With the advent of higher broadband speeds in the fixed-line and wireless systems, the need for higher-performing data converters has become apparent. In this article, we discuss the different standards that are pushing this trend and how single-tone testing, in some cases, does not accurately determine the performance in communication systems. This is particularly true in systems which employ multi-tone signaling or OFDM (orthogonal frequency division multiplexing), where the available bandwidth is split into a large number of sub-channels, which then allows data to be transmitted over many sub-carriers, Figure 1a and Figure 1b, thereby enabling higher data rates.
Figure 1: a) OFDM Signal (Time and Frequency);
b) OFDM output spectrum
(Click on images to enlarge)
3GPP LTE requirements
One application driving the requirements for higher-performance data converters is 3GPP Long Term Evolution (LTE). The LTE represents a major advance in cellular technology allowing higher download and upload speeds for mobile devices, thereby enabling true broadband mobile devices.
A major problem with OFDM-based systems is their high peak-to-average power ratios (PAPR),
which means that the OFDM signal has a large variation between the average signal power and
the maximum (or minimum) signal power. This tells us how well the signal is distributed over the amplitude range.
This is why LTE has chosen Single-Carrier FDMA (SC-FDMA) as its uplink technology to overcome the challenges faced by Mobile WiMAX (802.16e) using OFDMA (Orthogonal Frequency-Division Multiple Access), but the downlink still uses pure OFDM. LTE channel bandwidth is between 1.4 and 20 MHz, of which around 19 MHz of contains signal, whereas WiMax bandwidths are only up to 10 MHz.
ADCs and DACs for LTE
LTE standard could have bandwidths up to 20 MHz. For reasonable filter requirements, an ADC
sampling rate of 40 megasamples/sec (MSamples/sec) is a typical requirement, irrespective of the architecture employed.
As stated earlier, OFDM is still used in the down link for the LTE application, in which large PAPR is possible if all sub-carriers reach their peak amplitude simultaneously. If this occurs. then this can introduce more distortion compared to single channel application [Reference 1]. This higher distortion requires the need for higher signal-to-noise ratio (SNR) in order to have same quality of performance.
The number of subcarriers used in the system determines the required SNR that must be achieved. For large PAPR it is envisaged that a mobile terminal with greater then 60dB of dynamic range is required. Lower dynamic range requirement is possible if PAPR is reduced [Reference 2], but higher PAPR is required in order to improve spectral efficiency [Reference 3], thus resulting in low PA (power amplifier) efficiencies.
As mentioned previously, the LTE standard uses SC-FDMA in its uplink, thereby reducing the
PAPR requirement by much as 50% and maximizing the SNR requirement of the DAC [Reference 3]. As this is the case, then the same typical requirements for 802.11x and 802.16e are applicable for LTE.
ADCs and DACs for DOCSIS
Another application driving increased data converter specifications is DOCSIS 3.0 (Data Over
Cable Service Interface Specification) which gives cable-modem designers the option of implementing a wideband design. The DOCSIS downstream-channel spectrum can occupy up to a 100 MHz bandwidth [Reference 4] (0 to 32 MHz, 30 to 48 MHz, 40 to100 MHz) if a wideband tuner is used instead of four narrowband tuners. Within this wide bandwidth, of which the desired channels only constitute a fraction, the remaining undesired signals consume a substantial portion of the available signal path dynamic range. For EuroDOCSIS, the same procedure is being used, but with the following changes: the tones are: 5 MHz, 35 MHz, and 65 MHz S-CDMA mode.
Compared to the previous standard, this system requires a much higher-performance ADC and DAC. The ADC sample rate needs to be well above the Nyquist frequency (100 MHz) and a clock rate above 200 MHz would be required. Due to this large bandwidth, undesired signals could propagate into the spectrum, which increases the requirements on the ADC SNR. In order to accommodate the additional signal power and also PAPR considerations will further increase the margin necessary in the ADC's specifications [Reference 5].
The DAC also needs to cope with larger bandwidths and requires the SNR to increase by 6 dB,
compared with LTE. But because of these higher bandwidths, the DAC has to function at much higher frequencies, possibly up to 85 MHz for the extended DOCSIS, which is high compared to 10 to 20 MHz for WiMax and LTE. At these high frequencies, the DAC is required to have large dynamic range.
Data converter testing
Numerous communication systems use discrete multi-tone transmission schemes (DMT, OFDM,
etc), in which the available bandwidth is split into a large number of sub-channels. Traditionally, dynamic performance of high-speed data converters has been evaluated by analyzing the FFT (in case of ADCs) and the output spectrum (in case of DACs) with a single-tone sine-wave.
This is not enough to assess the data converter performance in modern communication applications. This is because the characteristics of the reconstructed multi-tone (carrier) spread spectrum are far different from that of a simple, full scale sine wave.
This is why multi-tone power ratio (MTPR) is an important feature in the evaluation and design of DMT systems [Reference 6] in communication systems. The MTPR specification determines dynamic range from peak power to peak distortion, Figure 2a and Figure 2b. As the OFDM signaling scheme uses narrowband carriers which are transmitted in parallel at different frequencies, this can result in subcarriers which could potentially add up to create distortion.
Figure 2: OFDM signal with equally spaced at tones using different PAPR
a) complete OFDM signal;
b) MTPR measurement in the low-band
(Click on images to enlarge)
In other words, better MTPR performance, in both the transmission and receive paths, results in higher data rates throughout. In these applications, the SNR of the reconstructed waveform--which includes the effects of both noise and distortion--will directly affect the bit error rate (BER) of the system.
Another specification that has gained popularity is adjacent-channel power ratio (ACPR) or
adjacent-channel leakage ratio (ACLR), which are used to determine the performance of a data
converter for certain applications such as DOCSIS 3.0. ACPR is the ratio of the reconstructed
signal power to the power measured in an adjacent channel measured in dB.
As an example, Figure 3a shows single carrier spectral response for a 12-bit DAC [Reference 7], for a single carrier centered at 5 MHz. The ACPR performance for a four-carrier signal is also shown, in Figure 3b. Since the signal power level, signal bandwidth, channel spacing, carrier frequency, and data converter sampling are all application-specific and affect the measurement results, the test conditions should be noted when comparing data converter IPs (intellectual property). The ACPR performance of a data converter is ultimately limited by its noise floor and distortion performance.
Figure 3: ACPR spectral representation
a) one strong carrier at 5 MHz;
b) and containing four strong carriers with the first carrier at 40MHz.
(Click on images to enlarge)
1. Louis Litwin and Michael Pugel, "The principles of OFDM", http://mobiledevdesign.com/tutorials/radio_principles_ofdm/.
2. Mikael Gustavsson, J. Jacob Wikner and Nianxiong Nick Tan, "CMOS Data Converters for
3. Tim Haynes, "Designing energy-smart 3G base stations", http://mobiledevdesign.com/hardware_news/radio_designing_energysmart_base/.
4. Curtis Ling and Patrick Tierney, "Advances in CMIOS TV tuners make multi-tuner DOCSIS
3.0 cable modem design a reality", http://www.embedded.com/columns/technicalinsights/205917195?_requestid=183008.
5. Rudolf Tanner, Jason Woodward, "WCDMA--Requirements and Practical Design".
6. Franco Maloberti, "Data Converters".
7. S3DA300M12BT65D (S3 Mixed-Signal IP Part Quoted in this Article)
8. Steve Halford, "OFDM Uncovered", http://www.commsdesign.com/design_corner/showArticle.jhtml?articleID=16504605.
About the author
Michael Connolly is a mixed-signal design engineer at Silicon & Software Systems (S3). He has been involved in Mixed Signal design in S3 for over 9 years, working extensively in the wireless and consumer sectors. His expertise is in data converter ICs and his interests are in pipeline ADCs and current-steering DACs. Michael holds an MSc and a BTech from Waterford Institute of Technology, Ireland. He is a member of the IEEE and currently works from the San Jose office in technical support for Mixed Signal IP and can be reached firstname.lastname@example.org .