[ath9k-devel] TX power and TPC

[ath9k-devel] TX power and TPC

[Benoit Papillault]

Hello there,

I'm working on EN 301 893 certification. There is a part regarding the 
RF Output Power to make sure that the RF Output Power is within some 
limits. Let's say 27 dBm at 5600 MHz for instance.

In 4.4.1.1, the RF Output Power is defined as the "mean EIRP during a 
transmission burst". This is exactly the RF power at the connector + the 
antenna gain.

So, I started with an AR9280 card / ath9k and hookep up to a Spectrum 
Analyzer with a Channel Power measure. It has been configured with RMS 
detector / Average trace over 100 samples. In order to be close to 100% 
duty cycle, I did a "ping -I wlan0 -b 255.255.255.255 -f -s 1472" (OK, 
this is not perfect, but it has been compared with continuous transmit 
on some other chipset and gives a 0.1 dBm difference).

I also did a Peak power measure using the Peak detector / Max hold trace 
of the spectrum analyzer. I made another measure using a USB power 
sensor over 10MHz - 12.4 GHz. Here are the raw results (iwconfig is the 
reported TX power, BTW, any way to do that with iw ? ) :

  mBm | iwconfig | Peak/Max Hold | RMS/Average | USB Power Sensor
    0 |    3 dBm |      -3.8 dBm |   -14.5 dBm | -14.2 dBm
  100 |    3 dBm |      -3.2 dBm |   -14.4 dBm | -14.1 dBm
  200 |    3 dBm |      -3.3 dBm |   -14.4 dBm | -14.1 dBm
  300 |    3 dBm |      -3.1 dBm |   -14.5 dBm | -14.1 dBm
  400 |    4 dBm |       0.6 dBm |   -11.4 dBm | -10.0 dBm
  500 |    5 dBm |       2.9 dBm |    -9.5 dBm |  -7.9 dBm
  600 |    6 dBm |       4.8 dBm |    -7.8 dBm |  -6.0 dBm
  700 |    7 dBm |       6.2 dBm |    -6.5 dBm |  -4.6 dBm
  800 |    8 dBm |       7.4 dBm |    -5.6 dBm |  -3.5 dBm
  900 |    9 dBm |       8.6 dBm |    -4.6 dBm |  -2.3 dBm
1000 |   10 dBm |       9.7 dBm |    -3.6 dBm |  -1.1 dBm
1100 |   11 dBm |      10.8 dBm |    -2.7 dBm |  -0.1 dBm
1200 |   12 dBm |      12.0 dBm |    -2.0 dBm |   0.9 dBm
1300 |   13 dBm |      12.7 dBm |    -1.2 dBm |   1.6 dBm
1400 |   14 dBm |      13.5 dBm |    -0.2 dBm |   2.8 dBm
1500 |   15 dBm |      15.0 dBm |     1.0 dBm |   3.9 dBm
1600 |   16 dBm |      16.3 dBm |     2.2 dBm |   5.2 dBm
1700 |   17 dBm |      18.1 dBm |     2.8 dBm |   6.1 dBm
1800 |   18 dBm |      18.4 dBm |     3.5 dBm |   6.9 dBm
1900 |   19 dBm |      19.7 dBm |     4.2 dBm |   7.6 dBm
2000 |   20 dBm |      20.1 dBm |     5.2 dBm |   8.8 dBm
2100 |   21 dBm |      21.0 dBm |     6.0 dBm |   9.7 dBm

I am not 100% confident in the way I did all those measures. If you know 
a better way to achieve them, please stand up.

Questions are :

1. since regulatory limitation are based on "mean EIRP over burst" ... 
why the driver is calibrated over "peak power" ? We are just loosing 11 
dBm or so.

2. What would explain the difference between the RMS power measured from 
the SA and the power measured by the power sensor ?

Regards,
Benoit

[ath9k-devel] TX power and TPC

[Holger Schurig]

To your question 2: are both the spectrum analyzer and the USB power
sensor calibrated?

Also, do all components have matching impedance?

Also, do they all use short coax cable that don't eat up much of the
power?  Do you know the cable type?  If a cheap cable is used, then
even a short piece of cable or a cheap plug can reduce the signal
considerable at 5 GHz. For example, RG 58 CU (which no one uses for
GHz things) has 100 dB loss for 100 meters at 2 GHz. Aircell 5  has
81.18 dB loss at 100 meter at 5 GHz. And so on.

[ath9k-devel] TX power and TPC

[Benoit Papillault]

Le 12/06/2012 11:28, Holger Schurig a ?crit :
> To your question 2: are both the spectrum analyzer and the USB power
> sensor calibrated?

The spectrum analyzer was bought with calibration certifications.
The USB power sensor has been Zero Calibrated once before all measures.

> Also, do all components have matching impedance?

Sure. 50 Ohms here.

>
> Also, do they all use short coax cable that don't eat up much of the
> power?  Do you know the cable type?  If a cheap cable is used, then
> even a short piece of cable or a cheap plug can reduce the signal
> considerable at 5 GHz. For example, RG 58 CU (which no one uses for
> GHz things) has 100 dB loss for 100 meters at 2 GHz. Aircell 5  has
> 81.18 dB loss at 100 meter at 5 GHz. And so on.
>

The same cable (2m, LMR-200 I guess) was used for both type of measures. 
No attenuators was used (the SA has some internal attenuators and the 
power sensor range is up to +20 dBm).

Regards,
Benoit

[ath9k-devel] TX power and TPC

[Holger Schurig]

Okay, that means the most obvious things are dealt with.

One question: for the SA, you said that it has a RMS detector. You
didn't say something like this for your USB probe. Can it be the case
that the USB probe is only designed for sine waves, i.e. is a
non-RMS-device?  In that case you'd need a low-pass-filter.

One (practical) web site that illustrates the differences between
unfiltered and filtered measurements in a practical way is
http://www.qrp.pops.net/RF-workbench-1.asp