ADL5350: Improve conversion loss

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

I want to use ADL5350. According to my test, the conversion loss is more than
10dB, much larger than the spec in datasheet. Would you please help review it
and give some suggestions on how to improve the performance? Thanks to you help
in advance. 


I advise to follow the IMPLEMENTATION PROCEDURE on ADL5350 data sheet (page
17-18), and tune the RF/IF/LO networks accordingly. In general, the RF and LO
input ports should be designed to pass the RF or LO signals, while the IF
output port should be designed to block RF and LO frequencies. It looks like
the customer misunderstood how this mixer works, and is simply using coupling
caps of 4.7nF on all three ports, which is not going to meet the design
requirements. With that being said, for each different frequency plans,
customers would need to set up the tuning networks for frequency specific. The
same principle applies to both receive (RX) and transmit (TX) applications,
which also means the networks tuned for receiving will NOT work for
transmitting within the same frequency plan. For instance, 125MHz IN with 50MHz
in RX should have different external networks from ones serving 50MHz IN with
125MHz OUT in TX. You may want to find out what application is this, and if
this is for TX, RX or both. If it is for RX, then you can easily pick Figure 59
as a start point, and tune the networks (lower L or/and C values a bit in 36nH
and 27pF combo) on both sides, so that the tuning is optimized for 125MHz input
frequency. If it is for TX, then you need to start over by following our
procedure for 50MHz input frequency. If you’re interested to test this
frequency band, they should use the schematic provided in Figure 65 (Board A)
on ADL5350 data sheet. Last, the mixer is not sensitive to RF/IF INPUT power
levels, so it is not meaningful to sweep the input power levels when debugging.
Sweeping LO power levels for seeking improved conversion loss might be a good
idea (see Figure 23) only after the external networks are properly tuned for
desired frequency bands, otherwise I would suggest to leave LO at 4 dBm as