Hi All ,
The following figure shows the INCR4 burst transaction.
Here the address increment is happening in each clk cycle As per AHB protocol Single outstanding address is allowed.
What are the possible adjustment has to be made for this transaction to slave accept the data properly.
Thanks & Regards
Muthuvenkatesh
If a slave drives HREADY low during the data phase of one transfer, this will stall the master's next SEQUENTIAL transfer address phase.
For an example of correct behaviour, refer to one of the first diagrams in whatever version of the AHB protocol you are working from, which will usually show read or write transfers with wait states. You can see that HADDR remains constant when HREADY is low in these simple waited transfer example waveforms.
So it looks like your master is not correctly reacting to HREADY low.
In fact as your master is correctly driving HWDATA, stalled by each HREADY low period, but isn't similarly stalling HADDR, it looks like the master design has problems, with the data phase operations looking correct, but the address phase operation broken.
One other issue is that if this is an INCR4 burst, HTRANS can only signal 3 SEQ transfers. Your waveform shows HTRANS remaining at SEQUENTIAL long after the 4-beat burst of transfers should have ended.
Hi Colin ,
Thanks for your guidance. I have corrected the master address phase latching problem .
Below I mentioned the pic and I mentioned the each transfer ending cycle.
First transfer to address 9000_0000 will complete in 4th clk
Second transfer to address 9000_0004 will complete in 5th clk
Third transfer to address 9000_0008 will complete in 6th clk
Fourth transfer to address 9000_000c will complete in 8th clk
Now Is it as per protocol and once finished INCR4 Burst what value we need to pass in htrans ?
Please confirm the same . I am verifying ahb slave using my real driver (ahb_master_driver)
First transfer data phase completes in 4th cycle and second transfer data phase in the 5th as you say, but the third transfer data phase completes in the 7th cycle (not the 6th), and the fourth transfer in the 8th cycle, again as you said.
An INCR4 burst type means that there will be 4 data transfer cycles; a NONSEQ and 3 SEQ transfers, possibly with some BUSY cycles in between these 4 data transfers.
Once those 4 data transfers have been signalled, HTRANS cannot signal SEQ as that would indicate a 5th transfer in a 4-beat burst.
So when HBURST indicates an INCR4 burst, after the 3rd SEQ transfer address phase ends HTRANS can only signal IDLE (if no new burst is starting) or NONSEQ (if a new burst is starting).
Yes third transfer will completes in the 7th cycle ..
Thanks for your information.
Regards
View all questions in SoC Design forum