exp
Full Member level 1
- Joined
- Jan 28, 2013
- Messages
- 97
- Helped
- 0
- Reputation
- 0
- Reaction score
- 0
- Trophy points
- 1,286
- Activity points
- 2,570
Hi all,
I am just wondering what are practical specifications of the components for a wideband transmitter (e.g. LTE 20MHz) in terms of:
1) DAC sampling rate
2) DAC reconstruction filter: Type, order, cutoff, image rejection
What puzzles me: Suppose we want to transmit LTE 20MHz (I know it's effectively 18, but let's go with 20). Then the effective bandwidth per I or Q branch is 10 MHz.
The sampling rate according to standard is only 30.72 MHz which suggests that the DAC will also run at 30.72 MHz. True? This is just 1/3 oversampling.
How an ideal zero-order-hold would give us images at 30.72MHz, 61.44MHz, 92.16 MHz etc. Having all 10 Mhz allocated, this means that the first image starts at 30.72-10 Mhz = 20.72Mhz.
This means the reconstruction filter would need to attenuate from 10MHz to 20.72Mhz.
Isn't this infeasible?
Suppose the cutoff frequency would directly be placed at the bandwidth (10MHz). First of all, this would already result in huge phase distortion, right?
Second, suppose we take a fifth order Butterworth filter (which I think is already impractical). Then the image would be attenuated not even by 30dB! (of course, it is a little bit more because the image is already a bit attenuated due to the sinc response of the ZOH).
How does that work in practice?
Thanks!
I am just wondering what are practical specifications of the components for a wideband transmitter (e.g. LTE 20MHz) in terms of:
1) DAC sampling rate
2) DAC reconstruction filter: Type, order, cutoff, image rejection
What puzzles me: Suppose we want to transmit LTE 20MHz (I know it's effectively 18, but let's go with 20). Then the effective bandwidth per I or Q branch is 10 MHz.
The sampling rate according to standard is only 30.72 MHz which suggests that the DAC will also run at 30.72 MHz. True? This is just 1/3 oversampling.
How an ideal zero-order-hold would give us images at 30.72MHz, 61.44MHz, 92.16 MHz etc. Having all 10 Mhz allocated, this means that the first image starts at 30.72-10 Mhz = 20.72Mhz.
This means the reconstruction filter would need to attenuate from 10MHz to 20.72Mhz.
Isn't this infeasible?
Suppose the cutoff frequency would directly be placed at the bandwidth (10MHz). First of all, this would already result in huge phase distortion, right?
Second, suppose we take a fifth order Butterworth filter (which I think is already impractical). Then the image would be attenuated not even by 30dB! (of course, it is a little bit more because the image is already a bit attenuated due to the sinc response of the ZOH).
How does that work in practice?
Thanks!