Monday, February 16, 2015

Improved Pneumatic Antenna Launcher

About four years ago I built a pneumatic antenna launcher that launched small tennis balls into trees for raising wire antennas.  You can find that earlier post >>here<<.

Recently a friend sent me a link to KR4LO's "Air Boss" Antenna Launcher which uses small 2 oz. "egg sinkers" in lieu of tennis balls.  The Air Boss looks like a well thought out design with great range, and for the price, an excellent deal.

But I already had a launcher that I'd built.  It got me thinking, though...could I modify mine to use the same approach?

Yes, I could.  And I did.  Here's the new design...



The barrel is now a 3 foot length of 3/4 inch diameter schedule 80 grey plastic pipe (threaded at both ends), cut down to about 30 inches (explained below).

My first iteration used white 3/4" diameter PVC pipe for the barrel (1/2" was too narrow), but its inner diameter was slightly too large -- too much air was escaping around the sides of the 2 oz. egg sinker I was using as a weight, compromising its range.  I had to wrap about 7 turns of masking tape around the "waist" of the sinker to bulk it out enough to get a good seal in the pipe (the sinker was still loose inside the barrel, but not too loose).

Compared to the white PVC pipe, the grey pipe has a slightly narrower inner diameter (and it seems stiffer, too -- less likely to bend).  So the egg sinker fits into it better, but I still add about 2 turns of masking tape around the waist of the egg sinkers, but this additional bulking-out might not be needed.

The barrel attaches to a 3/4 inch PVC ball valve (threaded, female connections), so, for storage, it's very easy to unscrew the barrel from the air-chamber/valve assembly.  Note:  my original design used a 1/2 inch ball valve.  This meant that I had to adapt the threaded 3/4" threaded pipe to the 1/2" ball valve.  Which I did.  Unfortunately, during testing I snapped off one of the glue joints near that valve.  Rather than rebuild it with the same 1/2" valve, I thought it better to use a valve that could mate directly with the barrel, without adapters.

This new ball valve is a bit stiffer to turn than my original one.  I was worried that this meant I couldn't turn it as fast, and therefore the "explosion" of air when I opened it wouldn't be as powerful.

I tried lubricating the ball with "Faucet and Valve" grease, but this didn't seem to have any effect.  So I instead jury-rigged a handle-extender with a piece of scrap kindling and a hose clamp:


Not elegant, but it seems to work.

The reel is inexpensive, purchased at Walmart.  The reel came with a cheap no-name, no-spec line pre-installed.  I removed this line and installed in its place 30 lb. test "Spiderwire" braid (about 125 yards).

The fishing reel is attached to the barrel with a couple of hose clamps.

The fishing weights are 2 oz. egg sinkers (I picked mine up at Walmart).  I didn't want the fishing leader to pass on the outside of the sinker, so it runs only inside the sinker, with lead split-shot clamped on the leader at either end to keep the sinker from moving around:


One thing I've learned with antenna launchers, be they sling-shot, pneumatic, or whatever...if you aren't careful, you can get your weight (or tennis ball, etc.) stuck up in a tree.  For me, this usually happens when I'm not satisfied with where the shot went -- then I try to pull the line back (with weight still tied to the end of the line) so that I can try again.  The weight will start swinging on the end of its line (like a pendulum) as I'm pulling it up through the tree branches, and then suddenly, it's done a loop-de-loop and wrapped itself around a branch!  Arrrgh!

This time I wanted to improve the system:
  • The weight is attached to a short length of leader (about 3 feet), at the other end of which is attached a barrel-swivel.
  • The main line (attached to the reel) has a barrel-swivel with snap at the end to which the leader will attach.
  • Prior to making a shot, the leader's barrel-swivel is clipped to the line's snap.
  • Then, after the shot has been made but before reeling a line back, simply unsnap and remove the weight and leader.  No untying knots!!!
Here's an image of the swivels and snap:


Note:  For the sinker's leader I first used some light weight mono-filament line that came with the inexpensive reel I'd purchased.  My thinking:  because I was using 30 lb. test line for my main line, if my sinker did get stuck in a tree, I wanted something that, if I pulled on it hard, would break before the 30 lb. line broke.

Unfortunately, this unknown line I used as my leader was just a bit too light weight.  On my second test shot, it parted and the sinker went sailing over the tree and out of sight.

The red arrow points to where the leader's new end.


Lesson learned.  Now I'm using 3' leaders (purchased at Walmart) that are 20 lb. test and have a barrel-swivel at one end and a barrel-swivel-with-snap at the other end.  I remove the swivel-with-snap from its end of the leader -- it's on this end that I'll attach the egg sinker.  (I can then attach one of these swivel-with-snaps to the end of my main line).  Also -- the 3 foot leaders I used come with loops in the line at their 1 foot and 2 foot marks (for attaching other hooks, I suppose).  I clip the loops open to prevent them from snagging on branch stubs, etc.

Here's a photo of the leader.  If you look closely you can see the two loops before I clipped t hem:

(click on image to enlarge)

Note regarding the leaders:  after about two dozen shots into trees, my leader (see the photo above) separated at one of the pre-installed loops -- I'm guessing that the knot used to make a loop weakened the mono-filament and it eventually parted.  For this reason, if I build more 2 oz weight leader assemblies, I'll probably just use 3 feet of the 30 lb Spider braid (that I use on my reel) in lieu of the mono-filament.

For better visibility, I spray painted the weights fluorescent orange:

Hot chile peppers in the blistering sun...
(with apologies to Dylan)

If I'm unsatisfied with the shot and decide to reel back the line for another go, I first unsnap and remove the leader and its weight and then I attach a short length of Flagging-Tape to the snap (this could be a short length of rope or string, too -- just something to add some weight to the end of the line).  This acts as a bit of drag on the end and  helps keep the snap-end of the line from flopping around and possibly wrapping around something as I'm reeling it back in.  (That's my theory, at least.)


Other notes:
  • Always close the snap before reeling the line back!
  • The grey 3/4" barrel was cut down from its length of 3 feet to about 30 inches so that, when the sinker has dropped down inside the barrel all the way to the barrel's bottom (make sure the valve is closed, or you'll drop through to the air chamber!), the swivel on the leader is just outside the barrel, not in the barrel.
  • Useful knots:  Double-surgeon's loop knot (great for attaching swivels or making slip-knots), and the uni-uni knot (for splicing line together, which I discovered I needed to do when my more-than-200-feet of line on the reel (which I thought would be adequate) came up short in one of my shots over a very tall tree.  I'm now using the entire 125 yards of the purchased Spiderwire.) 
Final results...I can easily clear the tree in the picture below...


But how high can I get it?  The tree in the photo below is significantly higher than the one in the photo above. With 50 psi of pressure, maximum height that I can achieve seems to be on the order of 90 - 100 feet, per this photo (after shooting the weight over the branches, I reeled up the flagging-tape "tail" until it seemed about at its max height).  Note that the measurement below doesn't take into account perspective, so actual height could easily be over 100 feet:

(click on image to enlarge)

Much better than my tennis-ball antenna launcher, but can height be improved?  I'm hopeful, but nothing I've tried has yet made a significant difference. 


