Hi Experts, one query.
Regarding HARQ, I want to know why in UL we always has Synchronous HARQ?
And DL is asynchronous.
I know that eNB send in PDCCH that which MCS,RV etc UE has to use in UL.
Is it because when sending in UL eNB knows CQI value report by UE,hence I can allocate which MCS, etc UE has to use in UL?
I think that: asynchronous is more adaptive than synchronous, retransmission data can be sent quickly, anytime, not fixed as synchronous.
But UL HARQ use synchronous, to save DCI resource because UE know that enb will feedback at n+4, enb no need to giant UL DCI again, UE can retrans with same PRB/MCS at n+8.
So, network resends UL DCI only in the case of adaptive retransmissions in LTE?
UL retrans can be adaptive or non adaptive.
In non adaptive, eNB just send ack/nack in phich.
In adaptive, eNB send phich & pdcch ul dci.
Somewhere already explained. Try to search techplayon.
Thanks. I got it there:
This is the only website explaining the reason for 8 HARQ’s.
Also they states enb takes 3ms processing time, eventhough UE also taking 3 ms processing time.
Question is - Enb which is too efficient, will it really take 3ms processing time?? I think reason should be something else. Not sure what. Just trying to figure out if i can find something.
Look at the Yellow highlighted part, please share your thoughts…
Phy take 1 msec and mac harq take 3msec
This is as per spec.
Sounds good, can you tell spec no?
Yes.
36.213
36.221
Synchronous HARQ reduces signalling overhead while asynchronous HARQ increases
flexibility in scheduling.
In adaptive HARQ, modulation, coding scheme, and freq resource allocation may be changed at each
retransmission. In nonadaptive HARQ, retransmissions are performed either by same previous
transmission attributes, or by predefined rules. Adaptive HARQ brings more scheduling gain at the
expense of signalling overhead.
In LTE, DL uses Asynchronous adaptive HARQ, and UL Synchronous (Adaptive/Non Adaptive)
HARQ.
There will be maximum 8 UL and 8DL HARQ PIDs for one HARQ entity. (15 for TDD)
This is the right answer.
Not only depend on processing time, I think it also depend on UE storage capability.
More harq process, UE need to save more data on buffer.
Guys, let me Phrase my question. 
In Asynchronous HARQ eNB replying to UE taking x ms time, where x may b any no.
Can this number be less than 3ms or it should be greater than 3ms then it can be any number?
Like 5ms, 6ms upto max 8 ms?
I am just concerned about enb processing time not UE.
3msec is fixed.
Thank you, so far its clear 3ms is fixed.
Can anyone share the breakdown of this 3ms like here mentioned.
Is it HARQ who is taking this much time or is it something else?
I am not able to find enough info regarding this.
My question is still same here:
Why do enb taking 3ms processing time. eNB is far more capable than a UE. UE also taking 3ms processing time.
So in this case eNB and UE both are taking same processing time i.e 3ms
This is what i am not able to digest, that’s why i am looking for a breakdown statement that why enb taking 3 ms processing time while replying to UE.
As per one of my friend working in Qualcomm this statement is wrong but he is also figuring out the exact answer.
The capability of eNB is scaled based on the cell size which determines the max UEs in a cell and also puts a cap on the max number of UEs which can be scheduled in a TTI by eNB.
This means eNB has a real time enforcement to complete processing for each UE within a certain time limit. (At eNB each layer takes certain processing time, so L1 informs and ticks the L2 two TTI before (~5 symbols before the prior TTI) the actual TTI and within 600-700 us the eNB L2 needs to complete its processing (scheduling) job.
Mind that 8ms RTT is for FDD and in TDD it varies but AFAIK it’s within k+6 TTIs of the original Tx where k is dependent on the TDD frame configuration.
So 3ms is an upper bound on eNB side to finish processing and let UE receive new Tx or reTx in the 4th ms after the last Harq feedback to keep things simple for FDD by 3GPP.
This is right!
3GPP TR 36.912 V14.0.0 (2017-03): Annex B: Latency performance of Rel-8.
This will gave 3ms breakup.
Thank you 
Will look into it, thanks for response.
Remember that eNodeB has to handle many UE’s. eNodeB may or may not take 3ms. This is purely implementation dependent. However eNodeB / UE are expected to provide HARQ feedback at SF+4. SF is the subframe where eNB/UE receives the data.
Trust me it is not a cakewalk for eNodeB to send HARQ feedback within 4ms as each sub system/module (RRH, PHY, MAC), transmission over fronthaul takes its own time.
