Evaluating high-speed RF converter front-end architectures

Article By : Rob Reeder

Here is a quick-start guide to the pitfalls of ADC front-end interface design along with some useful and familiar design approaches.

An earlier article titled “A close look at active vs. passive RF converter front-ends” discussed the advantages and disadvantages of amplifiers and baluns, along with how to improve phase imbalance when designing with radio-frequency (RF) converters. This article will leverage that information with a focus on defining and understanding the performance trade-offs when designing a new receiver front-end.

The article will also compare various active receiver front-end design approaches, including low noise amplifiers (LNAs), fully differential amplifiers (FDAs) and the classic passive wideband balun.

Comparing AC performance trade-offs

When comparing baluns, LNAs and FDAs to a single-ended-to-differential (S2D) amplifier like TRF1208, it’s important to go over the metrics involved when designing a wideband, high-performance analog-to-digital converter (ADC) interface. Here are five metrics that can help keep the design focused and on track if considered in advance:

  • The input impedance or voltage standing wave ratio (VSWR) is a unitless parameter that shows how much power is reflected into the load over the bandwidth of interest. The input impedance of the network is the specified value of the load, usually 50 Ω.
  • The bandwidth is simply the beginning and ending frequencies used in the system, typically –3 dB from some reference point.
  • The passband flatness is usually defined as the amount of fluctuation or ripple tolerable within the specified bandwidth: 1.0 dB or +5 dB, for example. These levels could be more or less defined with a slope.
  • In AC performance, for single-tone, signal-to-noise ratio (SNR) and spurious free dynamic range (SFDR) matter while third-order intermodulation distortion (IMD3) is important in two-tone settings.
  • The input drive level is a function of the bandwidth, input impedance, and VSWR specifications. This level sets the gain or amplitude required for a full-scale input signal at the converter. It is highly dependent on the front-end components—the balun, amplifier and anti-aliasing filter—and can be one of the most difficult parameters to achieve.

Just to be clear, these metrics encapsulate the entire front-end interface design, not just the ADC. Considering these metrics first may help you decide between an active or passive front-end.

Essentially, you only need to perform frequency sweeps consisting of front-end bandwidth, input drive and AC performance (SNR and SFDR) to quickly evaluate the differences in the overall front-end design.

Let’s look at five different front-end designs to compare these metric trade-offs, illustrated in Figure 1.

Figure 1 Five front-ends encompass designs based on balun only, an LNA, a balun plus an FDA, a single-ended FDA, and the TRF1208. Source: Texas Instruments

Next, Figure 2 shows the input bandwidth and input-drive-level trade-offs across a frequency of up to 10 GHz. The front-end bandwidth for each design gives an indication of –3-dB bandwidth and input drive level needed to reach –6 dBFS at 1.4 GHz. For example, looking at the TRF1208, it only takes a –16-dBm input signal to reach –6 dBFS of the ADC’s full-scale value. Conversely, it takes roughly +1 dBm to achieve the same level using a wideband balun. Between the two, this is a difference of 17 dBm of signal strength. The balun and wideband interface network create loss, and therefore drive up the noise figure number of the entire signal chain. The bottom line is that baluns create loss, which is also true of the LNA and FDA front-end designs, which include a balun for the S2D signal conversion.

Figure 2 Here is how frequency response looks like in five front-end designs. Source: Texas Instruments

Figure 2 illustrates the passband flatness, starting from approximately DC to 8 GHz. Even though all front-end designs can achieve 8 GHz, each has various peaks and valleys to contend with. In all fairness, it’s possible to nudge these peaks and valleys according to input network value changes and what the design ultimately calls for.

Baluns have loss, so the wideband balun interface will require higher signal-drive strength, with a +1-dBm signal level at the primary of the balun in order to achieve –6 dBFS on the ADC’s output. Since all other comparisons use an active amplifier device—all of which have inherently various gains—the input drive level required will be much less: anywhere from –5 dBm to –16 dBm. You could conduct further analysis and front-end work to “even out” the gains and input network losses. In the meantime, this information does give you some idea of what to expect before diving deeper into AC performance.

SNR and SFDR rankings

Conducting frequency sweeps over the same bandwidth captures SNR, SFDR and IMD3 performance. These are typical standard tests used to make comparison trade-offs when designing with high-speed converters.

Figure 3 shows the trade-offs in SNR between the various configurations.

