Extending useable frequency span of 1:1 wideband transformers used for distortion measurements
Note: for full article please click here for pdf.
With the continued migration of signal path components to differential stages, it is sometime necessary to convert that differential signal to single ended for distortion measurement purposes. Also, when a four-port network analyzer is not available, using 1:1 transformers is a typical way to map a differential I/O to single ended for more typically available network/spectrum analyzers.
While the transformer response can be calibrated out for response shape measurements, it is more useful to have the differential to single ended path giving a flat gain response for distortion measurements. In the course of testing a 4GHz Fully Differential Amplifier (FDA) for OIP3, no flux-coupled transformer could be found that held 0.5dB flatness above 200Mhz. A simple technique of shifting the flat region up will be shown with resulting OIP3 data through 300Mhz for the ISL55210 wideband FDA.
Typical 1:1 transformers used in device characterizations and their bandwidth limits.
While various 1:1 wideband transformers are available, holding better than 0.5dB flatness above 200Mhz seems particularly challenging for those devices that use the winding structure of figure 1. For the best OIP3 measurements, the test signals passing through the transformer should also be well separated from its rolloff regions to maintain adequate linearity in the transformer itself.
Click on image to enlarge.
Fig. 1 is the classic “balun” configuration where a balanced input signal on the left (in this case from the output of a Fully Differential Amplifier- FDA) is going to an unbalanced output – typically the single ended input of 50Ω measurement instrument. All transformers are specified from a particular source and load configuration (those would be equal in the 1:1 turns ratio case) with some midband insertion loss. Typically they report both the high and low -3dB frequencies for these bandpass devices using a selected source impedance where the load is assumed to be matched to that. Table 1 summarizes a small selection from the 1:1 balun universe with the relevant specifications pulled from the vendor data sheets.
Click on image to enlarge.
Baluns
The two bold rows are baluns that will be considered here. A clue for a path to higher passband frequencies can be found in comparing the characterization impedance. Increasing the source and load impedance will shift the passband up in frequency while moving it down will shift the passband down. While the ADT1-1WT seems remarkable at 800Mhz upper F-3dB, that is specified using 75Ω. Operating in a 50Ω environment, its passband frequencies will shift down. A simple Spice modeling technique allows a very close fit to actual response (ref. 1). Comparing the simulated response for the 2- focus devices gives the curves of figure 2, where the ADT1-1WT is modeled at its 75Ω specification but simulated here at 50Ω source and load.
Click on image to enlarge.
This is from the input of the transformer to the 50Ω load so the 6dB matching loss from the source is neglected. The red Macom device curve exactly matches its 50Ω specified F-3dB range of Table 1 (400kHz –> 500Mhz) while the green ADT1-1WT has a bit more range on the low end, but is now very similar on the high end moving to a 50Ω simulated (vs 75Ω specified) source and load impedance. This simple model does not include insertion loss so a 0dB midband gain is shown. Both transformers have fallen below 0.5dB flatness at < 200Mhz whereas it would be desirable to use this balun output approach for FDA OIP3 measurements through 300MHz.
Shifting the response flatness region to a higher frequency
Taking this source and load impedance up should extend the flat response region to a higher frequency. This cannot be pushed too far as the interwinding capacitance will come in at lower frequencies for higher source Z. However, setting out to operate the transformer with a 100Ω source and 100Ω load, while still presenting 50Ω final output source impedance for a spectrum analyzer input, will give the example simulation circuit of figure 3 using exact resistor values (ref. 2).
Click on image to enlarge.
The Rm element is not strictly required to achieve a stepped up impedance looking out to the transformer secondary and is set to 10MΩ here to take it out. The Rm element is required to move the balun passband down in frequency using lower impedances but still getting to 50Ω output impedance. The circuit of fig. 3 puts 100Ω source and load on each side of the transformer while still presenting a 50Ω source to the 50Ω cable that is usually used to move the signal to the spectrum analyzer. That element is Rt in fig. 3. To emulate the 50Ω specification condition for the 1:1 transformer we would set both Ro elements to 25Ω, the R1 = 0, and R2 = ∞. The V1 source emulates the FDA output and is set to a 6.84 value to normalize out the 16.7dB insertion loss to Rt for simulation. R4 simply provides a DC operating point and would not be used when the real source is present.
Comparing the 2- predicted response shapes for a 50Ω vs. 100Ω test, using a source value of 2 in the 50? case, and the 6.84 shown above for the 100Ω case, gives the response comparisons of fig. 4. As expected, the predicted F-3dB passband has shifted up 2X where now the upper 0.5dB cutoff is a more useable 380Mhz.
Click on image to enlarge.
Figure 4. ADT1-1WT response comparison for 50Ω vs. 100Ω source/load conditions.
Using off the shelf wideband baluns is a common way to go back and forth between single ended and differential signal paths. Using the approach presented here has yielded very solid intercept measurements through 300MHz for a very wideband, low power, FDA like the ISL55210.
For full article click to pdf here.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
