Designing the analog input interface for a high-speed converter is the most daunting task throughout the product design process. Even though 10-Gsps high-speed converters can deliver as much as 10 GHz of bandwidth, most of the limitations involve a front-end design.
As both amplifier and balun specifications catch up in bandwidth and performance, it’s important to understand some of the nuances involved when it comes to interfacing the front-end. This article will review some of the most significant considerations and trade-offs when designing a high-speed, wideband converter front-end.
Amplifiers are active, provide gain, have inherent noise, need a power supply and decoupling, and consume power. A balun, however, does not add noise nor consume power. These are just fundamental differences; so, it’s too early to decide between them just yet. What’s important is understanding the trade-offs set by the application (Figure 1).
Figure 1 Amplifiers are active while baluns facilitate passive RF converter front-ends. Source: Texas Instruments
Amplifiers and baluns: The advantages
On the active side, you can use an amplifier to preserve the DC levels, also known as the DC bin. This type of amplifier/ADC interface in the signal chain is often referred to as DC coupling. Therefore, these DC levels hold some important information in a particular application.
Amplifiers can also maintain better isolation in previous stages in the RF signal-chain lineup, which matters when using unbuffered converters, or if standing waves are evident when the “match” is just not quite right in the frequency pocket relevant to the application.
Although amplifiers inherently provide +6 dB of gain, that gain is independent of its output impedance. In other words, the amplifier’s bandwidth won’t suffer or drop like a rock when it comes to subtle fluctuations in impedance across the frequency band of interest. Because the amplifier is more autonomous with gain vs. output impedance, amplifiers typically enjoy a more “ripple-free” passband.
On the passive side, baluns cannot pass DC levels, and transformers are inherently flux-coupled devices, which make them AC couplers. In other words, DC blockers. So, for inherently AC-coupled RF signal-chain lineups, baluns are a great choice. They are naturally passive and need no power supply nor supporting circuitry such as decoupling capacitors or ferrite beads for power. And because they need no power supply, baluns are noiseless.
It’s common for surface mount technology (SMT) baluns to provide 10 GHz and even 20 GHz of bandwidth. If you opt for a modular balun type—for use in a lab setting or high-end instrumentation—those baluns are hitting upward of 80 GHz with a bandwidth of 300 kHz to 80 GHz
If the bandwidth requirement for the application is paramount and you need more than 5 GHz, the balun seems to outshine the amplifier. But hold on.
Amplifiers and baluns: The disadvantages
A balun is more like a window than a door, and provides little to no isolation. Therefore, RF signal-chain lineups sometimes need to include isolation from the balun’s passing standing waves, impedance mismatches and our friend “kickback.” Kickback is a common term used to describe the existence of charge injection from the opening and closing of the converter’s internal sampling switch capacitor if the ADC is unbuffered.
Unlike amplifiers, passive baluns can be lossy; a high-frequency balun needs a wideband matching pad (3 to 6 dB) to help “stiffen” the broadband impedance across multi-gigahertz bands. The balun is more gain-dependent of its output impedance, unlike an amplifier. Therefore, the pad adds more loss into the RF signal-chain lineup and can increase the overall noise figure of an analog receiver design. Finally, because of the broadband match approach, standing waves add and subtract from the passband flatness, causing ripple throughout the passband.
On the other hand, as an amplifier is inherently active, it also inherently outputs noise and spurious as well. The noise will vary depending on the amplifier’s design, but all amplifiers will have some amount of noise and spurious output, which ultimately the ADC will see and gain up. For example, if a particular amplifier has a gain of 12 dB, an output-referred noise of 5nV/√Hz and a 12-bit, 10-Gsps ADC with an 8-GHz input bandwidth, 1-Vpp differential full scale has a signal-to-noise ratio (SNR) of roughly 60 dB. These two devices will effectively add together.
In other words, worsen the ADC’s noise floor and lower the dynamic range, worsening SNR by 5.45 dB or, as expressed by Equation 1.
SNRtotal = 20 × log ( (1/2)/√2) /√NoiseADC2 + NoiseAMP2 ) = 54.5 dB (1)
Where, NoiseADC = ( (1/2)/√2)/10 (60/20) ) = 354 µVrms
NoiseAMP =5 nV × √1.57×8000M = 560.4 µVrms
Amplifiers not only have noise, but are prone to linearity as well. This linearity effectively adds to the ADC’s linearity, making it worse overall. For example, if an amplifier’s worst spurious output is –80 dB and the ADC’s worst spurious is also –80 dB, the best effective linearity in an amplifier-plus-ADC interface design at this particular frequency is –77 dB, as expressed in Equation 2.
Spurious = 20 × log ( √(10 (-80)/20) )2+ (10 (-80)/20) )2 ) (2)
To combat against any noise or spurious output in the band of interest, an anti-aliasing filter (AAF) between the two devices will help. How much it will help depends on how narrow or broad the filter design is and on the AAF’s roll-off criteria. Extra supportive components will add more need for area in between the amplifier and ADC interface.
Understanding the importance of phase imbalance
If a frequency plan includes the even orders—the second (HD2), fourth (HD4), sixth harmonics (HD6), and so on—then you must also look at phase imbalance when designing the analog front-end interface. Both amplifiers and baluns have a finite amount of phase imbalance between their output signals, which typically gets worse (deviates) across higher and higher frequencies.