Standard Caveats:

Use at your own risk.  If you build one of these, don't overstress the PVC by pumping in too much air (I usually pump it up to about 40-50 psi).  Also, follow the instructions with the PVC glue, and, after handling the lead sinker and split shot, I'd recommend washing your hands.

Plus, when the weight descends, it has a lot of energy.  It has buried itself a good half-inch into the back lawn here -- I almost thought I'd lost it, then ran into the fishing line leading down into a hole in the ground.  So when you're aiming, pay attention to where it could fall!

All of which is to say: use common sense!

And, as always, I might have made a mistake in my equations, assumptions, drawings, or interpretations.  If you see anything you believe to be in error or if anything is confusing, please feel free to contact me or comment below.

And so I should add -- this design and any associated information is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Friday, February 13, 2015

A Milliwatt to Kilowatt RF Power Meter using the Analog Devices AD8307

A friend of mine, Dick Benson, W1QG, built a pretty cool combined Forward and Reflected Power and SWR meter using two AD8307 Logarithmic Amplifier chips from Analog Devices.  He used a PIC to convert the dBm readings back into watts (and to calibrate the device, too).

N2PK also has an interesting design for a "Forward Power and Return Loss Meter."  It, too, uses two AD8307 ICs.

One of these days I'll build a "Power and SWR" meter, too, but for the lab bench I thought a basic Power Meter would suffice.  Almost all of my work on transmitters is either testing or repairing them, so I don't really need to measure SWR or Reflected Power.  With that thought in mind, why not design a 0 - 60 dBm (1 milliwatt to 1000 watt) power meter?

Here's a picture of the completed power meter.  It reads either 0-30 dBm (1 - 1000 milliwatts) or 0-30 dBW (1 - 1000 watts).  Yes, the scale is in milliamperes.  Someday I'll make an appropriate scale.  But meanwhile, it's easy enough to convert the 0-3 mA scale to 0-30 dBm/dBW.

(click on image to enlarge)

And here's the schematic:

(click on image to enlarge)

And here's a better version, drawn after I'd posted the one above...

(click on image to enlarge)

Choice of parts (apart from the AD8307) was pretty much dictated by what I had in my junkbox.  The case was originally from a Drake W4 wattmeter. (I no longer recall where I found the case -- I must have picked it up (sans meter and electronics) at a swapmeet sometime in the past).

The meter (0-3 mA) was chosen because it was in the junkbox and, most importantly, it fit the hole in the case's front panel!  I initially wanted to use a meter that was marked from 0 to 600 mA (actually a 100uA FS meter), because the scale would be perfect for a 0-60 dBm meter without a range switch.  But I really wanted to avoid cutting a meter hole, so the 0-3mA meter won out.  The 0-60 dBm overall range is now broken into two sub-ranges:  0-30 dBm and 0-30 dBW (i.e. 30-60 dBm).

I also wanted to used op-amps with rail-to-rail outputs (so that I could drive down to 0 volts).  Fortunately, I happened to have some TLC2272 op amps on hand.


Schematic Notes: 

The design uses 5VDC, provided by a 78L05 regulator that drops the input DC (e.g. 9V - 20V) down to 5V.  The 1N4148 is just there to provide circuit protection in case the wrong polarity DC is attached to the input.

The AD8307 has a 92 dB Dynamic Range (-75 to +17 dBm), but if you look at the charts in the datasheet you'll see that accuracy suffers a bit at the ends of this range -- I think a range from -60 to +10 dBm is more reasonable if you want to preserve accuracy.  So I decided to limit my top-end to +10 dBm, which would put my bottom end at -50 dBm.

50 dB of attenuation (externally applied) is required to drop the "application" power range of 0 - 60 dBm down to the meter's input range of -50 to +10 dBm.  Not a problem, as I'll explain later.

The 52.3 ohm resistor to ground at the BNC input parallels the AD8307 input impedance (1.1K || ~0.7pf) and brings it closer to 50 ohms.  (Note, the 0.7pF would be the differential input capacitance.  I.e. 1.4 pf to ground (for each input pin), in series).  Here's an S11 plot of the Input port's Return Loss:


Because of the lower power levels that the AD8307 would be working with, I decided to shield the AD8307 to (hopefully) prevent external RF fields from affecting the reading.  I built the sides of a small shielded box on the ground plane using copper-clad PC stock. Copper tape (soldered to the box sides) caps the box.

Power and the output signal from the AD8307 pass through the box via feedthrough caps (1nF, if I recall their value correctly).

I amplify the AD8307's output signal, whose slope is about 25mv/dB, by a gain of 4, which increases the signal's slope to 0.1V/dB (i.e. 1V/10dB).

For the op amp to generate 0-3V for each of the two ranges, given that the AD8307 output ranges from about 0.9V to 2.4V over an input range of -50 to +10 dBm, I have two switchable "offsets" that connect to the negative-input of the op-amp, thus shifting the AD8307's output down so that 0 dBm (or dBW) corresponds to 0V out of the op amp and a reading of 30 (dBm or dBW) corresponds to 3V.

A "Range" switch (0-30dBm or 0-30 dBW) selects a "course" DC offset, which then can be fine-tuned with the "ZERO" pot.  The pot spans the same amount of voltage, but shifted, when the range switch is toggled (because the total resistance in the voltage divider does not change).  Note that changing ranges requires re-zeroing the meter. 

Two more op amps round out the design.  The first drives a 10 uF cap, which acts as a peak-hold (with its "slow" decay determined by a parallel 1 Meg ohm resistor. This feature is useful when looking at peak-power.  The op amp drives a 2N3904 transistor which is in the feedback loop.  This transistor serves two purposes -- it provides adequate current to charge up the 10 uF cap (the TLC2272 is a bit wimpy), and its base-emitter junction blocks the cap from discharging through the op amp when the output of the op amp drops down below the capacitor's voltage.

Note that the 2N3904's V(BR)EBO is 6V (min), which is greater than the 5V powering the TLC2272 driving the 2N3904's base.  So there's no danger to the transistor's Base-Emitter junction when the 10uF cap's voltage is high and the op amp's output is low.  By the way, I used 2N3904 transistors because I have a bunch on hand.  2N2222  transistors or any other garden-variety NPN would be fine, just as long as it has a minimum Beta of at least, say, 20, and a V(BR)EBO of at least 5V.

A switch allows selection of "slow" or "fast" decay of the peak-hold cap by switching in a 10K resistor to parallel the 1 Meg "slow" decay.

The other op amp drives the meter and isolates the peak-hold cap from the relatively low resistance represented by the meter.  The 500-ohm pot acts as a "Gain" control and it ensures that the reading of the analog front-panel meter correctly corresponds with the input power.  Initial setup requires a bit of tweaking between this control and the "Zero" control to get the meter's needle to read correctly from 0 to full-scale, but once it's calibrated, one shouldn't need to touch the "Gain" control again (thus, it's on the back panel).

A second 2N3904, in the feedback loop of this meter-drive op amp, provides current gain for the its TLC2272.  These op amps are a bit anemic with respect to current-drive, and their output voltage can drop appreciably with loading.  The transistor's current amplification keeps my 3 mA meter from loading down the op amp's output.

