3 Example Art Using the ADC3669

Now for an example using a low cost balun and the ADC3669, 16-bit, dual channel ADC, for a wideband frontend match design of 1.5GHz of analog sampling BW.

In this case, we plan to use the TCM2-33WX+ from Mini-Circuits (MC). This balun has 3GHz of BW and a low insertion loss as compared to higher cost baluns that are easier to match with. This MC balun also has a very good phase imbalance, <5degrees, when compared to the other lower cost brethren across the same frequency range. For more details, see link to the MC balun data sheet.

Using the generalized circuit above, the components needed are not purely resistive to define the match. In this case, we can use an R-C-L (R2-R3-R6) approach which proved beneficial in our case, see Figure 3-1.

Figure 3-1 Finalized Passive Network Match

Here is where the fine art comes in. How do you approach this balun matching design conundrum? Do I model or simulate this? Some modeling can be done to help give guidance. But to be frank, PCB parasitics still come into play and unless you are a world-class warlord simulating expert, that has unlocked the key to backing out PCB parasitics subtleties, the typical approach is test a few different iterations on your board.

A good starting point is the following, grab both sets of S-parameters, if available, for the balun and ADC and use your favorite simulation SW. Be cautious on vendor s-parameters as noted above. Next, use the matching network format as given in Figure 2-1. Then, use the R2-R3-R4 matching approach for one of the following:

1. Use an attenuation pad approach, something like 8.6-140-8.6 ohms for R2-R3-R4 respectively, can give you a 3dB pad. For more on this approach, see the link.
2. Use an R-C-L approach for R2-R3-R4 respectively. Keep in mind this approach helps to resonate away the ADC’s internal parasitic capacitance or “C” using an inductor or “L” as the last component. This can flatten out the BW, allowing the balun to do the rated BW job. However, this approach does take a bit if iteration to get this correct.

As stated above, the goal here was to not use a lossy attenuation pad. Therefore, to give more context to the R-C-L approach, see Figure 3-2, Figure 3-3, and Figure 3-4 as varying the L, C and R respectively in the network (Figure 3-1) and the role in defining the ultimate BW and network match.

Figure 3-2 shows how changing the value of L around has an influence on the BW while keeping all other component values the same. Notice as L is increased in value the BW is slowly reduced. This means the L value is having an adverse reactive effect on the internal C parasitic of the ADC.

Figure 3-2 Passband Flatness Response with Various Inductance (“L”) Values at R4

In this next experiment, Figure 3-3 shows how moving the value of C around has it influence on the BW while keeping all other component values the same. Notice as C is reduced in value the BW is slowly improving at the cost of flatness of the BW. This means the C value is having a reactive effect on the balun’s return loss over frequency. These capacitors help preserve the balun’s BW vs. frequency.

Figure 3-3 Passband Flatness Response with various Capacitance (“C”) Values at R3

In this final experiment, Figure 3-4 shows how moving the value of R around has the influence on the BW while keeping all other component values the same. Notice as R is increased in value the BW is slowly improving at the cost of flatness or peaking in the BW response. The effect of R’s value is almost the same as the effect of L, therefore preserving the impedance requirements that both the balun and ADC want to have in conjunction with each other.

Figure 3-4 Passband Flatness Response with various Resistance (“R”) Values at R2

Ultimately, the R-C-L approach can be simulated as well, to give you a good starting point using the tune feature in your favorite simulation package. This does allow you to see the same roles each component plays in the network match. Settling on some good starting values can help define which direction to go when iterating and perfecting the match as needed for your application.

Next, during the matching design effort as iterations are made, the recommendation is from time to time to do an AC performance sweep across the application BW of the converter. This can give you insight as to how the performance is coming along dynamically and makes sure nothing has gone wrong with the ADC.

Figure 3-5 describes the ultimate AC performance (SNR and SFDR) measured across the bandwidth of the ADC3669 using this method to match the input network out to 1.5GHz.

Figure 3-5 Final Matched Network AC Performance (SNR/SFDR) vs. Frequency

Before closing, there is one more match method, a narrow band approach, for those that only need a portion of the BW that the ADC can deliver. If we expand on the listing above, we can call this number 3 or…

• 3. Use a narrow band approach, effectively use the same example network described in Figure 2-1. But instead, use R5 to resonate out the ADC’s internal sampling network parasitic capacitance or “C” by placing an inductor or “L” in shunt at R5. This is done by simply taking the frequency center point of the BW in the ADC’s S-parameters. Assuming the s-parameter S1P file is in the form of real/imaginary, then you can equate this C value to an L value. Here is an example:

Using the s-parameter file, find the center frequency in the file based on the filter’s design center frequency. In this case the 4pF was found at 110MHz center frequency for the narrow band filter design.

In the first step, find the reactive impedance at 4pF and 110MHz:

Next, Equate Xc to XL, as shown in the following equation:

Now solve for L:

As shown previously, by equating the two reactive impedance, L is found and can resonate out the value of the ADC’s internal C value, or in this case, 4pF. This sets the starting point for the value of L to be used. The value can be iterated from there to adjust the frequency a bit if need be to optimize the center point of your narrow bandwidth match.