[time-freq] RE: time-freq Digest, Vol 11, Issue 4

[John Ackermann N8UR]

Christopher Hoover said the following on 09/08/2006 07:49 PM:
>> I am just doing a quick email scan at the moment after being 
>> away for a bit, so this isn't a researched answer, but you 
>> should be able to get more output than that.
> 
> So with _JP3 installed_ and a +10 dBm @ 50 Z input, I should be able to
> get at least +10 dBm @ 50 Z on the outputs?
> 
> If so, something's definitely not right.  I'll try to find some time to
> debug this over the weekend.

Chris, I went back and tested my TADD-1 this morning and your numbers
aren't dramatically off, though they are still a couple of dB lower than
I get.

I put in +7.6dBm at 10MHz and got out a maximum of +4.5dBm with 50 ohm
termination at both ends.  At those levels there was no sign of
clipping, so we can assume things are fairly linear and I would
therefore expect you to see more like +7dBm out for +10dBm in.

For what it's worth, removing JP3 caused the output to go up by 4.6dBm.

Using a scope to compare the output levels with high-Z and 50 ohm
termination at the load, there was the expected 3dB power/6dB voltage
difference.  To be precise, with JP3 installed and +7.6dBm in, the
output voltage was 1.22 volts peak-peak into 50 ohms, and 2.44 volts
with no termination.

One possibility to consider -- is your drive signal designed to drive 50
ohms?  If it's a higher impedance, you'll see more loss if you put the
50 ohm terminator in line.

And, hang on one -- I think I just discovered the culprit.

In the final design changes for the production board, I added R4 and R6
which are intended to provide a high-Z load for the input and also form
a voltage divider that could be tweaked if necessary to allow higher
drive levels without driving the MAX477 into clipping.  I figured that
the suggested values of 10k in series and 100k to ground would have only
a minimal impact on the input level (about 9%).

BUT... I didn't notice that the R8 in the input bias network provided
another path to ground through only 15k.  Now we have a voltage divider
of 10k/13k (13k is the result of 15k + 100k in parallel) so that's
almost a 1:1 divider and therefore loses almost half the input signal.

So... if you need maximum output, making R4 smaller and/or the R1/R8
combination larger (they both need to be the same value) can almost
double the output level.  As a suggestion, changing R4 to 1k and R1/R8
each to 47k would result in a signal drop of only 1/32.  (I think that
47k will still provide a stiff enough bias, but we need to test that to
be sure -- if the output waveform starts clipping when it didn't before,
that means that R1/R8 are too large.)

This makes sense because my recollection from the testing done on the
Rev B boards was that the board would provide basically unity gain
instead of the ~-3dB that we're seeing.

Thanks for bringing this up, Chris!  (And this is more evidence that an
inverting amp might be a better design, so thanks too for bringing that
up...)

If you get a chance to make these changes, please let me know how they
work out.  I'll also experiment here at home.  If this works, I'll
update the manual accordingly -- and we will ship different values for
R4, R1, and R8 in the next batch of kits.

John