hf/rf ZONE Products for the week of November 22, 2004
Linear Technology Says . . .
LT5528: Direct Conversion Modulator Provides Compact
Basestation Performance
A new direct conversion modulator from Linear Technology enables the design of next generation, less costly 3G wireless basestation transmitters. The LT5528 is a direct I-Q modulator with uncompromising performance, enabling more compact basestation transmitter designs at lower system cost. The LT5528 is targeted for high performance wireless infrastructure applications, including W-CDMA, TD-SCDMA, GSM, and PHS. It is also suitable for broadband fixed wireless access, as well as point-to-point radio links and other high performance radios.
The LT5528 incorporates a matched pair of double-balanced mixers which form a carrier suppressed, image reject I-Q (In phase and Quadrature phase) modulator. The I and Q inputs accept complex modulation signals from the baseband circuitry, and upconvert them directly to RF (1.5GHz to 2.4GHz) using a LO (Local Oscillator) operating at the same frequency. This effectively eliminates one to two stages of frequency upconversion in the radio path, greatly reducing the number of external components and lowering costs.
The LT5528's I/Q gain matching (to within +/- 0.1dB) results in a typical carrier suppression of -47dBm at 2.14GHz. The precision 0°/ 90° phase shift of the on-board LO phase splitter produces 45dBc image rejection at 2.14GHz. The modulated RF I and Q signals are combined in an on-chip balanced RF transformer at the single-ended output. The LT5528's 50-Ohm RF port eliminates the need for external impedance matching components. All other signal inputs and outputs are also 50-Ohm matched, thus simplifying the design task.
Careful attention is paid to the design and biasing of the mixer to optimize
modulator linearity, resulting in an OIP3 (Output Third-Order Intercept)
of 20.5dBm at 2.14GHz. Combined with a low noise floor of -159dBm/Hz, the
LT5528 achieves a best-in-class ACPR (Adjacent Channel Power Ratio) performance
of -66dBc and an Alternate Channel Power Ratio of -68dBc, measured with
a 4-channel W-CDMA signal. Single channel ACPR performance extends higher
still to -73.5dBc. This ensures ample design margin for basestation manufacturers
to comply with the standards specifications.
analogZONE Says . . .
It's easy enough to build a direct conversion modulator; heck, I made one 30 years ago. But the trick is to make it work! The nasty little problem of the local oscillator being at the same frequency as the output has dampened many hardy souls. We've seen direct conversion work being described in increasing amounts at technical conferences but it hasn't been until this year that we have seen monolithic solutions in production. TI announced the SiGe TRF3702 (covering much the same range as the LT5528) in March 2004 but analogZONE did not review it at the time because we did not like the relatively high LO output with suppression at -35 dBc, we didn't like the output impedance range of which just the real part covered something like 27 ohms to 80 ohms over frequency, and at the time no pricing was disclosed -- which suggested to us that it was being announced a little prematurely. We don't see the same issues with this LTC part.
The I and Q channels are differential throughout but the LO input is single-ended with a typical drive level of 0 dBm; S11 at the port is a nice -17 dB, worst case, and goes a little over -20 dB over part of the higher spectrum.
The killer difference with the LT5528 is that there is an output balun on board. It is optimized for about 1.9 GHz where S22 approaches -40 dB (when the part is enabled) with skirts falling to about -11 dB at 1.5 GHz and -13 dB at 2.5 GHz, very workable numbers. Nominal output power is -2.1 dBm with 1-V I and Q input signals with a 1-dB bandwidth of 1.7 GHz to 2.2 GHz, and -3 dB from 1.5 GHz to 2.4 GHz. The noise floor is at a fairly typical -159 dBm/Hz, conversion gain is -6.5 dB in power terms and -6 dB in voltage terms. The unadjusted image rejection is quoted at an incredible -45 dBc and carrier leakage is -42 dBm at 2 GHz. Output P1dB is 7.9 dBm (using an output coupling capacitor) while the baseband input P1dB is 3.2 Vp-p differential. The baseband bandwidth is a more than sufficient 400 MHz.
This is a 5-V part with a typical quiescent of 125 mA and a shut-down of 50 nA (LT can do sleep modes so well) although turn-on time is a relatively slow 250 ns. The data sheet does an excellent job of characterizing leakages and distortions over the frequency band.
I would be extremely comfortable using this part in a baseband design and it is in an extremely small package. Even the coupling capacitors for the LO and the single-ended RF output are also on chip. The part will require a calibration loop back to the DSP in use so that the dc offsets on the I and Q channels can be iteratively adjusted to maximize IO cancellation. The unadjusted numbers are so good already that this should be a breeze. From a design point of view most users will have the worst problems matching input currents to minimize output distortions and a belt-and-braces input transformer approach may be the easiest for a lot of designers. I have no doubt that Linear will offer some of the best help around.
The LT5528 will be outstandingly successful and is priced to be taken very seriously.
The LT5528 is in production in a 4 mm x 4 mm QFN-16 and is priced at $5.95 in 1000-piece lots.