FFT simulation in cadence

[the8thhabit]

Hello!
I am designing a noise shaping SAR ADC in CADENCE and I have a question in FFT simulation.
I used a 10-bit CDAC, OSR = 4, fs = 20MHz, sampling count = 2048.
Then the bandwidth is fs/2/OSR, so 2.5 MHz.
If I set k=31 in the equation k*fs/sample count, the fin (input frequency) would be 0.302MHz.
I ran the FFT simulation until 2048*50ns = 102.4us and then ran the FFT simulation.

In this case, I was wondering if it is correct to set the end frequency to 2.5MHz, which is the bandwidth, as shown in the photo below.

 

[sidun.av]

Hi @the8thhabit , your calculations seem right to me. I just have a few comments:
1. I hope when you mention fin = 0.302MHz you set it exactly to be 302734.4 Hz -> this will help avoiding spectral leakage and will result in more precise FFT.
2. I can recommend running your simulation for a bit longer than 102.4us to allow your circuit to settle (first few tens of points might be not be accurate due to the initial charging of the capacitors etc. at the startup) and take the last 102.4us from your simulation. This will help achieving better FFT results as well.
Hopefully, that helps.

 

[the8thhabit]

Hi @the8thhabit , your calculations seem right to me. I just have a few comments:
1. I hope when you mention fin = 0.302MHz you set it exactly to be 302734.4 Hz -> this will help avoiding spectral leakage and will result in more precise FFT.
2. I can recommend running your simulation for a bit longer than 102.4us to allow your circuit to settle (first few tens of points might be not be accurate due to the initial charging of the capacitors etc. at the startup) and take the last 102.4us from your simulation. This will help achieving better FFT results as well.
Hopefully, that helps.

Your answer really helped me a lot.
I will try that again.
Thank you so much!

 

[austinmariam]

Hello!
I am designing a noise shaping SAR ADC in CADENCE and I have a question in FFT simulation.
I used a 10-bit CDAC, OSR = 4, fs = 20MHz, sampling count = 2048.
Then the bandwidth is fs/2/OSR, so 2.5 MHz.
If I set k=31 in the equation k*fs/sample count, the fin (input frequency) would be 0.302MHz.
I ran the FFT simulation until 2048*50ns = 102.4us and then ran the FFT simulation.

In this case, I was wondering if it is correct to set the end frequency to 2.5MHz, which is the bandwidth, as shown in the photo below.

View attachment 197365

Yes, setting the end frequency to 2.5 MHz (fs/2/OSR) is correct since that defines your signal bandwidth. Just make sure your windowing and FFT settings match your chosen parameters for accurate results.

 

[the8thhabit]

Yes, setting the end frequency to 2.5 MHz (fs/2/OSR) is correct since that defines your signal bandwidth. Just make sure your windowing and FFT settings match your chosen parameters for accurate results.

ok!! Thanks for your reply!