John Miles wrote:
In article ,
says...

I just got some samples of the AD9952 and AD9954 DDS chips.
Both look identical, AD9954 has some extra features I don't need
but can simply not program. These parts are real small, will have
to find some VERY small tweezers to bend the leads and file down
a soldering tip to fit.


I'm a big fan of these tweezers:
http://www.folica.com/removal/rubis_costwe132.htm They have the finest
tips I've ever seen, and are great for working with these sorts of
parts.


I would try to solder all the gnd and unused
inputs down to a PC ground plane and bend all the power pins up
flat against the top of the chip to be soldered together. The signal
pins would have to be bent up halfway and fly leads soldered on
to go to the rest of the circuit on perfboard.

Anybody try to breadboard one of these 'dead bug' style like this?


Note that TQFP pin spacing (0.5 mm) is half that of the AD9850/AD9851
chips that Dan Tayloe mentions. Still, it can be done, and your
strategy sounds reasonable. I have hand-wired several 16-pin TSSOP 0.5-
mm chips by hand without losing one. Solder-wick and a fine-tipped iron
are mandatory. As Dan says, you can count on being able to bend each
lead only once, so make sure you get it right!

On the plus side, these chips appear to be very hard to overheat in the
soldering process, probably due to low lead mass and stout internal
connections. You can afford to take your time.

It might not be an issue with 1.8V logic, but these chips' 3.3-volt
predecessors (the AD9852 and AD9854) really got hot when operated near
their maximum clock speeds. On a PC board, the traces offer a definite
heat-dissipation advantage over hand-wiring. Don't assume you can run
these chips full tilt until you verify that they don't overheat in your
dead-bug configuration.

Since I was thinking of using a fair sized piece of PC board as a ground
plane
to which ALL the gnd pins would get soldered to, I might have a good enough
heat sink. I could also put some silicon heat sink grease on the bottom
of the
chip, or glue the chip down to the pc board with heat conductive epoxy.






Was thinking of using 8051 series micro, the built in serial port will
drive the AD98xx and AD99xx dds chips. Question....how do I convert
the 5v digital io of the 8052 micro to the 1.8v levels used by the
AD995x series, or is the DDS chip 5v compatible (on input maybe...
what to do about output? especially on the bidirection serial line....)


Just use a resistive divider. I don't see any reason to use FETs unless
programming speed is a concern. (CMOS inputs look like capacitors, so
resistive dividers will slow your signal edges down and limit the speed
at which you can program the chip. Some CMOS parts have a minimum-
risetime requirement, but I haven't bumped up against it yet personally
except for Vcc pins.)

I see in the data sheet that the I/O pins can be configured for 3.3V
logic levels, so that might help.


Was also thinking about maybe building the rig as a 6 meter single
conversion with an HF transverter front end. What do you think of
this idea?


Sounds like a plan!

-- jm

------------------------------------------------------
http://www.qsl.net/ke5fx
Note: My E-mail address has been altered to avoid spam
------------------------------------------------------