Calibration Method for SVDMIMO Revision 1'00

1
Calibration Method for SVD-MIMO-Revision 1.00-
Tomoya YamauraInformation Technology Lab.Sony
Corp.July 21, 2004
2
Background

• Our SVD is assuming channel reciprocity.
• It requires calibration function at both end.
• Here I would like to explain how to calibrate
  nodes briefly.

3
Notation

• H channel matrix
• H U D VH
• TA Error matrix of Tx-side in node-A (diagonal)
• RA Error matrix of Rx-side in node-A (diagonal)
• TB Error matrix of Tx-side in node-B (diagonal)
• RB Error matrix of Rx-side in node-B (diagonal)

4
One way to calibrate

• As Qualcomm proposed, both node are involved to
  do full-calibration.
• In this method, Error matrices might become
• TA RA TB RB I (unit matrix)
• H (and eigen-value matrix D) becomes identical to
  that of perfect transmitter/receiver.
• To do this, MAC support and transmitting frames
  including calibration data are required.
• But it is not necessary.
• Actually, I dont investigate it deeply.

5
My proposing way to calibrate

• What we really need is
• TA RA, TB RB
• Basically, identical to method of IMEC
• Detail would be provided later. Sorry
• This implies self-calibration is enough.
• Derived H and D are not identical to those of
  ideal case, but it is not matter.
• MAC support to send calibration information is
  NOT required.

6
Options to don self-calibration

• Option (1) IMEC method (11-04-0792-00-000n)
• Having Reference Tx and reference Rx
• Connecting them to Rx branches and Rx branches
  each other via coupler.
• By the way, IMEC said that calibration at AP side
  is enough, but it is for SDMA system only. For
  SVD, we need self-calibration at both end.
• Option (2) Sonys proposal.
• Option (2-1) On the fly using L-LTF
• Option (2-2) CTS-to-self and calibration with
  air.

7
Option (1) IMEC method
Tx-1
IFFT-1
Cal.
Coupler
SW
Tx-2
IFFT-2
Coupler
Ref-Rx
Ref-Tx
Rx-1
FFT-1
Coupler
SW
Rx-2
FFT-2
Coupler
8
Option (2-1) On the fly
Tx-1
IFFT-1
Cal.
SW
Tx-2
IFFT-2
Rx-1
FFT-1
Att
SW
Rx-2
FFT-2
Att

• RCVR(s) receive legacy preamble from single
  antenna
• It shows concept only. Detail might be wrong
• Averaging might help interference problem.

9
Option (2-2) CTS-to-self
Tx-1
IFFT-1
Cal.
Att
SW
Tx-2
IFFT-2
Att
Rx-1
FFT-1
SW
Rx-2
FFT-2

• After CTS-to-self, Tx branches transmit fixed
  pattern (preamble)
• RCVR(s) receive them
• It shows concept only. Detail might be wrong
• Averaging might help interference problem.

10
Comparison

• Which options do you prefer ?

(1) I
(2-1) On the fly
(2-2) CTS-to-self
Cal Accuracy
Best
So so (2x2 case, coupler may work well)
Additional H/W
many
So so (to avoid saturation)
Rx side
Tx side
MAC support
Required
None
CTS-to-self
required airtime
Very small
None
Very small
11
Other topics

• How often ?
• Once per day might be fine, if TempCo is small.
• What is the trigger?
• Deterministic Time-out, or Power-on only
• Adaptive Check non-diagonal elements
• What is the required accuracy ?
• Sorry but I dont have any data yet.
• IMEC says
• 2deg in phase, 0.3dB in power
• For 1dB degradation at un-coded BER10-2
• Our previous experience of Adaptive Array
  Antenna, slightly more cal. errors are allowed.