Products > Test Equipment
OWON XSA1032-TG 3.2G Spectrum Analyzer
<< < (8/8)
TomC:
Third-order intermodulation - Intercept (TOI) Test

I read several interesting articles regarding this subject and found that although different looking formulas are used to derive the TOI, they are just that, different looking. In essence, they are all describing the same thing. Attachments 1-3 are a sampling of the most relevant articles I used as reference.

The SA manufacturer's specs for the test are in attachment 0, the test description that I used to perform the test is on attachment 4. The manufacturer's specs call for a ≥ 50 MHz two tone input at -20dBm with a 100kHz separation between the tones (signals). I couldn't use 50MHz since I'm using my AG1022F generator that maxes out at 25MHz. This is a dual output generator, so I set channel 1 to 25MHz and Channel 2 to 24.9MHz.

The test instructions (attachment 4) call for a "Power Coupler", also known as a combiner, to connect the generators to the SA. I don't have a combiner, instead I used a BNC T coupler and two cables to make the connection (see attachment 6). This is not optimal but is allowed for lower frequencies (although 25MHz isn't that low). To keep the 50-ohm system as intact as possible, I changed the setup of the generator's outputs to 25-ohms. This should result in a 50-ohm combined output to match the SA's 50-ohm input.

Attachment 5 is a screen from one of my favorite dB calculators. It shows that for a -20dBm output about 63mVpp are required. I started out my generator's outputs at that level, but to get the -20dBm required by the specs I had to increase the level a little, -70dBm on channel 1 and -67dBm on channel 2.

Attachments 7 and 8 show the spectrums of the two fundamentals and the intermodulation products 2f₂-f₁ and 2f₁-f₂ as indicated in the test instructions. There are other intermodulation products, but these ones are the closest to the fundamentals and for that reason the most problematic (can't be easily filtered) when present on a DUT. From the marker readings the TOI can be calculated. In the attachments, as I mentioned before, there are a number of different looking formulas for this, but they all give the same answer. So I'm going to use the one in the test Instructions.

   TOI = Rl + [Ir₃/2]

   Rl is the power level of the fundamentals, -20dBm.

   Ir₃ is the distance between either one of the fundamentals and largest (smallest level) of the two intermodulation products (IMDs). I'm going to round that to 50dBM.

   So we have -20 + [50/2] = +5dBm

Well, that's half the +10dBm that we are supposed to have according to the specs but given that the test conditions were far from optimal, I think I'll regard it as a pass!
TomC:
1 dB Gain Compression Test

I wasn't able to perform this test as specified. The main reason, as I see it, is that the SA has safeguards that prevent the user from manually setting certain parameters. As I understand it, the test calls for 0dBm attenuation and a ≥ 50 MHz input signal that shouldn't show the 1dB compression, nominally, before it's level reaches +2dBm (see attachment 1). So to see the signal, the reference level has to be higher than at least +2dBm, but the SA won't let me set the attenuation to 0dB unless the reference level is set a lot lower, -20dBm or lower.

In addition, although there is an abundance of information for performing this test on an amplifier, I couldn't find specific information on using the test to characterize the SA itself. The Fluke publication I've been using as a guide for other tests, has a write-up on a "Compression Test", but is a different test that deals with compression of an adjacent smaller signal when a larger signal is present. Attachment 2 is the complete Fluke pub in case some of you want to read it. Attachment 3 is a write up I found that in my opinion comes the closest to explaining the procedure and reasoning behind the test I wanted to perform.

The rest of the attachments deal with the procedure I envisioned to perform the test. Attachment 4 is the conversion table I used to set my generator at the appropriate dBm levels, my AG1022F only accepts pp values or rms values. Attachments 5 and 6 show the 25MHz fundamental at dBm values of 1 to 5dBm, I used different Traces and scattered them on the screen for better visibility by changing the Center Frequency before doing the Max Hold on each trace. The one thing I couldn't do is set the Attenuator to 0dB.

With the 25dB attenuator setting I didn't expect to see any compression, but if you look closely at the last two peaks, it appears as if compression is starting to creep in. I thought maybe this meant that there was some kind of non-linearity in the display, so I changed the reference level to 8dBm to see if there was a difference (attachments 7 and 8 ). But no, it's the same. I hooked up my DSO to check the signal mVpp values and it agreed with the SA. Seems that either there is loss on the cables, or my generator is not quite outputting the calibrated values.

