Best regards.
Siyu
2017-06-08 21:41 GMT+08:00 Bastian Bloessl <address@hidden
<mailto:address@hidden>>:
Hi,
On 06/07/2017 02:32 PM, zhan siyu wrote:
Thanks for your reply. Let me explain what I 'm doing.
I have two B210s connected with two computers. I want to measure
the throughput between the two computers over the usrp
connection over gr-ieee802-11. But no matter how hard I try,
like tuning the parameters and turning off my own wifi card and
AP, I can only get 150K B/s, which should be around 300K B/s if
my theoretical calculation is right. There are no underrun or
overrun errors. The throughput measurement tool I'm using is
iperf, which is an application to measure the end to end throughput.
Could you give me some hints?
I'm afraid I can't help a lot, since I have no idea what iperf is
doing. I guess it floods the network stack with UDP packets. The GNU
Radio transceiver, however, has no back pressure mechanism, so it
will just drop the frames when they are injected into the GNU Radio
flow graph.
Also, I don't understand what exact analytical number you are trying
to reproduce here. In fact, I have no idea what this experiment is
supposed to show/prove or what part of the transceiver you are
trying to profile. If, at all, this experiment will tell you how
fast the TX-side can produce frames.
This was investigated in
Gonzalo Arcos, Rodrigo Ferreri, Matías Richart, Pablo Ezzatti and
Eduardo Grampín, "Accelerating an IEEE 802.11 a/g/p Transceiver in
GNU Radio," Proceedings of 9th Latin America Networking Conference
(LANC'16), Valparaíso, Chile, October 2016, pp. 13-19.
Turns out, you can improve the TX-side by improving the OFDM Carrier
Allocator.
Best,
Bastian
Best regards.
Siyu
2017-06-07 14:02 GMT+08:00 Bastian Bloessl <address@hidden
<mailto:address@hidden> <mailto:address@hidden
<mailto:address@hidden>>>:
Hi,
On 06/07/2017 03:04 AM, zhan siyu wrote:
Thanks. I just wonder why. Because I meet some performance
problem. I thought it maybe caused by my
misconfiguration of the
gr-ieee802-11 code. Now, it seems not.
I'm a bit confused why the fact that the transceiver is not
configured through iwconfig ruled out any configuration
issues, but
great that all seems to be set up now.
However, theoretically, as my current sample rate is
10M and
BPSK. So the coding rate should be 10M/2 = 5M b/s. The
throughtput should be around 5M/8 = 625K B/s. Assuming
the 12%
head cost, so the data throughput should be 625 * 88 %
= 550K
B/s. But as my experiment shows, the throughput is
only 150K B/s.
I'm new to the communication. Is my calculation right ?
BPSK 1/3 is 3Mbit/s gross at 10MHz. The overhead per packet
has to
be subtracted, i.e. the actual maximum rate depends on the
frame size.
If it were right, then what might cause the gap?
Since you don't explain what you are doing, this is very
hard to
tell. You would reach this theoretical throughput only if
you send
frames back-to-back (which probably only works if you
pregenerate
the sample stream). But also a WiFi card will insert
inter-frame
space, so that the actual throughput will not match the
theoretical
maximum physical layer throughput.
Best,
Bastian
One more question, I didn't run the volk_profile. Does
it matter?
Best regards.
Siyu
2017-06-07 4:23 GMT+08:00 Bastian Bloessl
<address@hidden <mailto:address@hidden>
<mailto:address@hidden <mailto:address@hidden>>
<mailto:address@hidden <mailto:address@hidden>
<mailto:address@hidden <mailto:address@hidden>>>>:
Hi,
On 06/06/2017 03:55 PM, zhan siyu wrote:
Hi all,
I just found I can't use the iwconfig tap0
rate 20M to
setup the
bandwidth of the tap0. The error message is :
Error for wireless request "Set Bit Rate" (8B20) :
SET failed on device tap0 ;
Operation not
supported.
But in their video , it can be set in this
way. May I
know how
to solve it ?
The WiFi transceiver is attached to the tun/tap
interface,
which is
a virtual Ethernet device. This device doesn't support
WiFi-specific
configuration through iwconfig.
If you wanted this level of integration, you would
have to
write a
kernel module that attaches the transceiver to a
virtual
WiFi card.
Some group already did that, but they didn't
release the
source code.
Best,
Bastian