**Bode Plot comparison**The most intriguing new feature of the Siglent SDS1004X-E series is the Bode plotter – others call it FRA (Frequency Response Analyzer). Checking frequency responses belongs to the daily business of analog circuit designers, as it's used for checking transfer characteristics and port impedances of passive and active components as well as circuits such as amplifiers, attenuators, splitters, filters and any combination of them, e.g. control loops.

I’ve not published a review for the Bode plot feature of the SDS1004X-E so far, simply because it is still work in progress and doesn’t quite meet my expectations yet. There are several optimizations and refinements that I’d like to see and also one major issue that prevents the Bode plotter from passing all my tests. I hope this will change with the next update.

Nevertheless several members here have already shown some experiments with the Bode plot feature, so I guess it’s time for me to make up leeway and publish a little teaser as well. My goal is not just to prove that it does exist, but also what level of performance can be expected (once I declare it mature) – at the very least, since I will not stop pushing Siglent really hard to keep improving it.

For now, I’ll just show a little benchmark test between the following contenders:

- Signal Hound SA44+TG44 – as a representative for the traditional SNA approach.
- PicoScope 4262 (16 bit) FFT using peak hold with external sweep generator.
- PicoScope 3206B (8 bit) and its internal AWG driven by the open source FRA4PicoScope frequency response analyzer software.
- Siglent SDS1104X-E Bode plotter with 7.6.1.20R1 firmware.

For this shootout I have built a simple 455kHz IF filter, consisting of a Kyocera KBF-455R-20A ceramic 6 element filter with two resonant 2nd order L-matching networks for the 50/1500 ohm impedance transformation at both the input and output. This is a rather complex structure with some unwanted responses typical for the ceramic filter, the analysis of which requires high frequency resolution and a wide dynamic range.

- 50 ohm through terminators have been used for all DSO inputs.
- For the FRA applications on the Pico 3206B and Siglent SDS1104X-E, a resistive wideband power splitter has been used to tap off the input signal for the reference channel.
- Only the FRA applications can provide phase information, so this is ignored for the first step of this comparison.
- The initial goal was to cover the frequency range from 250kHz to 750kHz, but I had to deviate from that on the Pico Scopes for various reasons.

Let’s start with the Spectrum Analyzer as a reference.

IF_Filter_455kHz_Ref 01a

The markers show the amplitude levels of the three major peaks. The speed of just 1.1s for one sweep is second to none, but it shows as the amplitude accuracy below -50dBc is not that great, particularly obvious at the falling flanks of the filter and its unwanted responses. On the other hand, there is little visible noise and the dynamic range of the measured spectrum is about 75dB, which would have been even better on a wider span.

Now compare this with the excellent 16 bit PicoScope 4262. A signal generator with a sweep time of 120 seconds has been used to cover the frequency range from 100kHz to 1MHz. This span has been chosen because the FFT only offers a limited choice of analysis bandwidths and it makes no difference for this test setup anyway. Either linear or logarithmic frequency axis is supported; for filter analysis we typically choose linear. An FFT with just 8kpts registers the peak amplitudes for the entire frequency span, but at least two generator sweeps are required to get a reasonably nice plot. A total of 10 minutes for 5 sweeps were required to create the plot shown in the screenshot below. More FFT points would have significantly increased the sweep time without any additional benefit. For 8kpts, the frequency step is 244Hz and resolution bandwidth is about 710Hz using the Flat Top window.

IF_Filter_455kHz_Ref 02_FTZ

While this test setup is certainly anything but fast, it has high frequency resolution and >90dB dynamic range which is the best in this comparison by a clear margin. So this measurement can be considered to be the true reference with regard to amplitude levels. On the downside, this solution cannot provide a phase plot and works up to 5MHz only, as this is the bandwidth limit of the 16 bit PicoScope 4262.

Now for the dedicated frequency response analyzer, which I would have loved to try with the 16 bit PicoScope 4262, but unfortunately the FRA4PicoScope does not support external waveform generators and the integrated low distortion AWG of the PicoScope 4264 has an upper frequency limit of only 20kHz. As a consequence, the 8 bit PicoScope 3206B had to be used. The phase plot has been disabled for visual comparability with the previous measurements.

