Can AD8318 CLPF pin be shorted to ground indefinately? We are making a passive intermodulation tester which uses an automatic level control loop. The loop incorporates a 50dBm power amplifier, an ADL5330 VGA, and an AD8318 Log Amp as described on Page 16 of the ADL5330 datasheet. We are experiencing PA burnouts which have been traced to a nasty transient when the PA is energized. This is pretty much the last thing which happens. Essentially, since the PA was initially off, the ALC loop is railed out, commanding the VGA to produce maximum power when the PA comes on.
The PA we're buying isn't the most robust thing in the world, and it seems to have a damage threshold near (or perhaps below) the the level that would be required to produce maximum output if the PA turned out to have the minimum specified gain. It other words, the minimum gain spec of the PA is 45dB, but they typically run 55dB or so. There is no maximum damage spec, but we have seen them damaged at +6dBm of input power. So, while we could add enough padding to the output of the ADL5330 to guarantee the PA wouldn't get damaged by the turn on transient, we might not have enough power to drive a PA that actually had a 45dB gain. Sadly, for a while, we are simply going to have to live with this amplifier as it is.
To solve this problem, the first thing I tried was to turn off the AD8318 Log Amp via its ENBL pin, which is supposed to put the amp into a low current state. However, to my dismay I've discovered the loop capacitor continues to charge--even when ENBL is held low. So, I still get the same transient when the ENBL pin is brought high again. Dang!
The next thing I tried was to short out CLPF on the AD8318. I actually tried this with a pair of tweezers, and it seems to do exactly what I want: shorting out the capacitor drives the VGA gain to its minimum value, effecting a level dip and allowing the PA to be turned on gracefully. This effectively closes the loop--except that the loop capacitor is now being held at ground. Once the PA is steady-state, the tweezers are removed, unshorting the capacitor, and allowing the power level to nicely climb to the desired value. This is much better for the PA than slamming it to the maximum value, and declining to the desired level.
BUT, it raised one last question: Can the current source feeding CLPF in the AD8318 tolerate being shorted to ground indefinately without damage? My guess is that, since we're talking about a controlled current source there is no restriction on the size of CLPF, it probably can.
I'm also curious why the ENBL pin does not perform as one might expect, and discharge CLPF to zero when ENBL is brought low. Surely other people are bothered by this type of turn-on transient, yes?
Thanks for your help.
Thanks for the quick resopnse.
I'm not sure I completely understand what you're telling me. First, the
log amp appears to be a curent source, so I would think the voltage
would be lowest when CLPF is shorted to ground. Can you be more
specific about what might break down? Second, it is important to note
that we can leave the log amp enabled all the time (as is done in the
datasheet). So, under these circumstances, shorting CLPF would seem to
be equivalent to connecting it to an infinately large
capacitor--something not specifically prohibited in the datasheet, no?
In any case, the potential to break down that transistor would seem to
exist for any loop capacitor that ever was completely discharged, yes?
Third, assuming we did use a 1K resistor, would we not still see a a
turn-on spike? (though it would 13 times smaller, we don't want to
overshoot at all.)
My original concern was that there was some potential to overload the
current source with a steady-state short. However it appears that there
may be more to it than that. What do you suggest we do to tame this
nasty little spike? As far as I can tell, nothing in the datasheet even
suggests it can happen--much less what to do about it. We can't be the
only customers who don't want over power transients in our leveling
In the short term, a breadboard version of this circuit has worked very,
very well in the lab. It will be hard to walk away from that. Would
you mind looking at the attached circuit one more time? Below is the
version we actually breadboarded up. If I have to tell my boss the fix
I found doesn't work after all, I'd like to be able to say exactly why
with some confidence, and if possible offer a better solution.