If better accuracy is desired, I can read power via an external DVM rather than use the front-panel's analog meter.  A separate BNC connects a DVM to the output of the meter-drive op amp output -- remember, this output goes from 0 to 3V with a slope of 1V/10dB.  There's a series 10K just to limit current in case ESD somehow hits this signal, but it has no effect on the DVM reading, due to the DVM's much higher input impedance.

Additional thoughts:
I used TLC2272 op amps because that's what I had on hand, but these devices cannot source much current (e.g. the 3 mA required for my meter) without experiencing significant voltage drop at their outputs.  A better choice would be the LMC662 family used by N2PK in his Power Meter.  This device would allow you to eliminate both 2N3904 transistors, with the 2N3904 used for the peak detector replaced with a simple diode, e.g. 1N4148.
Something else to try would be to replace the single "Zero" pot with two pots, each selected by the Range switch.  I'd tried to select my voltage divider so that the single pot wouldn't need much tweaking (ideally: none) when flipping between the range switch, but I wasn't successful.  If it turns out that the zero pot, over time, needs no retweaking for a given range, then there's no reason why the single pot couldn't be replaced with two pots mounted, say, on the back panel, instead of on the front panel.

Notes on Construction:

I like to build on copper-clad PC stock because it provides a great ground plane for the circuitry (and I have quite a bit of it in my junk box).  When mounting IC's, I'll mount them "top-up" so that I can see their part numbers.  Pins going to ground are bent down and soldered to the copper plane.  All other pins are bent out so that they are straight out from the sides of the IC (like the wings of an airplane).

To give the device stability (in case, say, only one pin goes to ground), I'll solder a power-bypass cap to the board so that it's lying on its side and then solder the IC power pin to the other lead of the cap.  Or I'll solder a high-value resistor (e.g. 1 Meg) to the board next to a pin.  The resistor sticks up straight and will support the side of an IC that has a pin soldered to to the top of the resistor.

Here's the start of the build...

(click on image to enlarge)

Bending pins is great if you're using DIP packages, but often I'll be using SOIC parts.  What I try to do for these is to purchase little prototyping boards designed to adapt specific SOIC packages (e.g. SOIC-8) to DIP spacing.  I'll then stand these proto boards off the copper plane using either leaded 1 Meg resistors or, where appropriate (e.g. Ground), stiff wire.

In the picture below the two small green boards are the prototyping boards for the two TLC2272 packages.  (You can often find these on eBay).


And here' the completed unit!


To make the front panel overlay I used the same technique that I describe here.


In Operation:

In the photo below the meter is reading 10 dBW (i.e. 40 dBm, or 10 watts), and the DVM is connected to the meter's "To DVM" port, whose output slope is 1V per 10 dB (thus the 1V reading for a 10 dBW signal).


By the way, in the photo above I'm not really driving the meter with a 40 dBm signal -- if no external attenuation is used, the meter's "0-3 mA" scale can be spanned by an input signal ranging from -20 to +10 dBm (when the "Range" switch is in its "High" position (30 dBW)).  Only with an external 50 dB attenuator will the scale accurately reflect the actual power (e.g. 0-30 dBW or 0-30 dBm).  So, in the photo above I'm actually using no attenuation and I'm driving the meter with a -10 dBm signal from my RF Generator.

So it's worthwhile noting that other values of attenuation can be used in lieu of 50 dB.  The table below shows how the measurement range would shift with different values of input attenuation.  For example, if I used 20 dB of external attenuation and set the Range switch to its Low (30 dBm) position, then I could directly measure (using the meter's scale) 0-30 dBu (dB microwatt).



External 50 dBm Attenuator:

As I mentioned earlier, this design requires some form of external attenuation for the 0-60 dBm measurement range.  Fortunately, the shack/lab here has a variety of ways to attenuate RF signals:



(Back:  Bird 200 watt, 30 dB attenuator.  Front-left: 50 watt dummy load with 50 dB attenuator.  Front-right: homebrew -24 dB directional coupler.)

When testing and repairing transmitters, I usually use this setup:


Although the dummy load (military DA-437/GRC-103(V)) is rated at 50 watts, it seems to work fine with 100 watt transmitters (admittedly I minimize "key-down" time).  The 40 dB attenuator is just a voltage divider that divides the voltage at the dummy load (e.g. 70.7 volts @ 100 watts) by a factor of 100 (40 dB).  I chose this scheme of attenuation because it doesn't require the attenuator to dissipate large amounts of power and the 2.5K ohms has little impact on the 50 ohm impedance seen by the transmitter.

Here's what the 40 dB attenuator looks like:

(click on image to enlarge)

The series 2.5K resistor was tweaked (by paralleling another resistor) to give 40 dB +/- 0.15 dB of attenuation (while attached to the dummy load) over the range of 1 to 54 MHz.  (To keep the frequency response flat over that range I had to add a capacitive "gimmick" to ground next to the series resistors -- essentially some copper tape tied to ground that I wrapped around the resistor bodies). 

Note, too, that if driving the dummy load with 100 watts of power, the 2.5K resistor will need to dissipate about 2 watts, so select values accordingly. And to get 40 dB of attenuation there needs to be a 50 ohm load (e.g. the input of a spectrum analyzer) connected to the output (attenuated) port. 


Converting dBm to Watts:

To convert dBm to Watts, you can use the formula:

Power(watts) = (0.001)*10(dBm/10) 

But the table below might be simpler to use:


These readings can be easily scaled to other powers.  For a given dB range, multiply the milliwatt reading by the appropriate power of 10.  For example:
  • For the range 0-10 dBm, multiply "milliwatts" by 1
  • For the range 10-20 dBm, multiply "milliwatts" by 10
  • For the range 20-30 dBm, multiply "milliwatts" by 100
So, if a reading is 11 dBm, the power would be 13 milliwatts.  Similarly, 21 dBm would be 130 milliwatts.

If the Range switch is set to 30 dBW, then "milliwatts" is replace by "watts", and a meter reading of, say, 11 dBW would correspond to 13 watts.


References:

N2PK Forward Power and Return Loss Meter  Design using two AD8307 Logarithmic Amplifiers.

Analog Devices AD8307 92 dB Logarithmic Amplifier, Datasheet

TI TLC2272 Dual Rail-to-Rail Op Amp, Datasheet


Links to my Directional Coupler blog posts:

Notes on the Bruene Coupler, Part 2

Notes on the Bruene Coupler, Part 1

Notes on HF Directional Couplers

Building an HF Directional Coupler

Notes on the Bird Wattmeter

Notes on the Monimatch


Standard Caveat:

As always, I might have made a mistake in my equations, assumptions, drawings, or interpretations.  If you see anything you believe to be in error or if anything is confusing, please feel free to contact me or comment below.

And so I should add -- this design and any associated information is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Friday, February 6, 2015

More Notes on Directional Couplers for HF -- the Bruene Coupler, Part 2

This is Part 2 of my notes on the Bruene Directional Coupler.  Part 1 is >>here<<.  I strongly suggest reading Part 1 -- it explains a more easily understood variation of Bruene's circuit.

So, assuming you're familiar with the basic principles of the Bruene coupler, let's move on to Bruene's design!

Let's start with the schematic of the Bruene Coupler:

(click on image to enlarge)

This coupler was used in the Collins 302C-3 Directional Wattmeter and it was also described in an article written by Warren Bruene in QST magazine ("An Inside Picture of Directional Wattmeters," QST, April, 1959).

This architecture uses the same principles as the ones described in Part 1:  to generate a voltage representative of a Forward Wave on a transmission line, a sample of the line current is added to a sample of the line voltage.

And to generate a voltage representative of a Reflected (or Reverse) wave on a transmission line, a sample of the line current is subtracted from a sample of the line voltage.

What differentiates Bruene's design from the other variations is that those variations keep the circuitry generating the voltage sample independent of the circuitry generating the current sample.  Which is to say, with those other designs I could take a voltmeter and independently measure the voltage sample and the current sample.  I cannot do that with Bruene's circuit -- voltage-sample and current-sample generation are bound together!

You're probably looking at the above schematic right now and wondering what the heck I'm talking about.  You see two voltage dividers, so they must be generating the voltage sample of the Transmission Line voltage, while the transformer (with the two 10 ohms resistors) is generating a positive and negative voltage related to the current on the Transmission Line.  Everything's copacetic, right?

But take a closer look.

Vfwd and Vref. which are the sum and differences of the voltage and current samples, are measured at the divider-taps of the two capacitive voltage dividers.  So the voltage dividers can't just be generating samples of the Transmission Line's voltage.  They must somehow be involved with summing in the Line current samples, too!

The key to the summing are the two diodes. But how are the diodes doing this?

Each diode is actually behaving as a "shunt rectifier" (see Chapter 2 in this >reference<).  And this shunt-rectification process is differencing (rather than summing) the voltage-sample and the current sample that are at the diode's Cathode and Anode, respectively, with the result appearing at the diode's Cathode (the junction of the two voltage-dividing capacitors).

To illustrate this concept better, I've done some basic simulations using LTSpice.  My simulation-circuit is a much simplified version of the part of Bruene's coupler that generates Vref.  Please note that:
  1. I use the same capacitor values for the voltage divider as Bruene.  Vsource is the voltage across the transmission line.  It's a large amplitude because the voltage divider has a huge division ratio.
  2. The voltage created by the current-sampler, originally generated by a current source (the transformer) driving its current through the resistors, has been replaced by its Thevenin equivalent, but I've made the series-R very small to keep its effect from confusing this discussion.  I've called this voltage source Vinduced.
  3. I've added a "switch" between the Diode's anode and Vinduced so that I can look at the voltage sample (at the voltage-divider tap) for a full 360 degree cycle without it being influenced by the current sample.
Below is my first simulation.  To illustrate the shunt-diode's operation, we will first simulate Vinduced as a DC voltage source of 0 volts.  For the shunt-diode I've chosen a 1N916 (a silicon diode).  The actual circuit uses Germanium diodes (I believe).

I've annotated the schematic below to describe the basic functions of each part of the simulation circuit.

(click on image to enlarge)

So what's going on in the plots?  Let's look...

(click to enlarge)

I've defined an "Offset" voltage across C2.  It can also be thought to be across C1, too, and it's effectively the amount that Vsample has been shifted.  I'll explain this further a bit further down the post.

Looking at the plots, at 100nS we connect Vinduced to the diode and so at the 150 ns mark, when the voltage at Vsample is driven negative by Vsource, the diode's cathode connected to Vsample will start to go negative during the negative half-cycle of Vsource.  Because the diode's anode is tied to a 0V source, the diode will turn ON when its cathode voltage drops about 0.7V below its anode voltage.  I.e. when Vsample is approximately -0.7V.

When the diode turns ON, it effectively "clamps" the voltage at Vsample to -0.7 volts.

However, Vsource is still going negative.  So we have the "+" terminal of C2 (see simulation schematic) clamped at -0.7V and C2's "-" terminal (attached to Vsource) continuing down, which causes charge to flow onto C2's plates.   When Vsource reaches its negative peak, this charge can be calculated from the equation Q = C*V:

Q(max) = C2*Vsource(pk)

Actually, in the real world the voltage is slightly less than this by Vf, the forward voltage  drop across the diode.

Now, as Vsource starts to rise again from its negative most transition, the diode turns off, but not immediately, which is the "real-world" diode turn-off effect you see on the Red line in the plot above.

With real-world diode effects, the plots can get a bit confusing, so let me introduce a simpler example that uses an "ideal" diode.  I'll define the diode to have a Vf of 0V:  if its Cathode is at all negative with respect to its Anode, it is ON, and if its Cathode is at all positive with respect to its Anode, it is OFF.

Shunt Rectifier with Vinduced as a DC source and an Ideal Diode:

Here's the circuit and the simulation:

 (click on image to enlarge)
 

Looks a bit cleaner, doesn't it?

So what is happening?

From 0-100ns the diode's Anode is floating because the switch is open, and it is essentially out of the circuit.  Vsample is just Vsource divided down by the capacitive voltage divider (C1 and C2), and therefore Vsample goes between -3.4V and +3.4V.  Its average value is 0V.

At 100ns the switch turns ON, and the diode's Anode is connected to Vinduced, which here is a 0VDC source.

From 100ns to 150ns, Vsource is positive and thus Vsample is still positive.  The diode is OFF (it is back-biased:  Cathode positive, Anode at 0VDC), and Vsample is still just Vsource divided down by the capacitive voltage divider (C1 and C2).

But starting at 150ns, Vsource goes negative and it tries to take Vsample below 0, too.  But Vsample cannot go negative, because as soon as it begins to go below 0V, the diode turns ON and clamps it to the voltage at its Anode, 0VDC.

So as Vsource goes negative, Vsample sticks at 0V, and the voltage across C2 (1.7pf) gets larger and larger, until, at 175ns when Vsource is at its negative peak, the voltage across C2 (Vc2) is at its max:

Vc2(max) = 1000 V.

At this point C2 will have its maximum charge on its plates.  From the formula Q = C*V, we know that:

Qc2 @ max V = C2 * Vc2(max) = C2*Vsource(peak)

(...given the voltage at the diode's Anode is 0VDC and that the diode is ideal with a 0V Forward-voltage drop.)

Also, note that the voltage across C1 (500pf) is 0V (Vsample is clamped to 0V).  Therefore the charge on C1 is 0 coulombs.

So at 175ns, when Vsource has reached its most negative excursion, the maximum charge is on the two caps:

Qmax = Qc1 + Qc2 = 0 + C2*Vsource(peak)    (equation 1)

Now, as time moves on from 175ns, Vsource just starts to rise up from its negative peak. Vsample, which was at 0 V, starts to go positive and the diode immediately cuts off (because Vsample = Vsource + Vc2(max)).  The diode is now effectively out of the circuit.

The diode is back-biased and out of the circuit.  There's no path for the charge of Qmax on C2 to go but to C1 (there's no other path but to C1).  So, as time continues, no charge will by lost by the C1/C2 combo, and no charge will be added -- the diode never turns on again.  And so the total charge on C2 and C1 (=Qmax) will flow back and forth between C2 and C1 as Vsource goes up and down.

