For instance, improvements in sample rate and resolution of a/d converters simplify the design of multicarrier, multistandard software defined radios. Similarly, advanced radars require better a/d converter sample rates and resolution to improve their sensitivity and accuracy.
In communications applications, using a single a/d converter to digitise entire spectral bands, rather than a limited number of channels, can simplify receiver design. To do this, the entire spectral band must fit within a single Nyquist zone of the converter, meaning the sample rate (Fs) must be at least twice the spectral bandwidth (BW) of interest.
However, a significantly higher sample rate simplifies the necessary antialiasing filters and the preceding stages in the receiver. For example, it is possible to digitise a 75MHz GSM band using a sample rate of 184.32Msample/s with highly selective filters. Limiting the filter to third order requires a/d converter drivers with second harmonic distortion (H2) better than -75dBc.
In Figure 1, the distance from the band edge to the first interfering harmonic that will alias back in band is only 25.74MHz. Increasing the sample rate to 491.52Msample/s (Fig 2) moves the closest alias image 140.04MHz away from the band edge, relaxing the combined filter plus driver requirements.
Fig 3 illustrates the frequency response of two hypothetical third order hourglass filters. For the case of 491.52 Msample/s, the increase of 114MHz in the distance from interferer to band edge allows the filter to generate much larger stop band attenuation.This allows the use of a lower power a/d converter driver with 23.5dB less second harmonic performance.
Furthermore, oversampling the spectral band of interest allows more flexible frequency planning. In many cases, oversampling will make it possible to position the aliased second and/or third harmonics out of the band of interest. This capability enables a given 14bit, 500Msample/s converter's spurious free dynamic range (sfdr) performance to improve by 8dB over a 55MHz bandwidth, centred at 194.5MHz.
A high resolution converter capable of sampling at 491.52Msample/s would allow a 75MHz band of interest centred on 122.88MHz to be free of all aliased second harmonics (fig 4). Because none of the second harmonics of the 75MHz band alias into the band of interest, they can be filtered in the digital domain. This further eases the filter and converter driver requirements, reducing cost and complexity.
An additional benefit of oversampling is an improvement in the noise floor in the desired channel. This can be understood by looking at the noise in per Hz units, instead of the typical Nyquist bandwidth signal to noise ratio (snr). Engineers can translate between the two using:
snr[dBFS/Hz] = snr[dBFS/Nyquist] + 10*log10(Fs/2)
For the 491.52Msample/s rate, an a/d converter that provides a Nyquist snr of 73dBFS offers an snr/Hz of 156.9dBFS/Hz. The output noise of an a/d converter with the same Nyquist snr sampling at 184.32Msample/s is 152.6dBFS/Hz. The increased sample rate provides an improvement in snr of 4.3dB for channel widths less than the Nyquist frequency of the lower sample rate converter. Alternatively, the snr improvement due to oversampling can be calculated as 10*log10(Fs1/Fs2). For Fs1=491.52Msample/s and FS2=184.32Msample/s, the improvement is the same 4.3dB.
Previous converter generations, while having lower sample rates and less dynamic range, consumed several Watts, which had implications for system performance. Finding a converter that can meet the performance targets while consuming less than 1W relieves additional design constraints.
One obvious implication is energy use. For applications that are rolled out in their tens of thousands, every mW counts. For instance, savings 1.5W per element in a network of 36,000 4G basestations would save 50kW. Furthermore, while technology dedicated to infrastructure is not typically thought of as battery powered, these systems often require battery back up.
Clearly, a/d converter advances translate into benefits for a variety of applications. For example, a 14bit, 500Msample/s a/d converter consumes less than 1W of dc power. This allows a base station receiver to digitise the entire 75MHz GSM band and use a low power a/d converter driver and simple channel filtering while avoiding the need for high current low drop out regulators and bulky heat sinks and fans.
Similarly, the same a/d converter would allow advances in state of the art phased-array radar design. Its very high sample rate and resolution allow improved sensitivity while the decrease in power consumption allow simplified thermal design, especially where the element spacing is constrained to be small.
Thus, improvements in a/d converter technology are enabling advancement in performance, reliability, and cost of systems that are dependent upon them.
Intersil is introducing a family of a/d converters that meet the demanding requirements of these applications. The first in the family is the ISLA214P50IRZ, a 14bit 500Msample/s converter that consumes 925mW while providing an snr of more than 73dB. With its high sample rate, high dynamic range, and low power consumption, this part is suited to applications such as broadband communications, high performance data acquisition, power amplifier linearisation and communications test.
Many of these benefits come from FemtoCharge, a charge domain signal processing technology developed by Intersil. The fundamental idea is that signals are represented as quantities of charge, rather than voltages. While this may sound like a subtle difference, it has a large impact on the power dissipation of analogue signal processing circuits, like an a/d converter.
This is because a voltage is not a physical entity; rather, it is a measure of electrical potential. In contrast. a quantity of charge is the physical entity making up that potential. As an analogy, voltage could be 100ml of water, whereas charge is the water molecules themselves.
In pipelined voltage based a/d converters, the signal must be recreated each time a signal processing operation is performed. In other words, each time a voltage based a/d converter wants to use its signal – 100ml of water – it has to pour a new glass of precisely 100ml. When finished, the 100ml is thrown away. However, in charge based converters, the signal – or water molecules – is reused for each mathematical operation; in essence, it is poured from glass to glass.
By conserving the water molecules for subsequent operations, the a/d converter is much more efficient. Since the signal is never recreated, care has to be taken not to 'spill' any of the 'water' when moving it from one glass to another – and that is the challenge in Femtocharge. However, in general, it requires much less power to transfer the signal accurately, rather than recreate it from scratch.
Edward Kohler is senior strategic marketing manager for high speed data converters , Mark Rives is a principal applications engineer supporting high speed data converters and Dave Carr is applications manager for data conversion products. All are with Intersil.
