AD8339:  channel to channel crosstalk

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

One of our customer reports cross-talk which occurs in adjacent channels in

The cross-talk is random, currently 40 parts have been tested, cross-talk are
observed in 3 parts. (Move the chip to another PCB, the cross-talk phenomenon

Testing Condition:
RF1=2.51MHz Sine Wave?RF2=NC?4*LO=10MHz LVDS Signal?LPF=100kHz Cutoff Frequency?
Modulator1?2 are turned on?

Testing Result:
Scope1?10kHz Sine Wave is observed?Amplitude=0.3Vpp?
Scope2?10kHz Sine Wave is also observed?Amplitude=0.003Vpp?

The customer said that actually the demodulated output is measured by a 5Vp-p
range, 16-bit ADC, so if the output is less than 1LSB, for them, it means there
is no crosstalk.

Their project is under assessment, they can't sure what arosstalk-level is
required now.


Thanks for contacting us, this is the first instance of crosstalk testing by
customers we’ve received on this part. Your input is very complete, and it
appears the customer measures about -40dB on the three devices of concern. It
would be interesting to know what level of crosstalk is provided by the other
37 parts and what level is required. Could the customer will provide these
numbers? Also, has this been an issue in the field or is the customer
conducting an evaluation test on the AD8339 prior to using it in a design?

Please note that we don’t provide a crosstalk specification on the data sheet;
I checked my data files and I find that there has been no crosstalk testing on
the AD8339.
The maximum signal attenuation or channel-to-channel isolation for ½ LSB in a
16B ADC is -102dB, and is likely to be difficult to achieve across the system.
Certainly multi-element probes don’t achieve this level of isolation,
element-to-element. In the experiment described below they measure a worst case
value of  -40dB with respect to a 300mV signal, so a full scale range of 5V
actually benefits low levels of crosstalk although 3mV would have to be
evaluated (-64dB with respect to 5V). However this approach doesn’t take into
account intrinsic  noise  and other error sources. Also, systems using CW
Doppler tend to incorporate analog switching between the transducers and the
amplifiers so there would be no need to rely on a high degree of adjacent
channel isolation. Rather, one could simply uses channels located elsewhere in
the array so as to take advantage of the a wider physical spacing.

Bottom line: the -40dB crosstalk will likely prove a non-problem.