No charge is being added, no charge subtracted.  Therefore, at all times (after we've first charged C2 via the diode clamp), the charge across C1 and C2 must satisfy this equation:

Qc1 + Qc2 = Qmax

And therefore (using Q = C*V):

C2*(Vsample - Vsource) + C1*Vsample = Qmax

Rearranging:

Vsample = Vsource*C2/(C1+C2) + Qmax/(C1+C2)   (equation 2)

So, if we know Qmax (and we do), we can calculate what Vsample will be for any value of Vsource.

The equations above do this for an ideal diode with 0V attached to its anode.  We can derive a more general equation for Qmax (rather than use equation 1 above) that takes into account the voltage connected to the diode's Anode and also its Forward-voltage drop (i.e. a non-ideal diode):

Qmax = C2*(Vsource(pk) + V(anode) - Vf(diode)) + C1*(V(anode) - Vf(diode))   (equation 3)

Using equations 2 and 3, we can now calculate Vsample for different voltages connected to the diode's Anode, per the table below (I'll keep the diode ideal, though, with Vf = 0):

(click on image to enlarge)

You can see that "Vsample without diode" ranges from -3.4 to 3.4V.  If we look at, say, the diode with 0V attached to its anode, the range of Vsample now becomes: 0V to 6.8V -- a shift of 3.4V, the peak voltage of the non-diode Vsample.  So Vsample's average voltage is now 3.4 volts instead of 0, and if we were to filter out the AC waveform on Vsample we would measure 3.4 VDC.

The previous example was with a DC source connected to the diode's Anode and demonstrates how the shunt-rectifier can create a DC voltage from an AC signal.  What happens if the Anode is instead connected to an AC source representing the voltage from the current-sample?


Vinduced as an AC source:

So all is fine and good if Vinduced is a DC source.  Let's change Vinduced from a DC voltage source to something instead the simulates what is actually happening in the Bruene Coupler -- an AC source of the same frequency as Vsource, but with a phase offset.

We should still see a DC offset voltage generated on the capacitors in an analogous way to when Vinduced was a DC source. But the level of this offset voltage should now depend on the phase difference between the Vinduced and Vsource.

To keep the analysis simple, I won't be changing the amplitudes of Vsource or Vinduced, just their phase relationship.  And note that the amplitude of Vsource has been chosen so that, if the diode were removed, Vsample = Vinduced in amplitude.  

Let's examine how the voltage offset changes with phase (see the RED line in the plots below).  With the two waveforms 180 degrees out of phase, LTSpice simulates this value to be 6.8V:

(click on image to enlarge)

With the two waveforms 180 degrees in phase (0 degrees delta), LTSpice simulates this value to be 0V:

(click on image to enlarge)
 

With the two waveforms 45 degrees out of phase, LTSpice simulates this value to be 2.61V:

(click on image to enlarge)

With the two waveforms 180 degrees out of phase, LTSpice simulates this value to be 4.8V:

(click on image to enlarge)
 

And finally, with  the two waveforms 135 degrees out of phase, LTSpice simulates this value to be  6.28V:

(click on image to enlarge)

Summarizing the DC offset calculated by the simulations:


But is this DC value that LTSpice calculates actually equal to the difference (or sum) of the two sine waves representing our voltage sample and current sample?  This is a requirement that must be met if we are to calculate Vfwd and Vref by adding or subtracting the voltage and current samples.

Let's check this by doing some separate math, using the same peak voltages and phases that I used in the LTSpice examples above...

If summing sine waves, I can use the following equations to calculate the the resulting amplitude of the summed waveform (amplitude can be expressed as Vrms, Vpp, etc.  I'll use Vpeak, which, from our simulations, is 3.4V).


Knowing the amplitudes of the sine waves and the different phases, I can plug these numbers into Excel and solve for c, the summed-amplitude (called Vc in the table, below):

(click on image to enlarge)

The "LTSpice" column is 180 degrees out of phase with with "formula" column because the "shunt-diode" circuit is actually a differencer, not a summer, which is equivalent to adding 180 degrees to the phase shift "α" in the formula for c.

If we compare the table results from the math with our simulation results (see the earlier table), we find that they are exactly the same values (after first taking into account the shunt-diode's additional 180 degree phase shift).  Which means that the shunt diode circuit is indeed summing (or differencing) the voltage waveforms of the voltage sample and the current sample!

And therefore, it is functionally equivalent to the earlier Bruene coupler variants that we looked at in Part 1.

Now that I've shown that the shunt-rectifier sums or differences the voltage and current samples, I'll refer you back to Part 1 for a recap of circuit operation in Transmission Line and non-Transmission Line environments.

The only thing to keep in mind is that, because the shunt-rectifier is actually a differencing circuit, not an adder, the side of the coupler that, in part 1, generated Vfwd (that is, the left side of the circuit in the diagrams), now generates Vref.

And the opposite side (the right side of the circuit in the diagrams) now generates Vfwd.

Other than that, the same analysis applies!


Bruene Variant, Daiwa SWR Meters:

While looking around the shack I came across a Daiwa CN-620B wattmeter that I pickup up several years ago.  Opening it up, I discovered it's a variation of the Bruene Coupler that uses Bruene's "shunt rectifier" method of creating a DC voltage from the voltage and current samples.

(click on image to enlarge)

Inside the Daiwa CN-620B (click on image to enlarge))

 The main differences that I can see between it and Bruene's design are:
  • Daiwa uses two transformers, one for Vref for Vfwd, compared to Bruene's one.
  • The Daiwa transformers have fewer windings than Bruene's (I believe the latter has 60).
A number of other Daiwa wattmeters (e.g. CN710, CN720) use the same or similar design.  Curiously, when I look at the values of the components on the CN-620B's PCB, they are significantly different than the values listed in the schematic! Don't know why.


Links to my Directional Coupler blog posts:

Notes on the Bruene Coupler, Part 1

Notes on HF Directional Couplers

Building an HF Directional Coupler

Notes on the Bird Wattmeter

Notes on the Monimatch

Notes on the Twin-lead "Twin-Lamp" SWR Indicator


And some related links from my Auto-Tuner posts:

Part 5:  Directional Coupler Design

Part 6:  Notes on Match Detection

Part 8:  The Build, Phase 2 (Integration of Match Detection)


Bruene Coupler References:

Bruene, Warren, "An Inside Picture of Directional Wattmeters," QST,  Apr., 1959.  Includes both a good explanation of the Monimatch operation and a design for a directional wattmeter whose directional coupler topology would later be known as the "Bruene Coupler."

Collins 302C-3 Directional Wattmeter, PDF Manual containing schematic.

Rush, James, Jr., "The Mini-Mono-Monimatch," QST, Mar., 1965.  Although called a Monimatch in the title, the design is actually more similar to Bruene's directional coupler.

Bold, Gary, ZL1AN, "The Bruene Directional Coupler and Transmission Lines," PDF. This PDF gives an excellent explanation of the Bruene Coupler.

Kiciak, Paul, N2PK, "An HF In-Line Return Loss and Power Meter," PDF.  Constructions details of a power meter using a Bruene Coupler.  Contains an explanation by the other of why he prefers the Bruene coupler of the Tandem-Match Coupler.  Also interesting because the author separates the voltage-sampler from the current sampler and uses the differential inputs of an AD8307 to do the required addition (or subtraction) to get FWD and REF voltages.

(This web page could be useful for understanding the sampling method used in the N2PK meter:  http://www.g3ynh.info/zdocs/bridges/magdiff/part1.html )

Lewallen, Roy, W7EL, "A Simple and Accurate QRP Directional Wattmeter," QST, Feb, 1990, PDF.  Interesting variant of the Bruene coupler.  Roy uses two transformers for the voltage sample in lieu of capacitor voltage dividers.

White, Ian, G3SEK, "Inside a Directional Wattmeter," RadCom, Sept., 2002, PDF.  Discussion and a bit of analysis of Bruene coupler.  Includes Bruene's phase-relationship diagrams.

http://www.g3ynh.info/zdocs/bridges/reflectom/part1.html  interesting analysis

http://www.g3ynh.info/circuits/Diode_det.pdf  Diode detectors -- includes some info on shunt detectors, which is what Bruene's design uses.


Other references of generally interest:

http://www.g3ynh.info/zdocs/bridges/Xformers/part_1.html  great discussion on current-transformers for directional coupler applications

http://www.g3ynh.info/zdocs/bridges/Xformers/part_2.html Part 2 of current-transformers

http://www.g3ynh.info/zdocs/bridges/Xformers/part_3.html  And part 3, the last part, of current-transformers

http://www.g3ynh.info/zdocs/bridges/index.html  Indexes numerous topics.  Lots of great info to be found here!

http://www.richtek.com/assets/AppNote/AN008_EN/AN008_EN.jsp  Common-Mode choke model


Final Caveats:

As always, I might have made a mistake in my equations, assumptions, or interpretations.  If you see anything you believe to be in error, or if anything is confusing, please feel free to contact me.

Wednesday, February 4, 2015

More Notes on Directional Couplers for HF -- the Bruene Coupler, Part 1

In a previous post I looked at how the popular "Tandem-Match" Directional Coupler worked.  Recall from that analysis:
  1. I analyzed the Tandem-Match coupler in terms of "lumped" circuit elements, not distributed elements.
  2. The coupler works by taking a sample of the voltage across the transmission line at a single point and a sample of the current through that point.  
  3. The Tandem-Match Coupler creates both of these samples with transformers.
  4. The voltage at the "Forward Port" (that we measure on our meter) is calculated by adding the current and voltage samples.
  5. The voltage at the "Reflected Port" is calculated by subtracting the current and voltage samples.
  6. The voltage at the "Reflected Port" is 0 when the load is a real resistance equal to the value of the resistors terminating the measurement ports (Forward and Reflected Ports).  Thus, the resistor values should be selected to be the same as the characteristic impedance, Zo, of the transmission-line system into which the the coupler is inserted.
  7. I explained operation of the coupler both in terms of Forward and Reflected waves and also without using waves, instead in terms of only the voltage driving the coupler and the load at its output port.
Another popular directional coupler topology for HF is known as the Bruene Coupler.  This coupler was used in the Collins 302C-3 Directional Wattmeter and it was also described in an article written by Warren Bruene in QST magazine ("An Inside Picture of Directional Wattmeters," QST, Apr., 1959).

I'll analyze the original design as well as some of the variants it inspired.  Again, I will use lumped-element analysis, and I'll present explanations in terms of forward and reflected waves and also in terms of the voltage being applied to the coupler and the load connected to its output port.

My goal is to present you with an explanation that is understandable -- this will not be a rigorous analysis.  So, when possible, I will take advantage of models and assumptions that will simplify the math and hopefully help you grasp the underlying principles.

So here we go!  Let's start with the schematic of the Bruene Coupler:

(click on image to enlarge)

There are two capacitive voltage dividers; one is used for generating the Vref voltage, the other is used for the Vfwd voltage.  Let's call these the voltage samples, but in fact, they are more than that because of the two diodes connected to them.

Line current passing through the 1-turn primary of a 1:60 turn transformer and induces current in the 60-turn secondary.  This current, in turn, is transformed into voltage as it passes through the two 10 ohm resistors.   One voltage is positive with respect to ground, the other is negative.  These two voltages represent the current sample.

Now the explanation gets a bit tricky.  Take a look at the schematic -- we're measuring Vref and Vfwd at the voltage-divider taps.  This means that the two caps comprising each voltage-divider aren't simply generating a voltage divided down from the line voltage, they are also involved in adding or subtracting the current-sample voltages, too, so that, when we measure the voltage across either 500 pf cap, it's actually the sum or difference of voltage and current samples.

It wasn't obvious to me how the summing/differencing of voltage and current samples at the voltage-divider nodes was being accomplished.  It turns out the diodes linking the voltage-dividers and the current samples play a crucial role -- they are creating a DC voltage at the voltage-divider "tap" to which their Cathodes are connected.  But not as series rectifiers -- the diodes instead are serving as "shunt" rectifiers.

The DC voltage they create is a function of the phase and amplitude differences between the voltage and current samples.  I'll explain the role these caps and diodes play in more detail as it's interesting, but it's also a  a bit complex.  So instead let me kick off this post with an analysis of a Bruene variant that isn't quite so daunting.

(I'll get back to Bruene's original circuit, but it will be in another blog post -- look for Part 2!).

And before I get any further into this discussion, let me also make this important point:

I will first look at the Bruene Coupler in terms of Forward and Reflected waves. The Bruene Coupler, being made of "lumped elements" (in the first example below: 2 capacitors, 1 transformer, and 1 resistor), is only looking at the voltage and current present at its output port.  It has no idea what the load is, or even how the load is connected to the port.  The load might be a resistor or other component simply clipped onto the output connector with test leads, or it might be a length of transmission line with a load (either known or unknown) at its other end.

Irrespective of what the load actually is (transmission line, clipped-on component, or whatever), the Bruene Coupler gives us voltage readings that can be interpreted in terms of Forward and Reflected waves.  It is important to remember:  these readings should only be interpreted as representing actual Forward and Reflected waves when the Bruene Coupler is connected in a transmission line with the same characteristic impedance, Zo, as the Coupler's designed-for target impedance!

I'll start this analysis assuming the Bruene Coupler is inserted into a transmission line of the designed-for (target) characteristic impedance, Zo.  I will follow that with a look at its operation in a non-transmission line environment.


Bruene Coupler Variant, ZL1AN

ZL1AN has written an excellent article explaining a variation on Bruene's original coupler design, and I strongly recommend you take a look at it.  Tthe Heathkit HM-102 (and I believe also the Drake W4) used this version of the circuit.  I'll simplify the analysis a bit by assuming that the current-sampling is done with an ideal transformer.

Here's ZL1AN's circuit:

(click on image to enlarge)

Let's take this circuit and add a bit more information...

