit and NCO unit.
>>>
>>> The phase detector gain Kd is shown as: 1 unit/rad
>>>
>>> The NCO gain is 20 MHz/unit.
>>>
>>>
>>> This DPLL is in fixed point number. What does 1 unit mean in above two
>>> lines? Is it the whole digital range, or the least bit (LSB)? Or it is
>>> something else?
>>
>>It's whatever the original designer wanted it to be. If the fixed point
>>scheme has a fractional part (i.e., if it's anything but integers), then
>>I would guess that 1 unit is 1. If not, and particularly if the
>>designer was not a native English speaker, your guess is as good as
>>mine.
>>
>>--
>>
>>Tim Wescott Wescott Design Services http://www.wescottdesign.com
>
> What Tim said. I like to us an MSB as a unit for a fixed-point
> implementation, although one can get creative and use many other things.
> The main thing is to be consistent across the analysis.
>
> This may help:
>
> http://www.compdsp.com/presentations/Jacobsen/abineau_dpll_analysis.pdf
It helps that the phase detector is in unit/radian and the NCO is in Hz/
unit -- that means that whatever a "unit" may be, the phase detector gain
times NCO gain is known to be 20MHz/radian. That, in turn, means that as
long as the controller is behaving linearly, the system behavior depends
only on the controller transfer function.
Of course, it's a PLL, and with those the devil is in the (nonlinear)
details.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Reply by Eric Jacobsen●March 23, 20142014-03-23
On Sun, 23 Mar 2014 00:50:38 -0500, Tim Wescott
<tim@seemywebsite.really> wrote:
>On Sat, 22 Mar 2014 19:59:54 -0700, fl wrote:
>
>> Hi,
>>
>> This question is about digital PLL design. The DPLL is in fact a
>> software PLL according to some PLL's definition. It calculates the phase
>> error with hardware. This phase error passes through a typical digital
>> PI filter, arrives at a NCO. I want to design a similar PLL. Before
>> implementation, I would like to simulate this PLL in Matlab. The
>> questions are about the phase detector unit and NCO unit.
>>
>> The phase detector gain Kd is shown as: 1 unit/rad
>>
>> The NCO gain is 20 MHz/unit.
>>
>>
>> This DPLL is in fixed point number. What does 1 unit mean in above two
>> lines? Is it the whole digital range, or the least bit (LSB)? Or it is
>> something else?
>
>It's whatever the original designer wanted it to be. If the fixed point
>scheme has a fractional part (i.e., if it's anything but integers), then
>I would guess that 1 unit is 1. If not, and particularly if the designer
>was not a native English speaker, your guess is as good as mine.
>
>--
>
>Tim Wescott
>Wescott Design Services
>http://www.wescottdesign.com
> Hi,
>
> This question is about digital PLL design. The DPLL is in fact a
> software PLL according to some PLL's definition. It calculates the phase
> error with hardware. This phase error passes through a typical digital
> PI filter, arrives at a NCO. I want to design a similar PLL. Before
> implementation, I would like to simulate this PLL in Matlab. The
> questions are about the phase detector unit and NCO unit.
>
> The phase detector gain Kd is shown as: 1 unit/rad
>
> The NCO gain is 20 MHz/unit.
>
>
> This DPLL is in fixed point number. What does 1 unit mean in above two
> lines? Is it the whole digital range, or the least bit (LSB)? Or it is
> something else?
It's whatever the original designer wanted it to be. If the fixed point
scheme has a fractional part (i.e., if it's anything but integers), then
I would guess that 1 unit is 1. If not, and particularly if the designer
was not a native English speaker, your guess is as good as mine.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Reply by fl●March 22, 20142014-03-22
Hi,
This question is about digital PLL design. The DPLL is in fact a software PLL according to some PLL's definition. It calculates the phase error with hardware. This phase error passes through a typical digital PI filter, arrives at a NCO. I want to design a similar PLL. Before implementation, I would like to simulate this PLL in Matlab. The questions are about the phase detector unit and NCO unit.
The phase detector gain Kd is shown as: 1 unit/rad
The NCO gain is 20 MHz/unit.
This DPLL is in fixed point number. What does 1 unit mean in above two lines? Is it the whole digital range, or the least bit (LSB)? Or it is something else?
Here are the original data:
Parameter Value
Lock-in Range ΔωL 7.5 KHz
Damping Factor ξ 0.707
Natural Frequency ωn 3.3 x 10^4 rad/s
Phase Detector Gain Kd 1 unit/rad
NCO Gain Ko 20 MHz/unit
PI Filter Coefficient C1 2^-11
PI Filter Coefficient C2 2^-14
Thanks for help me.