Professional Documents
Culture Documents
Lecture 16 ECE265A - Rx3 Homodyne-B
Lecture 16 ECE265A - Rx3 Homodyne-B
Lecture 16
Receiver Architecture (3)
Direct-conversion Rx (b)
Vincent Leung
Direct-conversion receivers
o LO leakage, DC offsets (last lecture)
o 2nd-order distortion
• Mixer Feedthrough,
o Flicker (1/f) noise
o IQ (LO) mismatches
o Mixing spurs (not important for homodyne Rx)
Beat component
(corrupting the down-
converted spectrum)
1
0
1 𝟐 2 2
𝑆(𝑡) = + cos 𝜔 𝑡 − cos 3𝜔 𝑡 + cos 5𝜔 𝑡 + …
2 𝝅 3𝜋 5𝜋
© Leung, ECE265A Winter 2019 4
Single-balanced mixer
Known as the “single-balanced” mixer, the differential LO “commutates” a
single input to two outputs
o Input is “single-ended”, LO/ output are balanced (differential)
𝟏
o , 𝟐
𝟏
o , 𝟐
o , ,
_
1
-1
𝟒 4 4
cos 𝜔 𝑡 − cos 3𝜔 𝑡 + cos 5𝜔 𝑡 + …
𝝅 3𝜋 5𝜋
© Leung, ECE265A Winter 2019 5
2 Nonideal feedthrough mechanisms
The cancellation is sensitive to asymmetries. If there is any mismatch, a net
feedthrough can arise in the differential output
For example, the switches may exhibit a mismatch between their on-
resistances ( , , ):
o , , 𝒓𝒐𝒏,𝒑 𝒓𝒐𝒏,𝒎
+ 𝑇 > 𝑇 ⁄2
_ 𝑣
𝒓𝒐𝒏,𝒎 𝑇
𝑇−𝑇
𝑉 (𝑡)
* Please do not confuse this 𝛼 𝑥+𝛼 𝑥 𝐿𝑃𝐹
with the “cascaded” IIP2 formulation 𝑦(𝑡)
discussed in Lecture 9, page 19-20.
cos 𝜔𝑳𝑶 𝑡
Beat component
rejected
o ,
o , [Note: ]
o ,
Apparently, the 1st term is the flicker noise contribution, while the second
term the thermal.
The flicker noise raises the thermal noise (evaluated over the signal
bandwidth) by the ratio of:
.
,
o
, _ 𝜶 = 𝑺𝒕𝒉 𝒇𝒄
𝑷𝒏,𝒕𝒐𝒕𝒂𝒍 𝒇𝒄 𝒇𝒄
o
𝑷𝒏,𝒕𝒉_𝒐𝒏𝒍𝒚 𝒇𝑩𝑾 𝒇𝑩𝑾
The higher the ratio, the higher the “flicker noise penalty”
Example: assume a flicker noise corner frequency of 200 kHz
o For baseband bandwidth = 10 MHz, (say, for 20MHz LTE)
, . .
• (Not too big a deal)
, _
A good receiver design will maximize the gain in the RF frontend (linearity
permitting), such that 𝒕𝒉 is dominated
For good Rx design,
by the source noise, and to lesser this is dominated by
extent, the LNA and mixer noise. “gained up” source
noise:
Therefore, the higher the RF gain, the (~4𝑘𝑇𝑅 𝐺 𝐺 )
higher the 𝒕𝒉 *, the lower the 𝒄 (less
flicker noise penalty)
* Thus relatively less baseband circuit noise contribution. To be
studied in Rx “level diagram”, time permitting.
© Leung, ECE265A Winter 2019 12
Flicker noise for GSM
o , _
Therefore, the so-called “flicker noise penalty” is:
, .
o
, _ 𝜶 = 𝑺𝒕𝒉 𝒇𝒄
As expected, the penalty is
much more severe than the
previous 2 examples.
o The solution is called
“low-IF” receiver, to be
discussed. 𝒇𝒄
© Leung, ECE265A Winter 2019 𝑩𝑾 < 𝒇𝒄 13
I/Q Mismatch in quadrature downconversion
Let’s model the gain and phase mismatches on the LO by (where the factor
of 2 is only to simplify the results later):
∈
o ,
∈
o ,
𝒙𝑳𝑶,𝑰
𝒙𝑳𝑶,𝑸
o the baseband symbols are corrupted by the symbols in the other output
Q