So that's the best I could do! But please, if any one of you has any insights or opinions on how to do this test correctly, post it.
TomC:
Resolution Bandwidth Test Part 1

I found this test very interesting. Besides the test specs and test instructions (attachments 1 and 2), I read an articles that compares the performance of the analog filters used on older SAs, to the current digital filters (Attachment 3). On Part 2, I'll try to duplicate the example showing how the new digital filters allow the spectrums of very close adjacent carriers of different amplitudes to be viewed and analyzed and do so with much faster repetition rates than what could be achieved with the older generations.

The spec identifies the different RBW filters available (10Hz to 500kHz (1-10 steps by sequence), 1MHz, 3MHz). I did test the selectivity of each one of these filters, but I've only attached a sampling of the spectrums I obtained to illustrate the main things I found (Attachments 4-9). To check the "Bandwidth Selectivity", I didn't follow the Instructions on attachment 2 to the letter because this SA has an NdB Marker Function that allows the bandwidth at the dB level of interest to be determined automatically.

According to the spec, the "Resolution Filter Shape Factor (60 dB : 3 dB)", this is what the instructions call the "Bandwidth Selectivity", should be "<5: 1 typical". The instructions show that this is calculated using: f60dB / f3dB, for this to be less than 5, the 60dB bandwidth can't exceed the 3dB value by more than 5 times. It also states that the typical ratio is 1. I'm afraid that I never saw 1 on any of my measurements, 1 would mean that the 60dB and 3dB bandwidths are the same. As far as the ratio being <5, that checked out fairly good with some caveats. The details are as follows:
  Edit: I didn't look at the punctuation close enough on "<5:1 typical", It's a colon, not a semicolon as I
           thought, so it claims the ratio is <5:1 typical, which is OK, not 1 typical as I mistakenly interpreted it.

   RBW = 10Hz: Attachments 4 and 5 show the 3dB and 60dB bandwidths. 143 / 10 = 14.3 not <5. So for RBW=10Hz the selectivity
   is not as good as stated. With the span   set to 500Hz, we have 50Hz per division. Notice that about half way between -60dBm
   and -70dBm the bandwidth is about 50Hz, that would be the 35dB bandwidth of the filter and the point where we actually
   have a 5 ratio. So the filter can still resolve different frequencies as specified but only for larger signals.

For all the remaining spectrums I'm not showing the 3dB bandwidth because I found out it's the same as the RBW setting or very very close. The next few possible RBW setting have an improved signal shape, and are closer to the stated spec, but still somewhat off. Starting with RBW = 50Hz specs are met for several possible settings.

   RBW=50Hz" Attachment 6 shows the 60dB bandwidth, the 3DB bandwidth is the same number as the setting (50dB). 247 / 50 <5.

For the settings prior to 300Hz the specs are met, but starting at 300Hz the noise floor starts to get in the way. To get a 60dB reading with the current settings, the noise floor has to be less than -90dBm. But right at RBW = 300Hz this is no longer the case, and it gets worse from this point on, since the noise floor gets higher with every further increase in RBW.

   RBW = 300Hz: Attachment 7 shows the 60dB bandwidth, the 3dB bandwidth is 300, 1700 / 300 = 5.66, not quite <5.

Even though the 60dB bandwidth can't be accurately determined on subsequent RBW settings due to the noise floor, the shape of the filter is still good, and if the bandwidth is measured just above the noise floor, the ratio between this higher level bandwidth and the 3dB level bandwidth is still <5. Attachments 8 and 9 are examples of this situation.
TomC:
Resolution Bandwidth Test Part 2

To get some idea of how the different RBW filters perform in real live, here i'm trying to recreate the example described on the "Agilent AN 1318 Optimizing Spectrum Analyzer Measurement Speed" article I attached to the previous post. Namely, to check the SA's performance when trying to visualize two CW signals that are 240Hz and 20dBs apart. In the article they use signals around 1 GHz at -35dBm and -55dBm, due to equipment limitations I'm using around 25MHz at -30dBm and -50dBm.

Attachment 10 contains the spectrums shown in the article as displayed by an Agilent 8563E SA at RBW=30Hz and RBW=100Hz with a 5kHz span. The equivalent spectrums on my XSA1032-TG are attachments 14 and 15.

Attachments 11 to 13 show the same signal using RBW=10Hz, here I tried span 500Hz, as well as the 5kHz span they used on the article. Notice that even at RBW 10Hz, the sweep time is nowhere near the 16.7 seconds it takes for RBW=30Hz at 5kHz span when using the older Agilent 8594E (Figure 5 in the "Agilent AN 1318 ..." article).

Attachments 16 and 17 show the same signals using the RBW=30Hz and 100Hz used in the article, but using a 1kHz span instead of the 5kHz span. I included these because it seems that the Delta measurements are a bit more accurate at this setting.
Navigation
Message Index
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod