Forum

I am trying to calculate accurate figures for how long it takes to transmit a frame, however I am getting caught up between timings and size.

For example, in order to transmit a frame with of 200 bytes ( 172 payload + 24 byte 802.11 header info + 4byte FCS), I first convert to bits: 200 x 8 = 1600 bits

802.11 transmits 1 million symbols per second, so am I right in assuming that this would take 1600�s to transmit ?

So, at 1Mbps, I have so far got:

DIFS: 50�s
CWmin: 100�s
Preamble: 192�s
[u]Frame: 1600�s[/u]
Total: 1942 �s needed to transmit this frame.

Comments ?

I also need to work out the same calculations for other PHY's, so if anyone has any resources or figures that would help, please let me know.

Thanks !

Very good question!! I'm very interested in the replies.

of course the contention window will be always a different value

as soon as you have the correct value for 1Mbps data rate, for the others data rates it's just a matter of dividing the value by the new data rate. DIFS values will be different for others modulations

After figuring out how to do all that calculations we could create an excel sheet to facilitate the maths with all the possible variables

This timing will depend upon the type of modulation and use of short/long preambles. Let�s take a look at a specific example: 172 byte 802.11g frame using Short Preambles at 1 Mb

SIFS = 10�s (802.11-2007, 19.8.4)

Short Slot Time = 9�s (802.11-2007, 19.8.4)

DIFS (802.11g with Short Preamble) = 28�s
- This accounts for 2 x Slot Time + SIFS Time (802.11-2007, 9.2.10)

CWMin = 135�s
- CWMin is 15 (802.11-2007, 19.8.4).... 15*9=135

PHY Preamble = 72�s (802.11-2007, 19.3.2.5)

PHY Header = 24�s (802.11-2007, 19.3.2.5)

Payload = MAC Header + Data + FCS = 24 + 172 + 4 = 200 bytes = 1,600 bits = 1,600�s
- 200bytes*8=1,600 bits

SIFS = 10�s (802.11-2007, 19.3.2.3)

ACK Frame = 112�s
- ACK Frame is 14 bytes (802.11-2007, 7.2.1.3)� this is multiplied by 8 to get bits�. 14*8=112bits

Whew� now you get: 28 + 135 + 72 + 24 + 1,600 + 10 + 112 = 1,981�s or 1.981ms

Related to references, @revolutionwifi has the perfect chart here for all SIFS and Slot Times: http://revolutionwifi.blogspot.com/2010/08/wireless-qos-part-4-arbitration.html

Hope this helps out!

Thanks Brian, that's very useful.

So, an 802.11g frame using DSSS/BPSK at 1 Mbps will use a shorter DIFS than an 802.11 prime device at 1Mbps ? I have a source stating 802.11 DIFS is 50�s, you have 28�s, but you quote short preamble where my information states that all 1Mbps frames use long preamble.....

This is where I can't nail the information and it seems to contradict depending on where I read.

1Mbps is 1Mbps, I would have thought that it would not depend on the higher PHY rate that it has downshifted from, if we take your figures then we theoretically get more frames per second if the device is capable of higher PHY's, which doesn't seem to be logical.

But thanks for chipping in, and please continue to do so - treat this like a coffee house round table discussion, anyone welcome to participate.

Thanks again !

By the way, I forgot to add, I don't count the sifs + ack, it's redundant, as I am trying to work out theoretical time to transmit one frame, it does not need to be ack'd, for example, multicast or broadcast.

Brian, I have looked at your calculations and I differ:

[quote]This timing will depend upon the type of modulation and use of short/long preambles. Let�s take a look at a specific example: 172 byte 802.11g frame using Short Preambles at 1 Mb:[/quote]

[i]1MB will always use a long preamble as the short preamble does not support 1 Mbps transmission. The frame is to be transmitted at 1Mbps.[/i]

SIFS = 10�s (802.11-2007, 19.8.4)
[i]Agreed[/i]

[s]Short Slot Time = 9�s (802.11-2007, 19.8.4)[/s] - No
[i]Slot Time = 20�s (HR/DSSS)[/i]

[s]DIFS (802.11g with Short Preamble) = 28�s[/s]
- This accounts for 2 x Slot Time + SIFS Time (802.11-2007, 9.2.10)

[i]DIFS = 50�s[/i]

[s]CWMin = 135�s
- CWMin is 15 (802.11-2007, 19.8.4).... 15*9=135[/s] - No
[i]CWMin is 31 .... 31*20 = 620�s[/i]

[s]PHY Preamble = 72�s (802.11-2007, 19.3.2.5)[/s] - No
[i]PHY Preamble 144�s[/i]

[s]PHY Header = 24�s (802.11-2007, 19.3.2.5)[/s] - No
[i]PLCP Header = 48�s[/i]

Payload = MAC Header + Data + FCS = 24 + 172 + 4 = 200 bytes = 1,600 bits = 1,600�s
- 200bytes*8=1,600 bits - Agreed

Now we have:
50 + 620 + 144 + 48 + 1600 = 2462

So, 2462 �s to transmit this one frame at 1 Mbps.

Thoughts ?

Enesem,

Your post snuck in here before I posted this, but please read on.

For the one of the best explanations on computing xmit time, I suggest you read Chapter 1 of Byron Putnams book [u]802.11 WLAN Hands-On Analysis[/u]. Or since the original CWAP book is available for free on this webiste, I suggest you download a copy and read Chapter 8.

Whether ACK'd or not, the device must wait the DIF's before starting transmission. You are oversimplifying your calculations into meaningless numbers. You also can't base everything on 1 Mbps frames.

Whether you have short or long slots/preambles makes a big difference. I have to deal with these realities all the time in my work. Luckily the Anritsu test set I use, can make all those calculations automatically.

The other element involed is performance of the individual AP or Client. In reality you have to allow for turn-around-time. Assuming that there is no other processing time involved should mean that all devices have the same throughput (or good-put if you prefer).

Obviously this is not the case. There is no way you can equate the performance of a $49 AP with a top of the line Cisco Enterprise, for example, unit.

I have found Excel an excellent tool for mapping the possibilities.

Good luck, and have fun.

Thanks, however I CAN base everything on 1 Mbps frames as in my case I am trying to work out the the times [u]specifically[/u] for 1 Mbps, then work out the equivalent for other frames.

I have a specific need for this.

Long preamble only as the data will be transmitted at 1Mbps.

I have been combing through the resources that you mention, and I think my planning is sound, and I don't agree that converting slot times to time is meaningless - how else do you expect to arrive at an answer ?

Do you not see the challenge here ? I need to know how many frames can be transmitted per second for each modulation type, starting from 1 Mbps, assuming best case scenario.

What's YOUR breakdown of the time taken to transmit one frame described above ?

Fun for all....

Agree with numbers and Long Preambles at 1Mb

Thanks Brian.

CWmin - I don't think this should be in here. I am calculating for streaming data, so unless there has been a detected frame with a Duration Value, I do not believe a CWmin value will be set, so actually should be 0.

So my frame transmit time should actually be 1842 ?

Hmmm......