Saira,
GMSK modulation is FSK modulation, modulation index 0.5, Gaussian filtered. Access it from the ASK/FSK controls.

1. Yes. Powers of 2. 16, 32, 64, 128, …
2. Yes.
3. Number of points varies a little based on window bandwidth. For the legacy API, it is closer to 3.5 * span / RBW
4. 440.4 to 440.8 MHz would probably be 2 slices (440.4 – 440.6, 440.6 – 440.8). If each slice is 56 ms, your sweep time would be 112 ms.
5. The API expects FFT size. The user interface expects RBW. Use the relationship from question 2.
6. IF bandwidth is set using decimation (IF BW = 240 kHz / decimation)
7. I would expect this to work. Configure, sweep, configure, use measuring receiver (actually more of a modulation analyzer).
Let me know if you have more questions.

Also, below 28 MHz, the output is DC coupled (as it goes down to 10 Hz). This small DC offset has a 50 ohm impedance, so it adds a little to total envelope power, but will not damage the SA44B input.

Ivan,
The TG44A produces 10 Hz to 4.4 GHz, -30 dBm to -10 dBm. This is why 0 dBm and -40 dBm do not read accurately–they are beyond the amplitude range of the device.

The TG44A, as a tracking generator (not a signal generator), has a very simple job: generate a signal that tracks the SA44B’s input frequency, at a specific frequency and a repeatable amplitude. Since the SA44B can filter out harmonics, we do not filter them on the TG44A. This reduces cost, size, weight, and power, without really affecting performance as a tracking generator.

Below 28 MHz, the frequency is generated with a DDS, thus the sinusoidal output. Above 28 MHz, it is produced by a clock or divided LO, and will be closer to square than sinusoid. A external low pass filter can easily remove harmonics to produce a sinusoid if needed.

Yes, there is. It is aptly named the TG standalone API. Email support at signalhound dot com for a copy of this.

Ivan,
The actual value of the 20 dB pad is not important. It is measured and nulled as part of the store 20 dB procedure. We just need enough attenuation so we don’t overdrive the ADC when setting it to the high gain state. This is usually 12-16 dB minimum. A 30 dB pad would even work, but the readings would be noisier.

The BB60C residual responses are specified to be below -106 dBm, are typically below -110 dBm (at -50 dBm ref level) and most of them go away when you turn “spur reject” on. Unfortunately, it is difficult to do much better at our price point, and impossible to eliminate spurious / residual responses entirely, at any price.

Thank you for the suggestion. Our solution for frequency hopping spectrum analysis is the BB60C. It has hardware image rejection and very fast sweeps.
There are a number of solutions for frequency hoppers using the SA44B, but they are generally case-specific and do not work in the general case. If you can keep the hopping to a 20 MHz band, the SA44B can be used with image rejection off (disregard the image responses ~21.4 MHz away), or if the transmissions are long at each hop (>sweep time) you can keep image rejection on and analyze 40 MHz worth of spectrum.
Beyond this, the BB60C is required.

BBowar,
The USB-SA44B has higher LO feedthrough than other models we sell, and the LO is offset by only 10.7 MHz. You can minimize it by setting a reference level around -30 dBm (which will turn on the RF preamplifier), or manually turning gain to 1, preamplifier on, attenuator to 10 or 15 dB. This should reduce the LO feedthrough to around -60 dBm (typical).
Choosing a product like the BB60C will keep LO feedthrough below -80 dBm, and will keep the LO > 1.2 GHz away from the RF.
You can also add a circulator, ore even an external attenuator and/or preamp, to reduce LO feedthrough if needed.

The SA44B would work for measuring steady-state noise, such as from a clock, but intermittent noise would not measure accurately. The BB60C would be the recommended product for measuring intermittent / pulsed noise.

Ryan,
Unfortunately, spurs like this are normal for the USB-SA44B. The frequency is usually a little higher, but it depends on your cable lengths and computer. It looked like your spurs peaked around -100 dBm with a -30 dBm reference level for a 100 MHz span.

When you use a span of 200 kHz or less, most of these spurs will be drastically reduced or eliminated. This is because the SA44B mixes differently for narrow spans.

The best workaround is to show multiple captures. You could do several 200 kHz steps with low RBW. If you wanted you could export them as CSV, paste them together, and then plot using spreadsheet software or your favorite plotting tool.

I realize the RBW/span limitation at low frequencies seems odd, but the combination of hardware limitations at those frequencies, and the sweep performance we wanted for Spike, made it necessary.

With a shared external frequency reference, in streaming mode only, there should be zero phase drift. However, every time you change center frequency, there will be a new phase offset.

If there is phase drift from floating point frequency correction rounding errors at some frequencies, I would think it would be very small and identical across devices and therefore cancel, but I have not tested this (maybe 1 degree per minute?).

At even multiples of 20 MHz (e.g. 200.00 MHz), this correction should be zero, so you should get a true zero phase drift. If you test this, let me know. If there is phase drift, there should be a way to easily get rid of it, possibly by using streaming IF instead of I/Q…

The standard RBW works well for the BB60C, but not for the USB-SA44B. Because of the software-based image rejection, the default noise bandwidth is 2 x RBW, but then there is the software image reject algorithm. To get NBW close to 2 x RBW, you would need to either disable “spur reject” / image reject, or set video processing to power average and turn VBW down to 1/10 RBW.
To get noise BW close to RBW, you would need to filter out the image frequency. For wide sweeps, this is 21.4 MHz above or below the image frequency. For 200 kHz spans or less, this is 5.8 MHz above or 21.4 MHz above the measured signal. With image/spur reject on, filtering out either image will result in a good measurement.

Another thing to note: For accurate power measurements, you need to keep the total power into the BB60C below +10 dBm. Your picture showed over +16 dBm. You will want to use a 10 or 20 dB pad for accurate measurements.

For 6 dB bandwidth test, I believe the test is marker peak, then marker to the first point (left-to-right) less than 6 dB below peak, then delta, then last point (right-to-left) less than 6 dB below peak. You may wish to verify this, but I believe this is the 6 dB test.

Bbowar, to simplify the thought process, we use wide flat top windows, so the noise bandwidth is very close to the resolution bandwidth.

12V is not too bad, it’s people connecting to AC mains that can be really scary. An attenuator (or a limiter if sensitivity is important) is always a good idea though.

Bbowar, you can also look at the simple, less functional Labview example in your C:\Program Files\Signal Hound\Spike\api\sa_series\labview folder. It doesn’t have all of the functionality of Lior’s program, but it might be easier to get running if you are experiencing difficulties.

Jared,
AJ pretty much summed up what you see. The center frequency gets set, then the frequency modulator (which offsets the center frequency) gets activated, then the modulation pattern begins looping.
I am looking at adding packet-based digital modulation with a 2-508 symbol packet followed by an “off” period of up to 65,000 symbols. This could easily give you a 1% duty cycle. Let me know if you are interested. The only (minor) challenge is I would have to convert FSK math from frequency modulation to its quadrature amplitude equivalent to enable this.

[Author]

Posts