[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] Two-Tone Test Results (Transmitting)

From: David Bengtson
Subject: Re: [Discuss-gnuradio] Two-Tone Test Results (Transmitting)
Date: Wed, 19 Sep 2012 22:30:14 -0400

Comment's interspersed below.

On Sun, Sep 16, 2012 at 8:51 AM, Michael Hill <address@hidden> wrote:
Hi guys,

Thanks for the responses! Again, let me point out i am quite new to alot of the IIP3 concepts and still have my areas of confusion regarding the board's operation, so if I say anything weird or counter-intuitive please correct me!

To start with some info from Dave's queries:

1- Thats right.. these are the sorts of things i'm looking to characterise. I'm not sure if this would be considered some fault with my USRP/daughterboard or within the expected variance in behaviour for the daughterboards :(

so a starting point would be to do a single tone measurement, sweeping the gain. That would let you make sure that the gain is relativly linear and figure out the relationship between the gain steps and output power. You can also see how bad the DAC spurs are with the transmitter, something that's important. to understand.

2. An example shot is attached. I'm using a real Spec Amp. Thats right regarding the tones.. I noticed that too. However one thing that might be confusing about this test is that i'm not directly increasing the power by 1dB at a time (in which case I'd expect the 2F1-F2 and 2F2-F1 tones to increase by 3dB..) im changing the output power by varying the gain.. (rather then power directly). So that seemingly would impact the IIP3 values..(e.g. im gathering this would explain why my 2F1-F2 and 2F2-F1 tones aren't increasing by a 3:1 ratio..)

 In general, the most accurate IP3 measurements are when the IP3 tones are only 3 to 6 dB above the noise floor. Doing this lets you be assured that you're not measuring anything else, (IP5 etc). averaging is also a good idea, as that reduces noise in the measurement. You're measuring with a 100 Hz RBW/VBW, you might be better served by increasing that to 1 or 3 kHz and trace averaging. I'd also reduce the output power until the IP3 tones are 6 dB above the noise floor, and use that single measurement to calculate your OIP3. As your output power increases, Higher order intermodulation products will start to come up at the same frequency and give you a bad measurement, something that's no longer OIP3.

3. Will that affect my measurements for the 2F1-F2 and 2F2-F1 tones though for this calculation?

Yes, IP5/IP7 tones can start showing up at higher output levels, close to compression. If you widen up the bandwidth of the measurement to go out to catch the IP5/7 tones, I bet you'll see them with your current measurement setup. 


1. (Re: OIP3) thanks thats a good point.. o_0 I should adjust that.

2. (Re: Useful) But knowing the OIP3 will surely still give me an indication of when my transmitter will start showing non-linearities... won't it? I'm not planning on using a USRP as a receiver...

OIP3 measurement is a reasonable transmitter quality metric. There are others that are system specific you could use, but it's all measuring the same thing.

3. Re: Clipping. I expect your right, i've used an amplitude of A = 0.25 to avoid the DAC from saturating / causing clipping.. so ideally.. (if i'm right) any non-linear behaviour will be due to the OIP3 rather then the DAC.

it's worth making sure of that by measuring the DAC output directly, if possible.


On Fri, Sep 14, 2012 at 3:19 AM, Ben Hilburn <address@hidden> wrote:
Michael -

Hm, I don't understand what you are attempting to measure. Measuring the input IP3 (IIP3) for a transmitter doesn't mean anything. All that matters is output IP3 (OIP3). The e-mail you linked to is Marcus explaining how to measure the IIP3 through the receive-only port of a daughterboard, using the same daughterboard as a signal source. You are attempting to measure the IIP3 of a transmitter only - whatever data you collect won't tell you anything meaningful.

David -

Gain levels are approximate, and it is also possible that at the upper amplitudes the ADCs are clipping anyway. 


On Wed, Sep 12, 2012 at 7:56 PM, David Bengtson <address@hidden> wrote:
Let me preface this by saying I mostly lurk on this list, and haven't touched a USRP of any variety, so I'm unfamiliar with their gain lineup. However, looking at this data, a couple of things come to mind.

1-You're sweeping the gain by 35 dB, and the amplitude is moving by 28 dB, which seems to indicate to me that something strange is going on with the levels.

