The AK2401 uses a direct-conversion architecture instead of the more standard heterodyne approach.

A fundamental requirement in any RF system is translating the received signal from the RF frequency to the baseband frequency. The former is the higher frequency corresponding to the carrier signal; the latter is the frequency of the original information-carrying signal. Wireless transmission requires the RF frequency, but the signal must be returned to the baseband frequency before it is sent to the data-processing circuitry.

A very common receiver architecture involves an intermediate frequency (IF) that is somewhere between the RF and baseband frequencies. This is referred to as a heterodyne system; RF receivers of this kind are very common, and there are significant benefits associated with this approach.

It is interesting, then, to see that AKM (Asahi Kasei Microdevices) is emphasizing the fact that their new AK2401 RF transceiver does not employ a heterodyne architecture. Instead it uses the ostensibly simpler “direction-conversion” architecture, i.e., it converts the received RF frequency directly to baseband. You can see this in the receive-chain section of the block diagram:


Diagram taken from the AK2401 datasheet.


So the RF signal is received from the antenna and then amplified by a low-noise amplifier (LNA). It then goes to an external matching circuit before reaching the mixer. The mixer translates the signal from the RF frequency directly to the baseband frequency; the result is differential I/Q baseband signals that are amplified and passed through an anti-aliasing filter before being digitized by the 24-bit delta-sigma ADC.

The AK2401 is marketed as a high-performance, highly integrated transceiver IC for narrowband devices such as handheld two-way radios, public-safety radios, maritime radios . . . basically any type of “you talk to me, I talk to you” system. “Narrowband” basically translates to “low data rate,” since bandwidth corresponds to the rate at which information can be transferred. Voice communication does not require a lot of bandwidth, so it’s a natural application for a chip such as this one, but I don’t see why you couldn’t use the AK2401 for generic data transfer.


But Why?

If the heterodyne approach is so standard and advantageous, why did AKM decide to use a direct-conversion architecture?

Well, the information from AKM identifies only one benefit in this particular design context, but it’s an important one: improved integration, which translates to smaller form factor, which translates to smaller radios.


It’s small: QFN package, 7 × 7 mm.


Direct conversion involves fewer components, and most importantly, it does not require bulky image-reject filters. Heterodyne receivers suffer from the well-known problem of image signals that must be suppressed via filtering. As far as I know these image-reject filters are, in general, not compatible with IC techniques, and consequently they must be added by the board designer as off-chip components.


The DC Offset

If direct conversion is so beneficial in this one important aspect, why isn’t it more common? Why is AKM able to present the AK2401 as something new and unusual on account of its direct-conversion architecture?

A major difficulty is DC-offset cancellation. The inputs to the direct-downconversion mixer are the local oscillator (LO) signal and the received RF signal. However, it is possible for the LO to leak through to the RF input of the mixer, and mixing the LO with the LO results in a DC offset in the output signal.


This diagram depicts one common cause of DC-offset generation in a direct-downconversion receiver.


This problem doesn’t exist in heterodyne architectures because the IF spectrum doesn’t extend to DC, and thus DC offsets are easily removed via filtering. Direct-conversion produces a spectrum that extends to DC, so if you try to use a high-pass filter to remove the DC offset, you inevitably attenuate portions of the information-carrying signal.

AKM addresses this problem by means of an RDOC (Real-time DC-Offset Cancellation) system. It’s based on proprietary technology and I assume that they don’t publish many details about how this technique is actually implemented. This RDOC takes place in the digital realm, more specifically, as part of the “DC OFFSET CAL” block:


Diagram taken from the datasheet.


Somehow or another the RDOC algorithm removes the DC offset “in response to changes in the surrounding radio wave environment.” I like that the RDOC functionality is completely integrated and automatic—no designer involvement required. However, I did notice one little caveat in the datasheet (page 59): RDOC is “effective” for received signals that do not exhibit amplitude fluctuations (i.e., it works with frequency- or phase-modulated signals). Page 1 of the datasheet indicates that the AK2401 is compatible with QAM (quadrature amplitude modulation), but maybe QAM signals will be more susceptible to DC-offset problems.



Feel free to leave a comment if you have any thoughts on the heterodyne vs. direct-conversion issue.