Phase imbalance is the term used to quantify the amount of phase imbalance between two signals. Since the ADC’s analog inputs are typically a differential interface, the two inputs ideally should be equal in amplitude and 180 degrees out of phase. For example, if Ain+ = –2 degrees and Ain– = 185 degrees, that produces a 7-degree shift, which translates into the frequency domain or fast Fourier transform (FFT) plot as a worse even-order distortion; that is, the second harmonic gets worse.
Unfortunately, there is really no real way to quantify how much phase imbalance your signal chain can take before it starts to degrade system performance. That’s because every component with a differential input or output interface—be it active or passive—will have some finite amount of phase mismatch at some frequency. There is really no way to perfectly balance an IC internally, a balun’s windings, or even cables to the absolute perfect length.
So, when performing balanced or differential test measurements in the lab, where you plan to use cables or adapters in the test setup, these “extras” need to phase-matched as well.
If there is still doubt, and you like a bit of math, TI’s E2E high-speed converter forum can provide help for having a full derivation model of the ADC. Here, the model uses a third-order transfer function and a pair of sinusoid signals to prove how phase imbalance gives rise to even-order distortion as shown in Figure 2.
Figure 2 The differential input signaling mathematical model shows how phase imbalance gives rise to even-order distortion. Source: Texas Instruments
Phase balance with amplifiers and baluns
Returning back to the topic of amplifier and balun trade-offs, baluns come in many forms, packages and designs. Classic ferris-type baluns are usually prone to phase imbalance, so it’s best to consult the datasheet before making your final selection based only on insertion loss or the bandwidth that the balun can cover. Smaller packages that use lithography structures have tighter tolerances and better repeatability, which typically means that it’s possible to improve the phase imbalance. However, it comes at the cost of smaller or more narrow bandwidth choices, which isn’t always ideal if the design calls for lower frequencies in the UHF bands near DC.
Module baluns produce some of the best phase imbalance results, but are big, bulky and costly—as much as $2,500 for just one piece. These really expensive baluns provide some of the widest bandwidth-hitting DC frequencies and maintain phase flatness into the multi-gigahertz regions. Figure 3 compares the phase imbalance of some types of baluns available on the market.
Figure 3 A comparison of balun phase imbalance may help designers pick the right component. Source: Texas Instruments
If your design calls for a wide bandwidth but there are also cost constraints, one neat trick is to put two baluns or transformers back-to-back to improve the phase imbalance. See Figure 4 and Figure 5. The only downside is the doubled amount of PCB area in order to implement this type of front-end structure.
Figure 4 Various double-balun configurations can improve phase imbalance. Source: Texas Instruments
Figure 5 Here is a view of single-balun vs. double-balun phase imbalance improvement. Source: Texas Instruments
Using either balun configuration A or balun configuration B from Figure 4, or the red and blue curves in Figure 5, you can see that a phase imbalance of 5 degrees or less can be extended out to 3 GHz over the original single-balun configuration, or the green curve. If using one of these balun configurations, note that each balun combo will have varying degrees of improvement.
Phase balance is also prevalent on the amplifier side. Because low-noise amplifiers and gain blocks have single-ended inputs and outputs, you might assume that these types of amplifiers won’t have good phase imbalance and have high even-order distortion, which is why HD2 isn’t specified in amplifier datasheets.
Fully differential amplifiers (FDAs) are a classic amplifier input interface to the ADC with differential inputs and outputs. Even though the FDA may allow you to tie the INPUT pin to ground in some fashion, the FDA enables conversion of the single-ended signal to a differential signal. The FDA inputs are sensitive to this reference shift, and therefore will exhibit more even-order distortion than what’s published in the datasheet.
FDAs are typically characterized with wideband modular baluns in order to capture their published performance metrics. There is a new sheriff in town, however. The TRF1208 amplifier uses a compensated input structure, allowing for a single-ended input interface by default and removing the dependency of the balun on the inputs like a traditional FDA. TRF1208 input structure fits well when interfacing to RF analog receiver cards, which are classically single-ended.
Figure 6 compares even-order distortion, HD2, with analog input frequencies up to 10 GHz. Using the same ADC, the figure directly compares a typical wideband balun interface, the TRF1208 interface, a low-noise-amplifier-plus-wideband-balun interface, and an FDA with and without a balun on the differential inputs.
Figure 6 A comparison is made among even order distortion vs. FDA, low-noise amplifier plus wideband balun, balun only and the TRF1208 interfaced with the ADC12DJ5200RF 12-bit, 5.2-Gsps ADC. Source: Texas Instruments
You can see that, in configurations below 1 GHz, the performance is fairly equal. As the analog input frequency climbs past 2 GHz, however, there is a distinct growth in the even-order distortion for all combinations except the TRF1208 and low-noise-amplifier-plus-wideband-balun interface.
Bringing the trade-offs back in phase
To summarize, both active and passive front ends have their pluses and minuses. Before making a choice in haste, please understand the application at hand, which will help determine which path to follow. Once you have decided between active or passive interfaces, look at the individual trade-offs related to phase imbalance sensitives, gain, power and input power required to achieve full scale. Collecting this information in a spreadsheet will help you quickly determine which device is best for your next +5-GHz design.
Rob Reeder, application manager for high-speed converters at Texas Instruments, is author of Signal Chain Basics blog # 172 for Planet Analog.
- What’s in an RF Front End?
- The future of agile RF front ends is now
- The latest about RF front-ends in 5G, Wi-Fi 6E and IoT
- 5G RF front-ends: Design challenges and their solutions
- Selecting the right software defined radio solution for your application