Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[STDS-802-11-TGBE] discussions on 11-21/1229r3

Hi Xiaogang,

 

Thanks for the discussion yesterday. Here are the points I did not get chance to address fully regarding adding the subfield “TB Sounding Feedback Rate Limit “ to PHY Capabilities element (11-21/1229r3):

  1. During the PHY meeting you mentioned the efficiency issue using different methods.  However, to deal with STAs with limited capability (e.g., slow process in PHY à MACàPHY), additional overhead is needed by either using padding on TF (existing method in 11ax) or setting low data rate on TB PPDU (new proposal in 11-21/1229r3). As indicated in 802.11be D1.2, the solicited TB PPDUs have to finish at the same time. Hence, the duration of the solicited TB PPDU is upper bounded by the lowest data rate of the solicited TB PPDU.  The following excerpts from 802.11be D1.2 supports my points
    1. P111L39, “The UL Length subfield of the Common Info field indicates the value of the L-SIG LENGTH field of the solicited TB PPDU.”
    2. P379L1, “A non-AP EHT STA that responds to a Trigger frame that solicits an EHT TB PPDU shall set the TXVECTOR parameters below as follows:

                                                               i.      ...

                                                             ii.      —The L_LENGTH parameter is set to the value indicated by the UL Length subfield in the Common Info field of the Trigger frame.”

    1. P507L14, “The TXVECTOR parameter L_LENGTH field of an EHT TB PPDU has the same value as the UL Length subfield of a Trigger frame.”

 

Therefore,  the proposal for CID 5770 and 5779 in 11-21/1229r3 is not more efficient than 11ax existing solutions or the solutions with increased padding duration.

 

 

  1. There are existing solutions in 802.11ax to provide extra process time via adding padding field. Below is one example of the implementation:
    1. The PHY -> MAC processing can start right after the reception of NDP, i.e., starting at t1 time as indicated below

 

 

    1. MAC-> PHY processing can start even earlier than t2 if there is a padding field in Trigger frame, i.e., MAC->PHY can start right after the detection of padding field.

 

Best,

Zinan

 

 

From: Zinan Lin
Sent: Monday, October 11, 2021 4:31 PM
To: Chen, Xiaogang C <xiaogang.c.chen@xxxxxxxxx>; STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
Subject: RE: question on 11-21/1229r2

 

Hi Xiaogang,

 

Thanks for your response!

 

My inline comments are given below.

 

Thanks,

Zinan

 

From: Chen, Xiaogang C <xiaogang.c.chen@xxxxxxxxx>
Sent: Monday, October 11, 2021 11:34 AM
To: Zinan Lin <Zinan.Lin@xxxxxxxxxxxxxxxx>; STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
Subject: RE: question on 11-21/1229r2

 

External Mail

Hi Zinan,

You are right about TF padding, but there is an upper bound for the padding duration and it’s used to full fill TF processing which is different with the concern of the implementation in the CR. for non-TB sounding, STA can chose a rate that fit its hardware limitation such that the STA could response within SIFS. This is exactly the point to add a capability for TB sounding where AP can poll with a high rate which cannot be full filled within SIFS.

 

[ZN]: Yes, you are right, there is an upper bound for the padding duration. But as what is indicated below (Table 9-322a in 802.11ax D8.0, Subfields of the HE MAC Capabilities Information field), the upper bound can be increased by using the reserved value of MinTrigProcTime.  I do not understand why fulfilling TF processing is different from the concern of the implementation in the CR mentioned in your email. Could you please elaborate more?

 

For both TB sounding and non TB sounding sequence, the STA can start to process CSI feedback right after the reception of NDP. In other words, PHY-> MAC process can be started right after the NDP is received regardless of the existence of the Trigger frame. The problem left is how fast the STA can finish MAC->PHY, which maybe similar to general TB based transmissions. If the padding duration is long enough in Trigger frame, it should be sufficient for non-AP STA to finish MAC->PHY even when the data rate of TB PPDU is high.

 

Please note 802.11ax D8.0 (P371L36) gives the solutions to satisfy MinTrigProcTime requirement, “An AP may use any type of padding to satisfy the MinTrigProcTime requirement of a non-AP STA, such as using the Padding field in a Trigger frame, post-EOF A-MPDU padding, or aggregating other MPDUs in the A-MPDU.”

 

 

 

For capability element, IMO, it’s better to keep PHY related ones within PHY capability element.

[ZN] As indicated in 1229r2, the concern of implementation involves “PHY->MAC->PHY”. It is not a purely PHY layer problem and is related to MAC layer as well. Therefore, using the reserved bit in EHT MAC Capabilities may be a better choice.

 

BRs,

Xiaogang.

 

From: Zinan Lin <Zinan.Lin@xxxxxxxxxxxxxxxx>
Sent: Thursday, October 7, 2021 8:22 AM
To: Chen, Xiaogang C <xiaogang.c.chen@xxxxxxxxx>; STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
Subject: question on 11-21/1229r2

 

Hi Xiaogang,

 

Thank you for presenting 11-21/1229r2 in the last week’s Joint meeting. It is an interesting topic.

 

Regarding CID 5770, I am still confused by the motivation of adding TB Sounding Feedback Rate Limit subfield in the EHT PHY Capabilities element.

a.       I am not sure why the STA has to indicate its TB Sounding Feedback Rate Limit capability to the AP

                                       i.      For TB-based sounding, non-AP STAs can start process CSI feedback after NDP frame. Other solutions are available for AP to provide extra processing time for non-AP STAs. For example, in 11ax, there is a Padding field in Trigger frame (Figure 9-64a—Trigger frame format and 26.5.2.2.3 Padding for a triggering frame in 802.11ax D8.0) ) and User Info field with AID12 4095 as padding field (Table 9-29h—AID12 subfield encoding). Both of these can be used to give the non-AP STA more processing time.

                                     ii.      For non-TB based sounding, STA can choose the MCS/Nss to transmit the feedback. Note the STA needs to start processing CSI feedback and responding right after NDP within SIFS time in this case.

 

b.       There are only two reserved bits remained in EHT PHY Capabilities element. Why it has to use the precious bit in EHT PHY Capabilities element to indicate the TB Sounding Feedback Rate Limit? 

                                       i.      There are 11 reserved bits in EHT MAC Capabilities element. Is it better to use the reserved bit in EHT MAC Capabilities element if this capability has to be indicated to the AP?

 

 

Best regards,

Zinan

 

To unsubscribe from the STDS-802-11-TGBE list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1