IF_Filter_455kHz_Ref 03

The frequency span had to be 100kHz to 1MHz because otherwise we wouldn’t get any annotation on the frequency axis. The application supports logarithmic frequency axis only, the grid does not look particularly nice and no cursors for precise measurements are available. There’s a lot of visible noise below -68dB and the dynamic range of the displayed spectrum is just about 66dB because of this. This is the worst dynamic in this comparison, particularly obvious at and below 250kHz. This is mainly because of the PicoScope 3206B ‘s mediocre sensitivity of 10mV/div and the rather low output level of its internal AWG of just 2Vpp.

Even more importantly, the AWGs in the PicoScopes have an output impedance of 600 ohms, which makes the level drop by another 22dB. So the dynamic range can be phantastic in high impedance networks, but it is nearly unusable for the widespread 50 ohm standard. Since we are constricted to the internal AWG of the scope when using the FRA4PicoScope, the upper frequency limit is only 1MHz and the practical applications are very limited because of this as well.

The limitation to 1000 frequency steps/decade rules out more extreme narrowband analyses.

On the positive side, there is the pretty fast sweep time of only 11.7s for 500 data points – the screenshot above shows 1000 data points (19.5s) because of the frequency span which is twice as wide as has been planned initially.

Finally we take a look at the Bode plotter in the Siglent SDS1104X-E and its current state with firmware version 7.6.1.20R1. It gives us the choice between linear or logarithmic frequency axis and for filter analysis, linear is what we want. For the first test, the phase plot has been shifted out of view to maintain visual compatibility with the other contenders. Center frequency and span are both 500kHz in accordance with the spectrum analyzer measurement.

SDS1104X-E_BP_455_500kHz_BP_500_nophase

We get a rather nice low noise trace and about 75dB dynamic range for the displayed spectrum. This looks quite usable in my book, even though the unwanted responses appear a bit too low in amplitude compared with the results from the spectrum analyzer or the 16 bit DSO. I have confirmed that this does not change the least bit with higher frequency resolution.

If we want a phase plot, then there are only two opponents left, the FRA4PicoScope and the Siglent SDS1004X-E. Let’s have a look at the PicoScope first.

IF_Filter_455kHz_Ref 03_Phase

Well, that certainly doesn’t look right and the phase plot is pretty much useless. But there is also an option to “unwrap” the phase plot, and that looks like this:

IF_Filter_455kHz_Ref 03_Phase_unwrap

It looks a lot clearer now, yet the phase plot shows only street numbers except for the filter pass band and the unwanted response. It looks like the system attempts to measure the phase of the noise, which is of course not going to lead anywhere. All in all, this solution is not useful for analysing 50 ohm systems.

Finally the complete Bode plot including phase for the Siglent SDS1104X-E. There is no “unwrap” option (yet), so it looks a bit confusing, especially at the spots where the phase jumps between -180° and +180° (which is the same of course) because of noise and/or minor inaccuracies, but at least it is much closer to the truth than the FRA4PicoScope Result.

SDS1104X-E_BP_455_500kHz_BP_500_2

Anyway, this screenshot already hints on a number of desirable improvements, like dynamic adaption of the threshold for the wrap-around when the phase plot is shifted, then of course the “unwrap” option as well as the possibility to completely hide the phase plot, as it isn’t always needed and for complex structures like this, the picture is much clearer without. In fact, nobody really cares for the phase when analyzing a high order narrowband filter like this.

It can also be seen that the phase plot overwrites the amplitude trace, particularly obvious for the 2nd unwanted resonance at 650kHz; this flaw needs to be resolved as well.

Finally, the table below summarizes the general features and the results from this test.

Bode Plot Comparison

The SDS1004X-E shines with its three analysis channels and offers a fairly complete feature set by providing a phase plot and giving the choice of linear or logarithmic frequency axis. It covers a wide frequency range up to 120MHz (a limit that is most likely to be further expanded in the near future) and offers convenient tracking cursor measurements.

The frequency resolution might appear a little on the low side, but it is certainly adequate, especially since this is the only device in this test restricted to a small 7” TFT, whereas all other contenders are essentially PC applications.

The speed is slow but bearable and I hope we’ll see some improvements in this regard.

Amplitude accuracy appears a bit off at the lower level resonance peaks in this test, but that is not conclusive yet and needs further investigation.

EDIT: Max. vertical sensitivity for PicoScope 3206B corrected to 10mV/div.

EDIT 2: Added some important information on the FRA4PicoScope: the AWGs are 600 ohm output impedance, which explains the very poor dynamic when used in a 50 ohm system. The 1000 poits/decade limit makes extreme narrowband analysis impossible.