#1 Dec. 28, 2017 14:15:11

prix
Registered: 2015-09-09
Posts: 12
Profile   Send e-mail  

LWB5 performance with iperf

Hello,
we use LWB5 on our custom iMX6Q Linux board.
We do 3.3V signaling with SDIO bus at 50MHz (High Speed).
This should give us bus speed of around 25MB/s max.

However when we measure LWB5 performance with iperf we get just around ~60Mbit/s (7.5MB/s).
We are connected with 802.11AC enabled access point (AP). All other 802.11AC enabled clients do around 300 Mbit/s on the same AP.

Comparison:

prix@workstation # iperf -s
————————————————————
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
————————————————————
local 192.168.0.100 port 5001 connected with 192.168.0.102 port 60409
Interval Transfer Bandwidth
0.0-60.0 sec 439 MBytes 61.3 Mbits/sec

local 192.168.0.100 port 5001 connected with 192.168.0.101 port 57182
0.0-60.0 sec 2.77 GBytes 396 Mbits/sec

where:
- 192.168.0.101 is Samsung Galaxy S7
- 192.168.0.102 is Linux board with LWB5


Debug output from Linux+LWB5:

# cat /sys/kernel/debug/mmc1/ios
clock: 50000000 Hz
actual clock: 49500000 Hz
vdd: 21 (3.3 ~ 3.4 V)
bus mode: 2 (push-pull)
chip select: 0 (don't care)
power mode: 2 (on)
bus width: 2 (4 bits)
timing spec: 2 (sd high-speed)
signal voltage: 0 (3.30 V)

# iw dev wlan0 link
Connected to 18:a6:f7:7f:25:42 (on wlan0)
SSID: TP-LINK_2542_5G
freq: 5220
RX: 13054521 bytes (8479 packets)
TX: 2104659 bytes (26468 packets)
signal: -37 dBm
tx bitrate: 433.3 MBit/s
bss flags: short-slot-time
dtim period: 1
beacon int: 100

Did you do any performance tests? What are expected results?
How to troubleshoot?

Regards,
Primoz

Edited prix (Dec. 28, 2017 14:45:58)

Offline

#2 Dec. 28, 2017 19:06:08

SupportEngineer
Registered: 2016-03-11
Posts: 61
Profile   Send e-mail  

LWB5 performance with iperf

Hi Primoz,

Given the information you have provided, this would agree with our own findings of being platform related. We have seen a throughput limitation of roughly 60Mbit/s over TCP using iMX6 platforms.

Please feel free to contact us with additional questions, or concerns.

Regards,

SupportEngineer

Offline

#3 Jan. 3, 2018 07:03:15

prix
Registered: 2015-09-09
Posts: 12
Profile   Send e-mail  

LWB5 performance with iperf

Hi,

thanks for your prompt answer, however I do have some additional questions and concerns.
So, is this expected result at 3.3V@50MHz or can one achieve better throughput by using 1.8V@100MHz or 200MHz on iMX6?
I cannot simply test 1.8V signaling on SDIO bus since this would yield significant board redesign.

Did you test with 1.8V or 3.3V signaling and achieved 60Mbit/s?

Thanks and regards,
Primoz

Edited prix (Jan. 3, 2018 07:04:27)

Offline

#4 Jan. 4, 2018 20:20:02

SupportEngineer
Registered: 2016-03-11
Posts: 61
Profile   Send e-mail  

LWB5 performance with iperf

Hi Primoz,

The recommended voltages per the Sterling-LWB5 datasheet (see link) on pages 25 and 44 are 3.3V VBATT and 1.8V VDDIO. The test was performed at 50MHz on our test platform. We have no further information regarding other conditions at this time.

https://www.lsr.com/downloads/products/330-0208.pdf

Please feel free to contact us with additional questions, or concerns.

Regards,

SupportEngineer

Offline

#5 Jan. 5, 2018 07:03:12

prix
Registered: 2015-09-09
Posts: 12
Profile   Send e-mail  

LWB5 performance with iperf

Hi,

yes, we also use 3.3V on VBATT and 1.8V on VDDIO. But those 1.8V signals are achieved by utilizing bidirectional voltage level translator (TXS0108E) between iMX6Q and LWB5. So my terminology or expression “3.3V signaling” implies we only use 3.3V signaling on SDIO bus and in terms of possible bus speeds.
Since iMX6Q doesn't have internal circuit to switch between 1.8V or 3.3V on SDIO, one must provide additional external circuit if 1.8V signaling is required for faster bus speed modes. For example UHS-I modes all require 1.8V signaling!

However even High Speed (HS) mode (3.3V@50MHz) should be able to achieve 25MB/s. But with LWB5 and iMX6Q we are still far from those figures…?

Do you plan to test LWB5 on iMX6 with other bus speed modes (SDR50, SDR104, DDR50, etc)? Can one even expect better throughput?

Regards,
Primoz

Offline

#6 Jan. 5, 2018 22:43:39

SupportEngineer
Registered: 2016-03-11
Posts: 61
Profile   Send e-mail  

LWB5 performance with iperf

Hi Primoz,

You are correct on the 1.8V signaling to attain UHS-I mode. This is what was tested and there are currently no plans on further testing. Seeming that this is a platform related issue, the support at NXP might be able to provide another angle to help you come to a solution.

Your text to link here…

Please feel free to contact us with additional questions, or concerns.

Regards,

SupportEngineer

Offline

Board footer

Moderator control