![]() |
|
|
|
#1 |
|
Hi, all, I am a beginner in this field.
When I was reading and simulating others' designs, I found many of them have their input and output registered. They use a Flip flop, every rising edge of the clock, they have the outputed signal stored in that flip flip, and then the output of the FF as the final output of the design. I am not quite understand why they want the registered signals, it cause some delay since the final output will wait for the rising edge of the clock. Is it for synchronization or something else? thanks. Alex |
|
|
|
|
#2 |
|
Posts: n/a
|
> I am not quite understand why they want the registered signals, it
> cause some delay since the final output will wait for the rising edge > of the clock. Is it for synchronization or something else? thanks. It is done for timing purposes. As you know, all actions should be completed before the next clock edge. In order to have as much timing margine as possible, designers often clock in their inputs, and clock in their outputs. That way the full clock cycle is available for routing purposes for data transfer between two blocks. Regards, Pieter Hulshoff Pieter Hulshoff |
|
|
|
#3 |
|
Posts: n/a
|
It's often helpful when completing the physical design of a chip. Say there are two blocks, A and B. A has unregistered outputs, B has registered inputs. The timing budget from A to B would be something like: T1 := period(clk) - thold_A - toutput_A - tsetup_B You can't do anything about thold_A or tsetup_B. toutput_A depends on the logic in A between the flop and the output port. Now say we replaced block A with a similar block C that has registered oututs. The budget is: T2 := period(clk) - thold_C - tsetup_B thold_C =~ thold_A Obviously, T2 is likely to be larger than T1. The more time you have between blocks the further they can be from each other or perhaps your clock speed will not be limited by block interconnect delays. You're right though, it adds clock cycles to the path. Sometimes that cannot be tolerated and then you really have to pay attention to the logic in toutput_A and the floorplan. CP In article <>, Alex <> wrote: > Hi, all, I am a beginner in this field. > When I was reading and simulating others' designs, I found many of > them have their input and output registered. They use a Flip flop, > every rising edge of the clock, they have the outputed signal stored > in that flip flip, and then the output of the FF as the final output > of the design. > > I am not quite understand why they want the registered signals, it > cause some delay since the final output will wait for the rising edge > of the clock. Is it for synchronization or something else? thanks. -- 'When religion and politics ride in the same cart, the whirlwind follows.' Chris Pruett |
|
![]() |
| Thread Tools | Search this Thread |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| CISCO 1801 DNS problem | marsav | Hardware | 2 | 07-05-2009 11:41 PM |
| Different types of video output - I'm confused | robotiser@googlemail.com | DVD Video | 2 | 07-29-2007 04:25 PM |
| Post-Route Simulation does not give output for the first clock cycle Options | velocityreviews | Software | 0 | 04-17-2007 05:47 PM |
| Sony Precision Cinema Progressive Output vs Component 480p Output | Otto Pylot | DVD Video | 1 | 04-18-2004 10:49 PM |
| Panasonic S25 DVD player w/o S-video output - will its replacement have S-video? | Mark | DVD Video | 1 | 02-11-2004 04:19 PM |