(click on image to enlarge)

In the image above:
  1. "V" is the voltage across the transmission line at our "point" of measurement.  We will assume that the voltages are the same at the input and output ports of the coupler -- there is no drop through the coupler.
  2. "I" is the current on the transmission line through that point.
  3. I've replaced the resistor load (of resistance "R" ohms) across the transformer secondary with two series resistors of value R/2.
  4. The voltage at the junction of these two resistors is Vc, because the voltage at the center-tap of the transformer is Vc, and these two resistor form a divide-by-2 voltage divider across the secondary that essentially places their junction at the same voltage as the center-tap.
  5. Current "I" through the transformer primary induces a current "Is" in the secondary.
  6. Is = I/(2*N), and it flows in the opposite direction of I (per the transformer "dots" that I've shown).
  7. We assume that no current flows out either the Vfwd or Vref measurement ports.
  8. Therefore there is no load on our capacitive divider's voltage "Vc" (that is, there isn't another path from Vc to ground that parallels the path through C2).
  9. Therefore the voltage at Vc is:
Vc =  V*(1/(jw*C2)) / ((1/(jw*C1))+(1/(jw*C2)))

If  1/(jw*C1) >> 1/(jw*C2), this simplifies to: Vc = V*C1/C2


 (click on image to enlarge)

Vfwd and Vref are easily calculate from the series addition of Vc and the appropriate voltage generated by the current sample:

Vfwd = Vc + Is*R/2

Vref = Vc - Is*R/2

Substituting in our equations for V and Is, we get:

Vfwd = V*(C1/C2) + I*R/(4*N)   (equation 1)

Vref =  V*(C1/C2) - I*R/(4*N)   (equation 2)

Analysis Using Forward and Reflected Waves

Let's first analyze this circuit in terms of Forward and Reflected waves passing through our coupler.

The Forward and Reflected waves each have a voltage and a current: Vf and If are the voltage and current of the forward wave, and Vr and Ir are the voltage and current of the reflected (or reverse) wave.

 (click on image to enlarge)

The total voltage on the line, V, at any point is the sum of Vf and Vr at that point.

V = Vf + Vr

And the total current on the line, I, at any point is the difference of If and Ir at that point (they subtract because Ir is flowing in the opposite direction of If).
I = If - Ir

We also know:

If = Vf/Zo

Ir = Vr/Zo

Where Zo is the characteristic impedance of the transmission line.

So, substituting and rearranging, equations 1 and 2 become:

Vfwd = (Vf+Vr)*(C1/C2) + (Vf/Zo - Vr/Zo)*R/(4*N)

Vref =  (Vf+Vr)*(C1/C2) - (Vf/Zo - Vr/Zo)*R/(4*N)

Regrouping terms:

Vfwd = Vf*((C1/C2) + R/(4*N*Zo)) + Vr*((C1/C2) - R/(4*N*Zo))

Vref = Vf*((C1/C2) - R/(4*N*Zo)) + Vr*((C1/C2) + R/(4*N*Zo))

Notice what happens if we select our components such that they meet the following requirement:

C1/C2 = R/(4*N*Zo) = K    (equation 3)

The two equations reduce down to:

Vfwd = Vf*2*K

Vref = Vr*2*K

So, if we select C1, C2, R, and N such that the satisfy the relationship above, then the voltage we measure at Vfwd is solely related Vf, the voltage of the Forward wave, and the voltage we measure at Vref is solely related to Vr, the voltage of the Reflected (or Reverse) wave!


Analysis in a non-Transmission Line Environment:

It's instructional to analyze the operation of the directional coupler just in terms of the components themselves, the voltage applied to the coupler, and the load at its output port without using the concepts of waves.  After all, the coupler consists of lumped-elements, so there's no real reason to think of its operation in terms of waves.

So let's draw our circuit like this, where V is the voltage source driving the input of the Bruene Coupler and Zload is connected to its output.


We'll use the same assumptions that we used above.  Therefore equations 1 and 2 still hold:

Vfwd = V*(C1/C2) + I*R/(4*N)   (equation 1)

Vref =  V*(C1/C2) - I*R/(4*N)   (equation 2)

This time, rather than expressing V and I in terms of waves, we will note that "I" is simply "V" divided by Zload:

I = V/Zload

If we substitute this equation into equations 1 and 2, we get:

Vfwd = V*((C1/C2) + R/(4*N*Zload))   (equation 4)

Vref = V*((C1/C2) - R/(4*N*Zload))     (equation 5)

Now let's take equation 3:

C1/C2 = R/(4*N*Zo) = K    (equation 3)

Let's take that second half:

R/(4*N*Zo) = K

And rearrange it:

R/(4*N) = K*Zo  (equation 6)

Where Zo can be considered to be our "Target" impedance.  Usually this is selected to be the impedance of the transmission line, but it needn't be.

Now let's plug equations 3 and 6 into 4 and 5 and reduce.  We get:

Vfwd = V*K*(Zload + Zo)/Zload

Vref = V*K*(Zload - Zo)/Zload

Well, these are interesting equations.  If Zload equals our target Zo that we selected our components for (e.g. 50 ohms), then Vfwd = V*2*K and Vref = 0.

We could use these equations as they are, but note what happens if we take these two voltages and divide one into the other.  We get something that ought to look familiar:

Vref/Vfwd = (Zload - Zo) / (Zload + Zo)

Which is the definition of Reflection Coefficient!

So we can use the voltages measured at Vref and Vfwd to determine an impedance relationship (i.e. imbalance) between the actual Zload and coupler's "target" impedance of Zo.  And this impedance relationship is exactly the same as the Reflection Coefficient.

SWR is easily calculated from the Reflection Coefficient:

SWR = (1 + |Reflection Coefficient|)  / (1 - |Reflection Coefficient|)

Note that, although SWR implies the presence of Forward and Reflected waves, we have no guarantee that what we measure (or calculate) to be SWR or Forward/Reflected power is actually what is happening in our system.   I'll quote G3YNH:
"[The bridge] can only infer the existence of reflected power from the difference between the actual load impedance and the target load impedance. To understand this point, consider an SWR bridge designed to balance when the load is 50+j0Ω. If we connect this bridge directly to a 100Ω load resistor, it will declare an SWR of 2:1. The resistor is not reactive however, and so will absorb all of the power delivered to it and reflect none. The 2:1 SWR reading is only true when the bridge sees an impedance magnitude of 100Ω (or 25Ω) at the input to a 50Ω transmission line. The bridge is just an impedance bridge, it has no special psychic powers, and its readings are only true when it is inserted into a line having the same characteristic resistance." 
So, summing up:

It should be evident from the above analysis that we don't need to rely on the concepts and Forward and Reflected waves to understand how the Bruene coupler operates.

Our typical "Bruene" SWR meter is really just calculating a relationship between the load at its output port (Zload) and its own design parameters (i.e. C1/C2 = R/(4*N*Zo) = K).  And this relationship is equivalent to the Reflection Coefficient if "Zo" in the relationship: "C1/C2 = R/(4*N*Zo) = K" is the same as the characteristic impedance of the transmission line, should the coupler be connected to a transmission line.

It's  important to note that Zload could be a load connected directly to the coupler's "OUT" port with a couple of wires, or it could be the impedance "presented" to the port by a long length of transmission line (an impedance determined, at that point, by the interaction of the Forward and Reflected waves).

The coupler doesn't care "how" the load is connected to its OUT port.  It's just looking at the voltage across the OUT port and the current through the OUT port.  It doesn't know anything else about the load except for this voltage and current relationship at its OUT port.  For example, if the OUT port happens to be connected to a transmission line, the coupler has no knowledge of the line's Zo. It doesn't even know if there's a transmission line attached, nor that the impedance it sees at its OUT port might be due to the interaction of Forward and Reflected waves.

