0
\$\begingroup\$

Consider the schematics below, the switching frequency is 250KHz, So, The cross over frequency be 250/5 = 50KHz. My ADC is able to sample at 1MS/s which is good as from the SNR perspective.Now, the doubts are below

  1. For ADC selection-sampling rate the switching frequency should be considered while determining the sampling rate or the Cross over frequency ?

  2. for the buffer used the Bandwidth should be 2 to 5 times the cross over frequency or the switching frequency?

  3. In case i want to use current shunt amplifier for current sensing, How to select from the Bandwidth point of view ?

In general i wanted to understand how the Opamp parameter be looked at for converter design - unity gain BW, 3db bandwidth etc, used in control loop ?

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
1
\$\begingroup\$

With a ratio of 5 or about 2 octaves with only a 2nd order filter you get -11 to -13 dB of ripple rejection at Fs and only half this at the Nyquist rate of Fs/2 so it will be noisy.

If your ADC is say 10bit with -60dB quantization level and you want to reject PWM noise, then a higher order active Bessel or Chebychev filter with the group delay and rejection is required or a sigma delta type ADC.

Search this site for better ADC designs and filters, and learn to make a list of test criteria or specs.!!!

  • define your specs for input SNR , BW and level range then desired output SNR + quantization noise. From this choose your filter and oversample Fs, with decimation is the best way done by Sigma Delta types or others, just as it is done in DSO’s with Gb/s sampling rates and xxx MHz signal BW.

Unless you also sense input /output current for forward PWM loop error correction with Vout , you will probably experience horrid startup currents and step load overshoot or fancier lead/lag compromises with ripple to improve stability.

You must use minimal lag (group delays) filters and latency for best 1st order stability. 2nd and 3rd order OA effects or 4th order ADC averaging effects all reduce stability.

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ Thanks for the response. Just consider the schematic as a generic buck converter..The ratio of 5 is just a number for representing a condition. I don't have any specific number regarding the performance parameters. My doubt is basically regarding the selection of OpAmp - as listed in the 3 questions..i just want to understand what would be the selection criteria is it switching freq or corss over freq ..specifically if you could respond to the 3 questions..it would be very helpful for me to move forward in further understanding.. \$\endgroup\$ – seeker Jul 14 '19 at 14:47
  • \$\begingroup\$ Op Amp specs of importance are GBW and slew rate limited by Iout/Cload= dV/dt and you specs for step load response in current or sensors. Small signal response is f-3dB= 0.35/Tr while large signal response is current /pF and maybe rail to rail dependencies. For a Buck converter phase compensation is critical with a 2nd order filter for stability and can be inherently unstable to step loads if you use this for feedback with phase shift or quantization delays. You must define your design specs. Better. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jul 14 '19 at 14:49
  • \$\begingroup\$ Lets put it like this, if you consider the above controller is digital with the compensation etc is implemented digitally in let say FPGA.. and i would need an OpAmp in the feedback as shown..So my question would be- The GBW of OpAmp be 'x' times the Crossover frequency or Switching frequency ? \$\endgroup\$ – seeker Jul 14 '19 at 15:15
  • \$\begingroup\$ If you’re using 1MHz GBW unity gain OPAMP then you get an additional -6dB/octave and 45 degree phase shift starting at 1MHz . Got it.? But ADC uses similar or better OPAMP your source impedance only needs to suppress small dynamic input sampling currents which can be done by C2 \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jul 14 '19 at 15:18
  • \$\begingroup\$ I get your point. From time domain perspective considering the case where (switching freq)Fs=250KHz and let (cross over freq)Fc=50KHz, as i understand the OpAmp BW be 5-10times 'Fc'(not Fs) could be taken as rule of thumb to select the OpAmp.Is my understanding correct? Ignore any other design spec for time being \$\endgroup\$ – seeker Jul 14 '19 at 15:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.