Forum Replies Created
- Posts
GSingh,
To determine whether the IMD3 products are from the generator or the analyzer, change the reference level on the analyzer by 10 dB. This will increase attenuation by 10 dB. If the levels stay the same they are from the generator. If they change significantly they are from the analyzer.Above 200 kHz span, only the 10.7 MHz IF is used, and the LO is injected first low side, then high side. The noise figure of the 10.7 MHz IF is maybe 5 dB higher, so this can affect measurements where the level of noise present is low.
As long as your VBW is much lower than your RBW, you can still make decent noise measurements, with the caveat that you’re measuring the combined noise from RF and image.We use two different IF frequencies when software image/spur reject is active and span is 200 kHz or less: 2.9 MHz and 10.7 MHz. The image frequency is [RF frequency] + 2 * [IF frequency]
Internal measurements will be made with the 2.9 MHz IF, followed by the 10.7 MHz IF. The lower reading for each frequency bin is displayed.
If you disable spur reject, only the 10.7 MHz IF will be used.The BB60 or SM200 is a better choice for noise measurements, since the image is rejected in hardware, and software spur reject is off by default.
Unless video bandwidth is much lower than resolution bandwidth, our image rejection algorithm (a minimum of 2 measurements function) will not preserve average noise energy.
My recommended settings are what I would recommend using. And if you decide to try a BB60C, you can avoid the problem entirely.Lsullivan,
At a high level, we use overlapping windowed FFTs for real-time mode. Typically, 50% overlap is used.
The SA44B is not ideal for broadband measurements. We generally only recommend it for narrow-band signal analysis (<250 kHz occupied bandwidth is ideal) due to the limitations of the SA-series software image rejection.
Because we use a software image rejection algorithm, making noise measurements can be tricky. A span of 200 kHz or less should be used. Good settings might be 1 kHz RBW, 30 Hz VBW, power, average. Then you have to factor in that the noise from the image will add to the reading, typically 3 dB.
Hope this helps. The BB60C might be a better choice for noise measurements, as it uses hardware image rejection so it can be used like a normal spectrum analyzer (the BB60C should not be used for noise measurements below about 100 kHz).
Another thing to note is that the preamplifier is off for sweeps that start below 500 kHz or so. So you will not have maximum sensitivity at low frequencies.Serpi,
The noise marker, using power, average settings, will give you a good noise measurement. The trace does not have to be in average mode, but it may help stabilize the reading a bit more.
You will find most frequencies have a noise figure quite a bit better than our specification. As a spectrum analyzer, we are required to meet the DANL specification at all operating frequencies for the life of the instrument, so padding these numbers a bit is prudent.
DDR,
You should not have big dropouts like that. This would indicate either hardware failure, bad correction data, or possibly bad sweep settings resulting in bad measurements.
To rule out the sweep settings issue, please test individual failing points with the VSG in CW mode, and the SA124B set to a 200 kHz span centered at the output frequency. You can email support at signalhound dot com and we can try to figure out if we can fix it with correction data or if we will need to repair one of the units.Serpi,
If you select average, power units, there is no 2.51 dB correction. This only applies to log averaging. Our DANL spec is worst case. NF is usually calculated off of typical DANL. You can use the noise marker with a reference level of -50 dBm, input terminated, at the frequencies of interest. Then add 174 to calculate noise figure of your actual unit.Hiroshi,
Yes, you can Store Thru and Store 20 dB pad, both with the 20 dB pad in place (leave the 20 dB pad in place for testing). This will enable high dynamic range sweeps, with 20 dB additional gain available.Hiroshi,
I’m glad you like it!
1. Yes. Remove the 20 dB pad after it has been measured. We use the 20 dB pad to measure both high gain and low gain and compute an internal offset.
2. Make sure you don’t exceed +20 dBm into the SA44B. You will need to compensate your measurement for the additional gain after the store thru, but otherwise it should work just fine.Please note that the BB60C or SM200B/C will provide lower EVM numbers for a clean GSM signal, as I believe it barely fits in the SA44B bandwidth so there is significant rolloff at the band edges.
Stanley,
There are two major factors that affect the accuracy and stability of scalar measurements:
1) The stability and phase noise of the TG
2) VSWR
There isn’t a lot you can do about the TG, but adding 3-6 dB fixed SMA attenuators on the output of the TG and the input of the SA and keeping them there for both the thru measurement and the DUT measurement can improve measurements quite a bit by reducing VSWR.Michael,
I suppose if you had a 2-channel oscilloscope, and used the API, you could stream data out of both with embedded triggers, use the scope (programmatically using SCPI) to calculate the time delta, and then add or skip the appropriate number of samples on one channel. You’d want an external 10 MHz reference to tie them together as well.It’d be a bit of a project… a 2-channel VSG or SDR would be a more straightforward answer.
Jim,
We typically see 1.2 amps, but depending on mode and input voltage, 1.3 amps would be at the high end of the normal range. About 1/2 of the power draw is from switching regulators, the other half from LDOs, so current can be a little higher if the voltage is lower.Michael,
The USB data comes back as packets with header and trailer data. If the software cannot “frame” the data such that the headers and trailers are where they’re supposed to be, you get this error.
Power cycle the BB60C, maybe try a different USB port, and let us know if the problem persists.Mcline,
If you look in vsg_example_complex_freq_hopping, you’ll see vsgSubmitIQ(). You could output a trigger pulse by adding vsgSubmitTrigger(). The API sends them to the VSG in the order received. So if you Submit some I/Q data, then submit trigger and then more I/Q data, the trigger’s rising edge will match the first sample of second chunk of I/Q data.It looks like you’re missing a scale factor on the I/Q data, and phase wrapping math on the FM plot.
DDR,
If a 30 MHz high pass filter would not affect the signals you need to produce, the VSG60A will probably work. Otherwise, you’ll need a dedicated arb / function generator instead.