| Message ID | 1565174405-2689-5-git-send-email-yhchuang@realtek.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Kalle Valo |
| Headers | show |
| Series | rtw88: enhance and stabilize T-Put performance | expand |
<yhchuang@realtek.com> wrote: > From: Yan-Hsuan Chuang <yhchuang@realtek.com> > > Add support for doing IQK in firmware > > Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com> The commit log does not answer "Why?". What is IQK and how does it help? No need to resend, I can update the commit log but just let me know what to add.
> > <yhchuang@realtek.com> wrote: > > > From: Yan-Hsuan Chuang <yhchuang@realtek.com> > > > > Add support for doing IQK in firmware > > > > Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com> > > The commit log does not answer "Why?". What is IQK and how does it help? > > No need to resend, I can update the commit log but just let me know what to > add. > Ideally the RF component's I/Q vectors should be orthogonal, but usually they are not. So we need to calibrate for the RF components, ex. PA/LNA, ADC/DAC. And if the I/Q vectors are more orthogonal, the mixed signal will have less deviation. This helps with those rates with higher modulation (MCS8-9), because they have more strict EVM/SNR requirement. Also the better of the quality of the signal, the longer it can propagate, and the better throughput performance we can get. Yan-Hsuan
Tony Chuang <yhchuang@realtek.com> writes: >> >> <yhchuang@realtek.com> wrote: >> >> > From: Yan-Hsuan Chuang <yhchuang@realtek.com> >> > >> > Add support for doing IQK in firmware >> > >> > Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com> >> >> The commit log does not answer "Why?". What is IQK and how does it help? >> >> No need to resend, I can update the commit log but just let me know what to >> add. >> > > Ideally the RF component's I/Q vectors should be orthogonal, but usually they are not. > So we need to calibrate for the RF components, ex. PA/LNA, ADC/DAC. > > And if the I/Q vectors are more orthogonal, the mixed signal will have less deviation. > This helps with those rates with higher modulation (MCS8-9), because they have more > strict EVM/SNR requirement. Also the better of the quality of the signal, the longer it > can propagate, and the better throughput performance we can get. I added this to patch 4 but there were some warnings so I had drop the patches. So please send v2 and remember to add this description to patch 4.
diff --git a/drivers/net/wireless/realtek/rtw88/rtw8822c.c b/drivers/net/wireless/realtek/rtw88/rtw8822c.c index e11bbc3..176ca5f 100644 --- a/drivers/net/wireless/realtek/rtw88/rtw8822c.c +++ b/drivers/net/wireless/realtek/rtw88/rtw8822c.c @@ -1876,6 +1876,22 @@ static void rtw8822c_false_alarm_statistics(struct rtw_dev *rtwdev) static void rtw8822c_do_iqk(struct rtw_dev *rtwdev) { + struct rtw_iqk_para para = {0}; + u8 iqk_chk; + int counter; + + para.clear = 1; + rtw_fw_do_iqk(rtwdev, ¶); + + for (counter = 0; counter < 300; counter++) { + iqk_chk = rtw_read8(rtwdev, REG_RPT_CIP); + if (iqk_chk == 0xaa) + break; + msleep(20); + } + rtw_write8(rtwdev, REG_IQKSTAT, 0x0); + + rtw_dbg(rtwdev, RTW_DBG_RFK, "iqk counter=%d\n", counter); } /* for coex */ diff --git a/drivers/net/wireless/realtek/rtw88/rtw8822c.h b/drivers/net/wireless/realtek/rtw88/rtw8822c.h index 5ee1de4..14a8894 100644 --- a/drivers/net/wireless/realtek/rtw88/rtw8822c.h +++ b/drivers/net/wireless/realtek/rtw88/rtw8822c.h @@ -176,6 +176,7 @@ struct rtw8822c_efuse { #define REG_TXF7 0x1ab0 #define REG_CCK_SOURCE 0x1abc #define BIT_NBI_EN BIT(30) +#define REG_IQKSTAT 0x1b10 #define REG_TXANT 0x1c28 #define REG_ENCCK 0x1c3c #define BIT_CCK_BLK_EN BIT(1) @@ -197,6 +198,7 @@ struct rtw8822c_efuse { #define REG_OFDM_FACNT3 0x2d0c #define REG_OFDM_FACNT4 0x2d10 #define REG_OFDM_FACNT5 0x2d20 +#define REG_RPT_CIP 0x2d9c #define REG_OFDM_TXCNT 0x2de0 #define REG_ORITXCODE2 0x4100 #define REG_3WIRE2 0x410c