Can you assist in selecting and understanding the AD9789 (and 9739, that should be similar) output stages selection?
The need is to choose the best output stage for using AD9789 in normal mode.
There are a set of recommended circuits both in the 9789 and 9739 data sheets, but most of them fail to match well both the DAC output and the circuit output.
Most of them make use of a 2:1 transformer. If the idea is to feed a 50 ohm transmitter, there is a strong mismatching both seen from the DAC and seen from the output. When using modulated signals with high PAPR, the level is reduced and the mismatching can cause additional return losses that take the good quality of the original signal down towards noise.
To clarify this point I would divide the problem in two parts, I assummed that for best response in normal mode the best was to use the recommended transformer (JTX-2-10T), with the LPF circuit, as in AD9789 data sheet fig. 112 that meets the evaluation board circuit also. I was considering using this circuit to start with.
1) analyzing this circuit, I can't undestand it exactly, as the DAC output differential impedance is 70 ohms. Then after the 90 ohm shunts, i.e. 180 ohms differential, the impedance seen by the output stages will be 50.4 ohm. What s the LC filter impedance? I assumed it was 50 ohms as this is the Rg it sees, but at the load it does not see 50 but 100 ohm. Then, no 2.1 transformation would be needed, but rather a balanced to unbalanced 1:1 ratio transformer. This transformation produces mismatch. On the other hand, the total net has losses due to the 90 ohm resistors, and the DAC load is also different than 70 ohms (50x2 in parallel with 180). It is true that the real transformer does not behave exactly as that, but the main mismatching produced by the high output impedance presented at the transformer DAC side port is still evident.
If one discounts the layout effects, that of course would have to be added, the impedance seen by the DAC in this fig. 112 circuit varies from a real value near 50 ohms at low frequency to something closer to 50-j50 for higher ones near 1Ghz.
2) Reviewing the data sheets, I have foud that the AD9739 one still recommends a different circuit for the normal mode, that uses BOTH a balun (1.1) and a 2:1 transformer. The problem remains basically the same as far as this transformer does not match the output impedance to the DAC impedance.Even if it does, due to the 90 ohm resistors if the output was really matched, the DAC will not.
3) The JTX-2-10T recommended circuit for normal mode AD9789's 2:1 transformer is not the same one recommended for normal mode AD9739 (ADT2-1T). And in the latter, a 1.1 wideband balun is included that does not appear in the 9789 case.
4) What is the impedance of the LC filter? It seems to be 50 ohm, as this is the impedance it sees at its input, but why the data sheet advices of using 50 ohm traces (and not 25) "from the DAC to the transformer" as if the filter was 100 ohm's and not 50?
Summarizing, ADI prevents the user to chose a different, well matched network, by recomending some transformers, but for the normal mode these transformers are 2:1 and produce mismatch at the DAC and / or the filter output. Can you just comment what should be the best optional network in spite of this, for a normal mode AD9789 50 to 900 MHz output?
It is impossible to match the circuit at all points if the resistors are included, but a better return losses are possible using a 1:1 instead 2.1 transformer, and the DAC mismatch due to the resistors are also smaller anyway. Would the DAC performance be worst with a better matched circuit?
Thanks very much in advance