[Vichal]

HI Experts,
I know that control signals should be stable through the burst, but it is not clear since when the control signals should be stable. Like in situation when a

Master(M2) is not having the grant and starts requesting the bus access HBUSREQ high (say at time T1), with HTRANS idle.
master then gets the grant at time T2 when last transfer of previous master is started.
and then at next HREADY high which indicates that last transfer of previous master is completed. M2 starts the transaction as indicated by HTRANS NONSEQ (say at time T3)


So my doubt is master should drive HSIZE and HBURST for the transfer at time T1 and keep them stable through the burst, even during the time when the HTRANS was IDLE and then NONSEQ.
OR
Master should drive HSIZE and HBURST for the transfer at time T3 and keep them stable through the burst.

Regards,

[JD]

The control signals need to be stable for the duration of a burst as you say, and the start of the burst is when the master changes HTRANS to indicated NON-SEQ, so T3 in your example.

The change of bus ownership doesn't actually change anything here, HBURST needs to be driven to the value required for the burst as soon as the master (granted or not) indicates the NON-SEQ access.

JD

[Vichal]

Thanks JD,

So do you mean to say that HSIZE and HBURST should be stable from T3 (HTRANS NON_SEQ) to end of burst
and not necessarily from T1(HBUSREQ high) to T3.

Regards,

[JD]

Yes. HBURST is meaningless during IDLE transfers, so it only needs to have a constant value when the actual burst is starting, and that is when HTRANS=NONSEQ, so T3.

Most masters would probably only assert HBUSREQ when they have a data transfer to perform, so I would expect HTRANS to change to NON-SEQ at the same time or very soon after HBUSREQ goes high (at T1). Otherwise the master is requesting the bus, but isn't ready to perform a transfer. However that isn't a protocol requirement, just what I'd expect from a reasonable master.

JD

[Vichal]

Thanks a lot JD for the clarification.