The big question with DDS is the spectral purity of the output. Even
weak artifacts can cause all kinds of birdies and other troubles in
today's RF environment. This is one reason so many folks find the old
designs so appealing - they are "clean" except for the obvious things
like images.

The big problem with DDS is getting a good anti-aliasing filter on the
output to satisfy Mr. Nyquist. Trying to build a good 'brick wall' low
pass filter is insane. Running the DDS with a clock many times greater
than the highest output frequency helps, with an injection frequency of
39mhz on 10 meters with a 9mhz if a 400mhz DDS clock is 5x the Nyquist
requirement. Rather than using a low pass filter, I'd use switchable
bandpass filters, much more rejection of unwanted spurs and a cleaner
output. Not as simple to build, you have to switch filters to change
bands, but hey, I can always GANG several tv tuner chassis for my
band switch! (I knew those old TV sets I caniblized would come in handy!)


Thought I'd use
three IF stages with a filter between EACH stage and a final one before
the detector.


Very good idea. The best filter goes first, then "cleanup" filters.

Actually those filters are identical. But another idea is a double
conversion with both 9mhz and 455khz if's. Then by tuning the second
conversion oscillator you can 'overlap' the bandpass of both sets of
filters to use only a portion of them, and if you also move the bfo
frequency you can move the position of the signal within the resulting
bandpass.


The above radio would probably end up being a transceiver (cause the
extra circuity isn't much) but the finals would end up being 1 or 2 1625
bottles 'cause I have at least a dozen of 'em in the junk box.


Great bottles but the sockets are a pain. Unless you hack up an ARC-5
tx.
My junk box has a few 1625 size sockets, two ceramic 'plate' types and a
large saddle mount. Actually these sockets are easy to find and not too
expensive. Antique Electronic Supply has them, and they show up on Ebay
often. No need to use ones out of ARC5 xmtrs.

I also have a good number of 6AG7 tubes, they would make a good driver
stage for the 1625s. (BTW 6AG7 is sorta an octal version of the 6CL6).
If I used a single 1625 in the final, it would be good for about 25
watts output without straining. Then if I needed more power, I'd build
a linear amp using a 2 or 3 811A's (that's 340-410 Watts PEP output).
Not a full gallon, but the extra few db ain't worth the cost! I have
some 813's in the junkbox, but the sockets for them are costly, and they
have such HIGH output capacitance that making a good tank circuit in a
single ended amp is rather a pain. Still a grounded grid tetrode
circuit looks interresting! (cathode driven with normal g1 and g2 voltages).


The biggest headache I've encountered in transceiver design is finding
a heterodyne combination that works in both directions and uses
available components. All of the classic ones are compromises in one
way or another, either on rx or tx.
If you have enough filters you can put separate filters in the tx and rx
chains, and even use double conversion on the rx but single conversion
on tx. Then you can use a separate tunable bfo for the receiver. The
bandwidth and center frequency of the tx filter can be selected for good
audio, while the rx filter(s) for selectivity. Tx bandwidth is a
function of audio channel filtering, alc, and NOT overdriving anything.