Figure 3 The SNR values are shown for five front-end designs. Source: Texas Instruments

Looking at the purple curve as the baseline performance, you can see that the wideband balun interface offers the best SNR performance over the converter’s entire bandwidth. The green curve representing the LNA approach is second, as these types of active devices typically have a very low noise figure, with about 1 dB to 2 dB of added noise. The FDA comes in third place, as it has more wideband noise than the LNA, but less than the TRF1208. There is a slight issue with common-mode noise cancellation when using the FDA in a single-ended input configuration, since its inherent design on the input anticipates a fully differential signal. Using this type of configuration will slightly affect the SNR.

The TRF1208 comes in last; however, it has more output noise because it has a higher gain than the FDA. Remember that a higher active gain will have a tendency to gain the device’s own self-generated noise. For example, with a 2-GHz analog input signal, the TRF1208 has a gain equal to 16 dB and a noise figure equal to 8 dB at –166.7 dBm/Hz, yielding 150.7 dBm/Hz of output noise. The FDA has a gain equal to 10 dB (S2D) and a noise figure equal to 11 dB at –163.3 dBm/Hz, yielding –153.3 dBm/Hz of output noise.

All of the designs are configured for the widest bandwidth possible, as shown in Figure 2. In any active design, reducing the bandwidth by using an anti-aliasing filter in between the outputs of the amplifier and the inputs of the ADC will help reduce the wideband noise outside the bands of interest. It will also help reduce what noise the converter “sees,” therefore pushing the SNR back toward the baseline performance, as shown in Figure 1 (WB Balun + 5200RF ADC).

Figure 4 shows the SFDR dynamic range from a linearity perspective over a 10-GHz frequency sweep between various front-end configurations. SFDR is a single-tone measurement that provides a good perspective on any limiting harmonics—second harmonic, third harmonic, fourth harmonic—within the frequency of interest.

Figure 4 The SFDR values are shown for five front-end designs. Source: Texas Instruments

Looking once again at the purple curve as the baseline performance, you can see that the wideband balun interface will yield the best SFDR possible over the converter’s entire bandwidth. The green curve representing the LNA shows very degraded performance, particularly at the lower band of up to 5 GHz, as the even-order distortion (HD2) will always dominate given the single-ended nature of the LNA. The HD2 eventually falls out of the ADC’s bandwidth.

The FDA seems to have some third-order domination in the 0.5- to 3.5-GHz area when using the differential front-end approach. More even-order degraded dominance is evident in the 0.5- to 5-GHz range when using the single-ended approach.

The TRF1208 stays on par with the passive baseline front-end all the way, showing why this amplifier is a preferable choice when it comes to wideband front-ends that need an active device.

Two-tone measurements

Another common converter test metric, two-tone measurement gives rise to IMD3 results or third-order intermodulation distortion, and quickly emulates real-world system application signals. Simply put, two-tone measurements actively assess two signals injected into the front-end interface at the same time. These two signals are typically offset 10 MHz from each other and driven to the same level, or –7 dBFS each. Figure 5 shows IMD3+ (2 × F1 + F2 or 2 × F2 + F1) results. While captured, the figure doesn’t include IMD3– (2 × F1 – F2 or 2 × F2 – F1) for ease of clarity in illustrating the performance differences.

Figure 5 This is how IMD3+ looks like in five front-end designs. Source: Texas Instruments

With the purple curve again illustrating the baseline performance, you can see that the wideband balun interface will yield the best IMD3 performance possible over the converter’s entire bandwidth. The green curve, representing the LNA, shows degraded performance relative to the wideband balun interface. The blue and black curves representing the FDA interfaces are degraded in performance as well relative to the baseline, up to 5 GHz. The TRF1208 stays on par with the passive baseline front-end for the entire frequency sweep. Again, it shows why this amplifier is a preferable choice when it comes to wideband front-end needs.

Additionally, the evaluated FDA has two power supplies, one negative, and consumes a whopping 1.8 W of power in order to keep the noise low. This is a classic way to drive the noise down and increase the amplifier’s headroom and throw more power at the design. The LNA dissipates the least amount of power; only 0.275 W with a single 5-V supply. The TRF1208 runs off a single 5-V supply and consumes just 0.675 W.

The goal of this article was to provide a quick-start guide to the pitfalls of ADC analog front-end interface design, along with some useful and familiar design comparisons and an introduction to the new TRF1208 differential amplifier. With any new wideband front-end design, it’s recommended to evaluate metric trade-offs and carefully plan upfront. Mind the phase imbalance, as it can wreak havoc if even-order distortions are in the application’s frequency plan. Given the characteristics of baluns and amplifiers and their pros and cons, it’s important to review the trade-offs and choose wisely.

This article was originally published on Planet Analog.

Rob Reeder, application manager for high-speed converters at Texas Instruments, is author of Signal Chain Basics blog # 174 for Planet Analog.


Leave a comment