ALM2403-Q1: MFB filter component selection and Q value

Hi Cherry,

Cherry Zhou said:

Is the cutoff frequency set to be a bit larger than 10kHz and smaller than 1/5 to 1/10 of the PWM input frequency ok? 

In switching power supply or SPWM input signal used to generate resolver sinusoidal excitation frequency, these high frequency signal associated with harmonic noises (谐波噪声). As a designer, you'd like to keep these harmonic noises away from the system's feedback and sampling frequency. Therefore, the higher the switching frequency would be beneficial to keep away these unwanted noises, and you can implement low cost LPFs to attenuate the unwanted switching harmonics signals.

For instance, if SPWM switching frequency is used at 100kHz, then 1/10 of the switching frequency is approx. 10kHz, which typical 10kHz, 2nd order LPFs are used to attenuate the undesirable switching harmonics. The 2nd order LPFs will provide -40dB/decade attenuation (or -6dB/Octave). At 100kHz, the LPFs are able to attenuate approx. 1/100th of input signals. 

If the 10kHz is used as the excitation frequency for the resolver, let us assume there is 30mVpp noises (voltage ripple) riding on the excitation frequency. At 100kHz, the 30mVpp will be attenuated at 0.3mVpp based on -40dB/decade LPFs. The 10kHz sinusoidal excitation frequency has its even and odd harmonics as well, since this is not pure sine waveform. If a system has a sampling frequency, say at 40kHz, the 2nd order filter is only capable to attenuate at approx. -24dB or ~1.9mVpp noises, which it may or may not be acceptable for the application.   

So LPFs with 1/5 of switching frequency may work, but you have to take into the design requirements. There are switching harmonics that you need to take into account and there are the harmonics from the excitation frequency and other noises in the system where you need to pay attention to. What are the acceptable noise levels in a design or Signal/noise ratios (errors) that are acceptable in a resolver design requirements? Is the 2nd order LPFs good enough? Higher order LPFs can be done, but it costs more and requires more components and larger PCB footprints. 

Cherry Zhou said:

a. The frequency of the resolver is 10 kHz, why is the PWM frequency set to 320 kHz?

To keep away the switching harmonics away from the resolver sampling system, see the reply above. 

Cherry Zhou said:

b. Why is the PWM input not set to the same operating frequency as the resolver? 

To attenuate the unwanted switching harmonics from the resolver's sampling system. 

Cherry Zhou said:

c. Are there any specific requirements for setting the PWM frequency? 

If the excitation frequency is 10kHz and the cutoff frequency of LPF is 16kHz, ideally the SPWM frequency should be 10X to 20X higher, or the minimum switching frequency should be configured at approx. 160kHz. 320kHz or higher would be better.

Cherry Zhou said:

So is the Q value selected to be =1 or <1? If < 1 or > 1, what would be the appropriate value? 

This depends on the filter construction and the design requirements. In the active filter, the Q is tunable as implied in the application note. Generally, you want keep the Q or qualify factor close to 1. Higher Q indicates a lower signal attenuation from the input. 

You have one or two high frequency SPWM signal and you want to filter the signal and convert into a sinusoidal waveform. Let us assume that the excitation frequency is 12kHz, then the filter's cutoff frequency will be optimized approx. 18kHz - 24kHz range (1.5X-2X of the excitation frequency). If customer wanted to filter out with lower cutoff frequency, it will be ok as well, but the input signal will be greatly attenuated (higher cutoff frequency provides a lower lower energy loss from SPWM signals). 

If the input signals have smaller voltage amplitude, and the customer has to increase gains in order to compensate it. The design of the LPFs are about compromises, and it is difficult to describe how to in a few paragraphs. If you need additional help, please send me a WebEx invitation and we could discuss the issues in real time. 

Here are two examples how SPWM signal can be simulated and filtered out.

SPWM Generator 10112023.TSC

Enclosed is 2nd order MFB filter, which you can simulate the R and C parameters. 

ALM2403-Q1 MFB Filter 10112023.TSC

If you have other questions, please let me know. 

Best,

Raymond

Hey Cherry, without getting into your why this filter questions, I needed to exercise my MFB filter tool every once in awhile, here is another set of RC values for the filter. 

1. The op amp is kind of higher noise, the R's have been shifted up to not hurt that noise but scale the loading and C values down, 

2. The ratios have been tuned to increase the loop gain inside the filter, 

here is this file, if you wanted to try some PWM inputs (or TI support)

ALM2403 MFB filter.TSC