What is the minimum sample rate for bandpass sampling?

The minimum sample rate must satisfy fs >= 2*BW (twice the signal bandwidth, not twice the carrier). Additionally, fs must be chosen so the signal does not straddle a Nyquist zone boundary. The valid sample rates are: fs must satisfy 2*fH/n >= fs >= 2*fL/(n-1) for integer n, where fL and fH are the lower and upper band edges. For example, a 10 MHz wide signal at 2.4 GHz (fL=2395 MHz, fH=2405 MHz) can be sampled at just 20 MSPS minimum, but the exact rate must be chosen to avoid zone-boundary aliasing. Practical rates are typically 3-4 times the bandwidth for margin.

What are Nyquist zones in bandpass sampling?

Nyquist zones are the spectral intervals [0, fs/2], [fs/2, fs], [fs, 3*fs/2], etc. When sampling at rate fs, all content in odd Nyquist zones (1st, 3rd, 5th...) aliases to the first Nyquist zone [0, fs/2] without spectral inversion. Content in even Nyquist zones (2nd, 4th, 6th...) also aliases to the first zone but with spectral inversion (the spectrum is flipped). The signal must fit entirely within one Nyquist zone; if it straddles a zone boundary, the upper and lower halves alias to different locations and overlap, destroying the signal. Choosing the right fs ensures clean aliasing.

What are the practical challenges of bandpass sampling?

Three main challenges. First, the anti-alias filter before the ADC must be a very sharp bandpass filter that passes only the desired signal and rejects all other Nyquist zone content, because any noise or interferer in any zone aliases on top of the signal. Second, the ADC's analog input bandwidth and track-and-hold performance must support the full RF carrier frequency, even though the sample rate is low. This is the ADC's 'analog bandwidth' or 'full-power bandwidth' spec. Third, clock jitter becomes critical: at higher input frequencies, the same amount of jitter produces a larger voltage error because the signal slope is steeper. The jitter-limited SNR is: SNR = -20*log(2*pi*f_in*t_jitter).

Digital Signal Processing

Bandpass Sampling

/band-pas sam-pling/ (undersampling)

Bandpass sampling (also called sub-Nyquist sampling or undersampling) digitizes a narrowband RF signal using an ADC sample rate that is at least twice the signal bandwidth, rather than twice the carrier frequency. By exploiting the periodic spectral aliasing property of sampling, the RF signal translates to a lower frequency band without an analog mixer or local oscillator, simplifying receiver architecture for software-defined radio and digital receivers.

Category: Digital Signal Processing

Requirement: f_s ≥ 2 × BW

Key Constraint: ADC analog bandwidth, jitter

Understanding Bandpass Sampling

Traditional Nyquist sampling requires a sample rate of at least twice the highest frequency in the signal. For a 2.4 GHz Wi-Fi signal, that means a 4.8 GSPS ADC. But the Wi-Fi signal itself is only 20 MHz wide. Bandpass sampling recognizes that you only need to preserve the signal bandwidth, not the carrier frequency. By sampling at just 80 MSPS (4x the bandwidth for margin), the 2.4 GHz signal aliases down to a manageable frequency range, preserving all its information. The ADC acts as both a digitizer and a downconverter.

Bandpass Sampling Theory

Bandpass Sampling:
Bandpass sampling (also called sub-Nyquist sampling or undersampling) digitizes a narrowband RF signal using an ADC sample rate that is at least twice the signal...

Key specifications:
2.4 GHz | 20 MHz | 80 MS

Capacity: C = B×log₂(1+SNR)

Sampling Architecture Comparison

Architecture	ADC Rate	Analog Mixers	ADC BW Req.	Complexity
Superheterodyne	2 × IF BW	1-2 stages	IF only	High (filters+mixers)
Direct conversion (ZIF)	2 × BW	1 I/Q pair	Baseband	Medium (DC offset, IQ)
Direct RF sampling	2 × f_H	None	Full RF	Low (but fast ADC)
Bandpass sampling	2 × BW	None	Full RF	Low (sharp BPF needed)

Key Equations

Signal-to-Noise Ratio:
SNR = P_signal/P_noise = 10log(S/N) dB

Spectral efficiency:
η = log₂(1 + SNR) bits/s/Hz (Shannon)

Error Vector Magnitude:
EVM = √(P_error/P_ref) × 100%

Comparison

Band	Range	Wavelength	Application	Standard
Bandpass Sampling	1 GHz region	300.0 mm	Primary use	ITU allocation
Adjacent lower	0.9 GHz	333.3 mm	Related band	Shared spectrum
Adjacent upper	1.1 GHz	272.7 mm	Related band	Guard band
Harmonic 2f	2.0 GHz	150.0 mm	Spurious	Filter required
Sub-harmonic	0.5 GHz	600.0 mm	LO option	Mixer design

Common Questions

Frequently Asked Questions

What is the minimum sample rate?

fs >= 2*BW (twice the bandwidth, not carrier). The exact rate must avoid Nyquist zone boundary straddling: 2*fH/n >= fs >= 2*fL/(n-1) for integer n. A 10 MHz signal at 2.4 GHz needs just 20 MSPS minimum, but practical rates are 3-4x bandwidth for margin.

What are Nyquist zones?

Spectral intervals [0, fs/2], [fs/2, fs], [fs, 3fs/2], etc. Odd zones alias without inversion, even zones alias with spectral inversion. The signal must fit entirely within one zone; straddling a boundary destroys the signal through overlapping aliases.

What are the practical challenges?

Sharp anti-alias bandpass filter needed to reject all unwanted Nyquist zone content. ADC analog bandwidth must support the full carrier frequency (not just the sample rate). Clock jitter becomes critical: SNR_jitter = -20*log(2*pi*f_in*sigma_j). At 2.4 GHz with 100 fs jitter, SNR is only 50.4 dB.

Related Terms

← Bandpass Filter Band-Reject Filter →