FRA Setting for Good Bode Plot

Hello,
I am using the Cleverscope to capture a bode plot of a switch-mode power supply by injecting into the FB loop of a controller and plotting the A and B probes, similar to what is recommended in the PSU gain/phase recommended setup.

It looks like the response begins to get very noisy above 10kHz, and I noticed that when I performed the sweep, the sweep speed greatly increased in the region from 10kHz to 200kHz, whereas it was very slow between 100Hz and 10kHz and recorded clean, precise data.

Is there a setting that I am missing that speeds up the sweep beyond 10kHz (or a certain frequency)?

Hello, I am using the Cleverscope to capture a bode plot of a switch-mode power supply by injecting into the FB loop of a controller and plotting the A and B probes, similar to what is recommended in the PSU gain/phase recommended setup. It looks like the response begins to get very noisy above 10kHz, and I noticed that when I performed the sweep, the sweep speed greatly increased in the region from 10kHz to 200kHz, whereas it was very slow between 100Hz and 10kHz and recorded clean, precise data. Is there a setting that I am missing that speeds up the sweep beyond 10kHz (or a certain frequency)?

Hi Timothy,
Good application, and thanks for sharing! What you are seeing is expected.
The noise is directly related to the Signal to Noise of the measured signal. You can see the SNR on the lower right of the FRA control panel, see below. If the SNR is low, then the response quality reduces.

A power supply usually includes an error amp with lots of gain at low frequencies (maybe 80 dB of gain), which falls off at 20 dB per decade, and so is much less gain at high frequencies. The Auto sweep amplitude method samples the gain at 7 frequencies spread across the frequency range. It may not be giving you high enough signal at the higher frequencies. One way around this is to use the AutTable Sweep method. This will calculate the amplitudes as in Auto, and transfer the values to the Amplitude Table. After that the table is used for amplitude. This gives you an opportunity to tweak. The table uses linear interpolation between table values. So you could increase the amplitude at high frequencies. Don't make it too high, or the error amplifier might saturate. Look at the Response V and SNR as you do the sweep. An SNR above 30 will give better results.

A frequency of 100 Hz has a period of 10 ms. But the error amplifier also has a lot of gain at low frequencies. It is trying hard to reduce the amplitude of our injected signal to 0, so we have to increase the injected signal level to be large , but not so large the amplifier saturates. All the same the signal we measure is small, so to improve SNR we decrease the bandwidth of the measurement (there is a lot of switching noise to eliminate). The Auto automatically adjusts the bandwidth based on the SNR. It might make the bandwidth as low as 5 or 10 Hz - you can see what the actual bandwidth is on the Sepctrum Display. Say it was 5 Hz, which has a period of 200msec. We need to capture at least that amount of time each time we make a measurement. We do that all the way to 10 kHz, and at that point we use the bandwidth set in 'Max Measurement Bandwidth' on the FRA control panel. We chose to do it this way to maximize accuracy in the high gain portion of the frequency response.
At 200 ms per point the low frequency portion takes much more time than the high frequency portion where we sample at 1/Max Measurement Bandwidth.

So as to what you can do to improve the 10 Khz and above:
1. You can reduce the max measurement bandwidth. This will make it slower, but as the bandwidth is lower, there is less noise being measured, and so the SNR will be better.
2. You can use the table to increase the amplitude at high frequencies.

Your picture was very low resolution, and hard to see actual values. Maybe next time you could make it a bit bigger? You can also save your setup after capture by using File/Save As.. and saving to a folder of your choice using a name of your choice. Then attach the resulting .apc file to the post.

I hope this helps.

Bart

Hi Timothy, Good application, and thanks for sharing! What you are seeing is expected. The noise is directly related to the Signal to Noise of the measured signal. You can see the SNR on the lower right of the FRA control panel, see below. If the SNR is low, then the response quality reduces. A power supply usually includes an error amp with lots of gain at low frequencies (maybe 80 dB of gain), which falls off at 20 dB per decade, and so is much less gain at high frequencies. The Auto sweep amplitude method samples the gain at 7 frequencies spread across the frequency range. It may not be giving you high enough signal at the higher frequencies. One way around this is to use the AutTable Sweep method. This will calculate the amplitudes as in Auto, and transfer the values to the Amplitude Table. After that the table is used for amplitude. This gives you an opportunity to tweak. The table uses linear interpolation between table values. So you could increase the amplitude at high frequencies. Don't make it too high, or the error amplifier might saturate. Look at the Response V and SNR as you do the sweep. An SNR above 30 will give better results. A frequency of 100 Hz has a period of 10 ms. But the error amplifier also has a lot of gain at low frequencies. It is trying hard to reduce the amplitude of our injected signal to 0, so we have to increase the injected signal level to be large , but not so large the amplifier saturates. All the same the signal we measure is small, so to improve SNR we decrease the bandwidth of the measurement (there is a lot of switching noise to eliminate). The Auto automatically adjusts the bandwidth based on the SNR. It might make the bandwidth as low as 5 or 10 Hz - you can see what the actual bandwidth is on the Sepctrum Display. Say it was 5 Hz, which has a period of 200msec. We need to capture at least that amount of time each time we make a measurement. We do that all the way to 10 kHz, and at that point we use the bandwidth set in 'Max Measurement Bandwidth' on the FRA control panel. We chose to do it this way to maximize accuracy in the high gain portion of the frequency response. At 200 ms per point the low frequency portion takes much more time than the high frequency portion where we sample at 1/Max Measurement Bandwidth. So as to what you can do to improve the 10 Khz and above: 1. You can reduce the max measurement bandwidth. This will make it slower, but as the bandwidth is lower, there is less noise being measured, and so the SNR will be better. 2. You can use the table to increase the amplitude at high frequencies. Your picture was very low resolution, and hard to see actual values. Maybe next time you could make it a bit bigger? You can also save your setup after capture by using File/Save As.. and saving to a folder of your choice using a name of your choice. Then attach the resulting .apc file to the post. I hope this helps. Bart