DDS Output Termination

[garyr]

According to the Analog Devices document "A Technical Tutorial on Digital
Signal Synthesis" the second-best way of coupling DDS DAC output currents
to a filter (page 47) is illustrated below (View with fixed font, e.g.,
Notepad). The document also states that the outputs should be terminated
equally. Given that R1 & R2 satisfy the filter termination requirements,
what should be the value of R3? R3 = R1 or R3 = R1/2?

+--------+ +-------+
| | | LP |
| A------+--| filter|---+--
| | |R1| | |R2
|AD9834 | \ +-------+ \
| B+--+ / /
| | |R3\ \
| | \ | |
+--------+ / Gnd Gnd
\
|
Gnd

 

[Joerg]

R1 kinda sorta serves as a pad to flatten the likely reactive input to
the filter. A more comprehensive approach would be a pi-type resistive
pad.

A more comprehensive example is a diplexer. View it as:A driving two
paralleled loads: a LP filter to a flat load of R3, and a HP filter to
a flat load of R3, with the crossover region chosen to keep the
impedance seen to A quite flat.

A simple example of this diplexer approach can be seen, e.g. in W7EL's
"Optimized QRP Transceiver". On the web at
http://www.arrl.org/files/file/Technology/tis/info/pdf/93hb3037.pdf
If you look there, the HP is just a 0.1uF driving a 51 ohm load
resistor, and the LP is a 100uH choke feeding a common base amp set to
have a 50 ohm driving impedance. This is not perfectly flat but it
pays off hugely in terms of termination of the unwanted products
coming out of the DBM, making sure they are absorbed and not reflected
back.

All that said... a little balanced broadband transformer is preferable
and really quite painless unless you need to cover, say, more than 6
octaves or something. Few applications have to cover more than 6
octaves.

How about a li'l emitter follower? A BFS17 is a lot cheaper than a
broadband transformer

 

[garyr]

How about a li'l emitter follower? A BFS17 is a lot cheaper than a
broadband transformer

7 cents for the broadband transformer core (Mouser 623-2643002402)
quantity one. Add some trifiliar winding and viola, you're done.

BFS17 is circa 20 cents quantity one.

(Not that either cost matters much compared to the price of an AD DDS
chip.)

And as for emitter followers... look closely at W7EL's common base amp
at the LP leg of the diplexer. There is a lot of wisdom there.

And you still want a little transformer on the output of the DDS. AD
is not joking when they say it is preferred.

Tim.

 

[Joerg]

7 cents for the broadband transformer core (Mouser 623-2643002402)
quantity one. Add some trifiliar winding and viola, you're done.

BFS17 is circa 20 cents quantity one.

(Not that either cost matters much compared to the price of an AD DDS
chip.)

Yeah, but now you have degraded the monetary value of your own work to
next to nothing

I've done multifilar windings a lot but never like the stench that
develops when burning off the enamel. In the summer with a window open,
well, ok, but not in the winter.

And as for emitter followers... look closely at W7EL's common base amp
at the LP leg of the diplexer. There is a lot of wisdom there.

Was looking for it but only found a mention of it, no schematic:

http://www.qrp.pops.net/30M_RCVR.asp

I use common gate stages a lot if the source likes a hearty load.
Despite the fact that one of my profs has publicly said common gate is
stupid, one of the designs ran at around 35k units/year for a long time.

And you still want a little transformer on the output of the DDS. AD
is not joking when they say it is preferred.

There I have become careful. I've found math bugs in datasheets. Also AD
often proposed split grounds, engineers believed it and ended up with an
EMI nightmare. Ok, I am not complaining since this also brought me
income (to fix it).