Scribd
Upload
449 views3 pages

Correcting I-Q Imbalance in Direct Conversion Receivers: S.W. Ellingson February 10, 2003

This document presents a method for correcting I-Q imbalance in direct conversion receivers. It describes how DC offsets (βI and βQ) can be removed from the I and Q outputs. It then shows that the remaining imbalance is characterized by amplitude error (α) and phase error (ψ). It provides equations to calculate α and ψ from the mean and cross-correlation of the I and Q outputs over time. Finally, it lists the complete 7-step algorithm for identifying and correcting the I-Q imbalance at a given baseband frequency using these parameters.

Uploaded by

andreic11
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
449 views3 pages

Correcting I-Q Imbalance in Direct Conversion Receivers: S.W. Ellingson February 10, 2003

This document presents a method for correcting I-Q imbalance in direct conversion receivers. It describes how DC offsets (βI and βQ) can be removed from the I and Q outputs. It then shows that the remaining imbalance is characterized by amplitude error (α) and phase error (ψ). It provides equations to calculate α and ψ from the mean and cross-correlation of the I and Q outputs over time. Finally, it lists the complete 7-step algorithm for identifying and correcting the I-Q imbalance at a given baseband frequency using these parameters.

Uploaded by

andreic11
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Correcting I-Q Imbalance in Direct Conversion

Receivers
S.W. Ellingson∗
February 10, 2003

1 Introduction
Analog direct conversion receivers are notorious for I-Q imbalance. Consider the
conversion of a single tone at RF to baseband. Ideally, the I and Q outputs of the
receiver are
I(t) = cos (ωt) and (1)

Q(t) = sin (ωt) , (2)

respectively. ω is the baseband frequency of the tone. With no loss of generality, we


have normalized the magnitude to unity and the phase to zero, as these quantities
are not relevant to this discussion. In contrast, a realistic direct conversion receiver
produces:
I 0 (t) = α cos (ωt) + βI (3)

Q0 (t) = sin (ωt + ψ) + βQ (4)

where ψ is the phase error, which we have assigned to the “Q” path, α is the magnitude
error, which we have assigned to the “I” path, and βI and βQ are the DC biases
associated with each path. The allocation of error mechanisms to paths is completely
arbitrary and implies no loss of generality.
∗
ElectroScience Laboratory, The Ohio State University, 1320 Kinnear Rd., Columbus, OH 43212;
Voice: (614) 292-7981, Fax: (614) 292-7297, E-mail: [email protected].

1
2 Theory
Correcting βI and βQ is very simple. For example, βI is simply the mean of I 0 (t) over
an integer number of periods. Given this estimate, the correction is simply a matter
of subtracting βI from the “I” path signal. The process is the same for the “Q” path.
Then, we are left with:
I 00 (t) = α cos (ωt) (5)

Q00 (t) = sin (ωt + ψ) (6)

This can be rewritten in matrix form as:


· 00 ¸ · ¸· ¸
I (t) α 0 I(t)
= (7)
Q00 (t) sin (ψ) cos (ψ) Q(t)

where we have used the identity:

sin (wt + ψ) = sin (ωt) cos (ψ) + cos(ωt) sin (ψ) (8)

Thus, we find that the correction is:


· ¸ · ¸ · 00 ¸
I(t) α−1 0 I (t)
= (9)
Q(t) α−1 tan (ψ) sec (ψ) Q00 (t)

Thus, it remains only to find α and ψ. To find α, let us first define:


Z t
1
< x(t) > = x(u) du (10)
N T t−N T

where T is the period 2π/ω and N is any integer greater than zero. Note that
¿ À
00 00 2 2 2 1 1 1
< I (t)I (t) >= α < cos (ωt) >= α + cos (2ωt) = α2 (11)
2 2 2

and by similar analysis

1
< I 00 (t)Q00 (t) >= α2 sin (ψ) . (12)
2

Thus, Equation 11 can be used to find α, and then Equation 12 can be used to find
sin (ψ). Assuming |ψ| < π/2 (hopefully any phase imbalance would be much smaller
than this!), we can obtain cos (ψ) directly from sin (ψ).

2
3 Summary
Here’s the complete algorithm for identifying the I-Q imbalance and correcting it:

1. Apply a tone that appears at baseband frequency ω. Note the measurement


and correction will only be valid at this frequency; however, these parameters
typically vary slowly with frequency.

2. Compute βI =< I 0 (q) > and βQ =< Q0 (t) >.

3. Compute I 00 (t) = I 0 (t) − βI and Q00 (t) = Q0 (t) − βQ .


p
4. Compute α = 2 < I 00 (t)I 00 (t) >.

5. Compute sin (ψ) = (2/α) < I 00 (t)Q00 (t) >


p
6. Compute cos (ψ) = 1 − sin2 (ψ).

7. Compute the correction matrix parameters:


A = 1/α
C = − sin (ψ)/(α cos (ψ))
D = 1/ cos (ψ)

8. The correction can now be applied to subsequent data as follows:


· ¸ · ¸· 0 ¸
I(t) A 0 I (t) − βI
= (13)
Q(t) C D Q0 (t) − βQ

You might also like