2-I graphed the levels of the tones, and they aren't moving in amplitude until the gain setting is 15 or so, which seems unusal to me. What does the spectrum analyzer display look like when you are doing this? Do you see visible tones at deltaF? What about at other frequencies? Are you using a real SA or the USRP?

3-The other thing to look at and worry about are IP5/IP7 tones, which can really be a problem when you're approaching P1dB on the output power. I'm suspicious that at higher gain settings, you're starting to see these tones show up.

Here's the google spreadsheet I imported your data into.


On Mon, Sep 10, 2012 at 10:32 AM, Michael Hill <address@hidden> wrote:
Apologies that copy paste formatting was attrocious..! Data attached in file.

P.S. earlier I meant to say f1 = 4MHz and f2 = 5 MHz
In the attached data

m1 = 2*f1-f2
m2 = f1
m3 = f2
m4 = 2*f2-f1


On Mon, Sep 10, 2012 at 10:56 PM, Michael Hill <address@hidden> wrote:

Can anyone provide any insight into why i'm getting the results I am?
The aim of this test was to measure the OIP3 during transmit, then determine the IIP3.

I conducted two-tone tests using f1 = 4 MHz and f3 = 5 MHz.
The LO was set to 2400, 2450, 2500, 4900, 5400 and 6000 MHz (I'm using the XCVR2450 with the USRP N210)
The Amplitude variable has been set to 0.25, and the gain has been stepped from 0 to 35dB.

The two tone test was essentially setup as the same in this link but with the the output measured by an oscilloscope, thus i'm sure i've done it right..

I've measured the power amplitudes, also the gain has been calculated at each step (since the gain request may not actually get the specific gain).
Measurements have been taken with an agilent spec amp at 300Hz.

I've then calculated the OIP3 and subtracted the actual gain to get the IIP3.

However, my data is a bit odd... An example has been provided in the table below. It seems that my intermodulation products are not increasing at 3dB for each 1dB..
What does this imply?

Examples of my transmit IIP3 results are attached, these seem somewhat in the jagged ball park of the receiver results below for the WBX.
Not an apple to apple comparison I know.. but I don't have much to go on..
am I missing something basic (I am still.. in my opinion.. new)

(Note: units below are MHz for frequencies, and dB for gain,  dBm for the output power.)

Carrier   m1 m2 m3 m4
2400   3 4 5 6
2400 MHz   2403 2404 2405 2406
0   -50.27 -11.88 -11.58 -49.31
1   -49.79 -11.05 -10.79 -48.74
2   -49.37 -10.37 -10.09 -49.06
3   -48.77 -9.62 -9.36 -47.62
4   -48.89 -8.93 -8.67 -47.66
5   -48.16 -8.08 -7.79 -47.26
6   -48.5 -7.43 -7.16 -46.71
7   -47.9 -6.63 -6.36 -48.1
8   -47.52 -5.98 -5.7 -46.06
9   -46.65 -5.24 -4.98 -46
10   -48.13 -4.58 -4.36 -46.07
11   -46.95 -3.84 -3.63 -47.07
12   -47.6 -3.23 -3.02 -45.6
13   -47.14 -2.51 -2.32 -46.78
14   -47.74 -1.81 -1.62 -48.15
15   -47.98 -0.98 -0.78 -45.74
16   -47.06 -0.31 -0.14 -46.42
17   -47.58 0.48 0.67 -47.57
18   -46.64 1.24 1.39 -47.44
19   -46.15 2 2.2 -46.19
20   -43.16 2.87 3.04 -44.33
21   -40.62 3.71 3.86 -42.54
22   -39.47 4.48 4.61 -39.99
23   -36.41 5.24 5.36 -37.01
24   -35.01 6.03 6.16 -35.63
25   -32.85 6.81 6.94 -33.63
26   -31.27 7.72 7.87 -31.36
27   -29.14 8.62 8.7 -30.24
28   -27.78 9.59 9.7 -27.9
29   -26.18 10.68 10.82 -26.36
30   -24.77 12.03 12.14 -25.26
31   -25.29 12.04 12.15 -25.62
32   -18.65 13.93 14.03 -19.96
33   -10.29 15.26 15.33 -10.31
34   -9.91 15.25 15.33 -10.09
35   -4.42 16.22 16.3 -4.33

Discuss-gnuradio mailing list

Discuss-gnuradio mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]