RF overload indicator circuit

[bson]

I'm looking for ideas for a simple overload indicator.  Here's my naive stab at it.  The RF signal comes from a TI THS3202D CFB wideband amplifier (MSOP-8 with thermal pad, 28dB gain from two stages insides an SMD PCB shield).  Its behavior as the output goes beyond ~18dBm (@50R) is it shoots straight to ~3Vrms and oscillates at ~500MHz.  Now, I don't really have a problem with this per se, it doesn't really cause any harm, I'd just like an indicator showing this is occurring.  The RF signal is DC up to 300-500MHz depending on what gain and compensation I populate the board for.



The series D1, D2, D3 will have a V_f of ~2.1V, or ~16-17dBm.  So they'll clamp just below 18dBm, causing the THS3202D output to sink into C1, charging it and triggering U1, lighting the indicator LED.  When the condition clears C1 bleeds out through R4.  The hysteresis of the HC14 guarantees the LED has a minimum pulse duration and won't blink at a rate too short to see.

Some potential issues:
* If the amp output sits right at the threshold it can, because of the hysteresis of U1 "drive" the detector circuit without any actual indication, clipping and distorting.  This is a very narrow boundary though since the output of the amp will very quickly charge up C1.  But it's theoretically possible.
* The 4148's have about a 4pF junction capacitance.  This will add a load on the output, though I suspect possibly insignificant.  I'm not entirely sure how the junctions will combine; three 4pF caps would result in a 1.33pF net capacitance, but these aren't capacitors...

If this works, then it would be a nice little compact addition that takes very little board space.  The LED would be a 3mm PTH device actually stuck into a front panel hole and hooked up with a simple 100mil 2 pin header.

Any diode more suitable than 1N4148 - with lower capacitance?  Clearly the continuous current from 2V down through 33k is trivial.  The peak current to charge C1 should be of no major concern.  I'm kind of new to poking around with RF in general and I'm slightly (but not overly) concerned with stability with the series 1N4148 loading the output.  Unfortunately it's kinda hard to test without actually making the board...  If it simply won't work I'd rather find out before I have a handful of boards!

The +5V is already on hand since I operate the THS3202D at +/-5V.
R2 will be determined empirically depending on what LED I use; 1k is just a placeholder.
U1 is a single-gate 74HC14 in SOT-353 (SC-70-3).

Any better, simple approaches I'm missing?

[bson]

Here's the current incarnation of the amplifier board.  It's about 22x32mm.  The idea is to mount two of these against a front panel with bulkhead SMA connectors, side by side in a shielded enclosure with a power supply.  They'll be fixed +14dB (or +10dB, +20dB or whatever else they're populated for) and can be used either individually or cascaded.  Each one would have its own individual overload indicator.  Since they'll be mounted directly using bulkhead connectors they won't need the corner holes... I'm inclined to remove those.  (I just habitually add corner holes to pretty much everything.)

 

(Apologies for the crappy phone photos...)

[bson]

One idea I had to handle the capacitive load is to split the 50 ohm source resistor into 10+40 and put the overload detector between them.  But I think gain would have to be rather low (and hence bandwidth high) for 4pF to significantly eat into the phase margin.

Another potential problem is that the detector only operates on the positive half of the output, so when it trips it may cause a sharp negative spike, but it should I think ( <= famous last words!) it's bound to be short enough to not really be much of a DC offset.  The amp can source 115mA so will when it trips the diodes it's going to load up C1 real quick.

I will probably have to run the HC14 off 3.3V.  Doh.
I need to rethink the triggering altogether, but that's a small problem; I can always replace the third diode with a transistor.

Actually, maybe just splitting the output and feeding one half of it into a half-wave rectified integrator with a bleed resistor, and using the integrator to trigger is a better approach.

[bson]

New plan.  I guess I'm talking to myself here.   

[Whales]

For detecting negative spikes: have you considered replacing your two diodes with a bridge rectifier arrangement?

Would the capacitance of the diodes matter?  They won't pass your carrier but they will pass your original signal.

[uncle_bob]

Hi

The standard way to do this for at least the last 50 years has been a capacitive coupled voltage doubler. The cap drives the junction of a series connected pair of diodes. One diode goes to ground, the other dries a cap with your output on it. There are SOT-23 high speed diodes designed for this sort of thing.

http://www.onsemi.com/pub_link/Collateral/BAT54SWT1-D.PDF

Is a really common / cheap / generic one. There are a lot of others out there.

Bob

[bson]

Nice, thanks!  I think this is more or less what I'm doing, except I'm detecting much higher signal levels.  Instead of the two series Schottkies I have three bipolar junctions, the third one being the B-E junction in the transistor, and tapping the series between the two last junctions.  I also don't AC couple it because I don't want to give the amplifier a significant capacitive load, or terminate it and add gratuitous insertion loss.  I do plan to add a clean-up Bessel low pass after the source termination though... but first I need to take a look at the output spectrum and I don't have a suitable network (or spectrum+TG) analyzer yet.  Just to know what actually needs cleaning up.

I'm making this not so much for radio/CW purposes as for general wideband amplification use.  One such use is to experiment with TVS diodes to create a wideband white noise source.

[uncle_bob]

Nice, thanks!  I think this is more or less what I'm doing, except I'm detecting much higher signal levels.  Instead of the two series Schottkies I have three bipolar junctions, the third one being the B-E junction in the transistor, and tapping the series between the two last junctions.  I also don't AC couple it because I don't want to give the amplifier a significant capacitive load, or terminate it and add gratuitous insertion loss.  I do plan to add a clean-up Bessel low pass after the source termination though... but first I need to take a look at the output spectrum and I don't have a suitable network (or spectrum+TG) analyzer yet.  Just to know what actually needs cleaning up.

I'm making this not so much for radio/CW purposes as for general wideband amplification use.  One such use is to experiment with TVS diodes to create a wideband white noise source.

Hi

The AC coupling adds no capacitance to the output of the amp. The diodes are in series with the blocking caps. The AC coupling *does* take out all of the DC offset issues.

With the 30V diodes I linked, you can go up to quite a bit of RF before anything bad happens. The more diode junctions you have in the system, the more DC offsets build up in your output. It's much more reliable to have a 3V output from the diode setup and a threshold after than to have a 1V output with two more diodes in the system.

I've done this a *lot* of times.....

Bob

[bson]

I just breadboarded it quick and dirty to check the power detection levels (just using a simple 100kHz signal), and it trips at 15.7dBm, which is slightly low.  But it's a very nice solid indicator, no flicker or hesitation.  It comes off, then when the power is dialed down it goes off cleanly with a hysteresis of about 0.2dBm.  I approve.  I see clipping above the detection level of course.  If I add yet another 1N4148 it comes up to 17.1dBm to trip and 16.9dBm to clear (so ~0.2dBm hysteresis).

I'll think about the blocking cap; that's not such a bad idea!

Another thought here is to cut the output... are there any relays good to these bandwidths, say 500MHz?  Even something like an Omron G6K is only spec'ed to 100MHz and basically goes off a cliff at that point.  A small FET?

[bson]

Yeah, this is low level stuff < +20dBm.

Panasonic seems to have a reasonable microwave relay collection... ~$7 in single unit quantities from Mouser, so quite affordable.
http://legacy.pewa.panasonic.com/assets/pcsd/catalog/ars-catalog.pdf

[bson]

Thanks for all the help and ideas - it works very well up to my ability to test it - which is only 200MHz currently.  I should really get myself some proper RF gear (SA/VNA - been eyeing the 8753D, a power meter, and an RF generator).  I did screw up the PCB by tapping the output after the source load instead of prior to it  , and have an updated board in the works (also tightening up spacing on the diodes and adding a bit of top side ground pour), but I just bodged it with an x-acto knife and a very short piece of 30awg single strand conductor.  The DC offset is there when it trips but I decided to just use an external DC Block instead for better bandwidth control.  Works like the proverbial charm! 

[tggzzz]

Consider using a 65LVDS2-series LVDS receiver, as recommended by Winfield Hill and John Larkin. Advantage: you can vary the threshold voltage, thereby varying the sensitivity. FFI: https://groups.google.com/forum/#!topic/sci.electronics.design/ZgNl_0PNnjg

Alternatively, the 400MHz (i.e. ~500MHz) Tek 485 scopes detect too much power and flip a mechanical relay before the 50ohm input gets too hot. You could easily reuse the circuit shown in the manual.

[bson]

Consider using a 65LVDS2-series LVDS receiver, as recommended by Winfield Hill and John Larkin. Advantage: you can vary the threshold voltage, thereby varying the sensitivity. FFI: https://groups.google.com/forum/#!topic/sci.electronics.design/ZgNl_0PNnjg

It looks like the maximum differential voltage on the 65LVDS2 is 1V, which is a bit low.  Also, it's not clear that it tolerates DC input on either input.  Its output drive is 12mA, which I suppose is workable.

I'd still probably need the emitter follower and '14 hysteresis, so it would really only substitute for a diode.  But on the plus is it has only 3pF input capacitance so its load is even less than a diode.  And, of course, no DC offset!

[ealex]

Hello

This is the overload module used by a R&S SMS2 signal generator:



I'ts able to detect DC offset applied to the input and RF overvoltage.

[tggzzz]

Consider using a 65LVDS2-series LVDS receiver, as recommended by Winfield Hill and John Larkin. Advantage: you can vary the threshold voltage, thereby varying the sensitivity. FFI: https://groups.google.com/forum/#!topic/sci.electronics.design/ZgNl_0PNnjg

It looks like the maximum differential voltage on the 65LVDS2 is 1V, which is a bit low. 

You could use a divider to make it less sensitive. Sensitivity is usually an advantage!

Also, it's not clear that it tolerates DC input on either input.

LVDS has a DC component, so yes, it can tolerate DC - and requires it for proper operation! You would bias the inputs to halfway between the min/max levels, and AC couple the RF input.

The data sheet gives the max/min input voltages.