For this reason, never assume that the meter reading is the actual Reflection Coefficient, Γ, or that the SWR reading is the actual SWR reading of the line.  It might not be.  We are really just measuring the relationship between Zload (as it appears at the OUT port) and the design parameters of the coupler.  Only if the "Zo" in the design relationship "C1/C2 = R/(4*N*Zo) = K" equals the actual characteristic impedance, Zo, of the transmission line would we truly be measuring the Reflection Coefficient.


And on that note, I'll end this non-Transmission Line analysis!


Frequency Insensitivity:

If you refer back to the equations for Vfwd and Vref, you should notice something interesting:  there are no j*w terms (where omega (w) = 2*pi*frequency).  This means that the voltage-divider voltage, Vc, and the voltage generated by the current-sample via the transformer are both constant over frequency.

Of course, in the real world nothing is perfect and there will be effects due to strays and parasitics.  Never the less, if designed correctly (to deal with strays), the frequency response should be flat over a broad range of frequencies.

Which leads to an interesting observation:  If the voltage divider is independent of frequency, why use capacitors?

Well, one doesn't need to use caps, we could just as easily use inductors (whose "jw" terms will cancel), or even resistors!

Which leads me to another variant of the Bruene coupler, which can be found in G3SEK's "In Practice" column in the September, 2002 issue of Radcom...


Bruene Coupler Variants, G3SEK:

One of the coupler's described in G3SEK's column looks very similar to the coupler described by ZL1AN, but there are a few differences:



The first is that a resistive voltage divider replaces the capacitive voltage divider.

The second is that the voltage sample from the resistor divider now feeds the junction of two resistors instead of the transformer center-tap.

I'll skip analysis -- the process is no different that what we've done earlier in this post.

Frankly, I don't know if it's better to feed the voltage-sample to the common-point between two resistors, as done above, or to the center-tap of the transformer secondary, as done by ZL1AN.  Our concerns are:  what is the effect on Directivity, and what is the effect on Frequency Response?


Other Riffs on the Same Theme:

Vc can feed both the resistors and the transformer center-tap.  (Any negative effects?  I don't know.)


The common point between the two resistors could be tied to ground.  But I'm not sure I'd recommend this -- it puts the two resistors in parallel with C2, which means that the frequency response of Vc will no longer be flat.  (The secondary of the transformer acts as an auto transformer, and thus, if Vc feeds its center-tap, it will look like a low impedance (i.e. short) to its two ends.  Which is to say -- it doesn't act as a common-mode choke when feeding the center-).


Here's another interesting variation, found on G3YNH's website (worth a visit!).  A single core is used, but the secondary consists of two windings that are not interconnected.  Thus, two voltage dividers are necessary in order to create voltage samples for the two independent windings.


Analysis is similar to the other variants.  Voltage and current samples add on the left-hand side.  And they subtract on the right-hand side.  I don't know what the advantage is to doing it this way, though.  But one advantage might be that the transformer's secondary, no longer center-tapped, doesn't act like a "shorting" auto-transformer to Vc's common-mode path to ground.  Now, there is some impedance in the path (due to the inductance of each of the secondary coils), but I'm not sure how effective this would be, as it will depend upon the resistance "R" of the two resistors which parallel these coils.

That covers the variants that I've seen that are obviously similar to the ZL1AN topology (and thus will analyze in a similar fashion).  I'll introduce below a few more variants that stray a bit further afield (but not by much).

The models presented thus far are in some cases simplified models of the actual circuits.  I've left off components that might be related to frequency compensation, or detection, or other functions deemed secondary, because I felt they would detract from understanding the underlying theory of operation.  If you're interested in more details, please click on the links I've provided!

As to the positives and negatives for each topology, I wish I had some answers, but I don't. If you have any experience or thoughts on the subject, please feel free to let me know, either via comments to the post or email.

Continuing on...


Bruene Variant, W7EL:

Here's an interesting take on the Bruene Coupler, published by W7EL in the Feb, 1990 issue of QST magazine.

Per the article, this version has +/- 7% accuracy over the range of 1 to 432 MHz.  Quite impressive!

At first glance the design looks similar to the Tandem-Match coupler, but it really is a variant on the Bruene topology; the current sample is either added-to, or subtracted-from, the voltage sample.  (With the Tandem-Match coupler, it's the voltage sample that is either added-to, or subtracted-from, the current sample).

The two transformers to ground create two voltage samples of the same polarity and whose value is V/N, where V is the voltage on the line.

The "series" transformer samples the line current, I, and its secondary generates a current that is I/N in amplitude.

If "I" is flowing from left-to-right in the diagram below (into the 1-turn primary's "dot"), the secondary current runs from left-to-right (out of the secondary's "dot").  This current creates a positive voltage of amplitude I*51/N across the left-hand resistor and a negative voltage (w.r.t. ground) across the right-hand resistor of amplitude -(I*51/N).

The voltage samples generated by the two voltage-sampling transformers are in series with their respective resistors, and thus Vfwd is the sum of the positive voltage across the left-hand resistor and the positive voltage across left-hand secondary, while Vref is the sum of the negative voltage across the right-hand resistor plus the positive voltage across the right-hand secondary.


The only negative that I can see with respect to the design is that you need to wind 3 transformers!


Bruene Variant, N2PK Power Meter:

N2PK cleverly used the differential inputs of the AD8307 Log Amplifier to do the summing and differencing of the voltage and current samples in his homebrew Power Meter.



That's it for the analysis of Bruene variants!  Analysis of the actual Bruene design will follow in Part 2...


Links to my Directional Coupler blog posts:

Notes on the Bruene Coupler, Part 2

Notes on the Bruene Coupler, Part 1

Notes on HF Directional Couplers

Building an HF Directional Coupler

Notes on the Bird Wattmeter

Notes on the Monimatch

Notes on the Twin-lead "Twin-Lamp" SWR Indicator


And some related links from my Auto-Tuner posts:

Part 5:  Directional Coupler Design

Part 6:  Notes on Match Detection

Part 8:  The Build, Phase 2 (Integration of Match Detection)


Bruene Coupler References:

Bruene, Warren, "An Inside Picture of Directional Wattmeters," QST,  Apr., 1959.  Includes both a good explanation of the Monimatch operation and a design for a directional wattmeter whose directional coupler topology would later be known as the "Bruene Coupler."

Collins 302C-3 Directional Wattmeter, PDF Manual containing schematic.

Rush, James, Jr., "The Mini-Mono-Monimatch," QST, Mar., 1965.  Although called a Monimatch in the title, the design actually is more similar to Bruene's directional coupler.

Bold, Gary, ZL1AN, "The Bruene Directional Coupler and Transmission Lines," PDF. This PDF gives an excellent explanation of the Bruene Coupler.

Kiciak, Paul, N2PK, "An HF In-Line Return Loss and Power Meter," PDF.  Constructions details of a power meter using a Bruene Coupler.  Contains an explanation by the other of why he prefers the Bruene coupler of the Tandem-Match Coupler.  Also interesting because the author separates the voltage-sampler from the current sampler and uses the differential inputs of an AD8307 to do the required addition (or subtraction) to get FWD and REF voltages.

(This web page could be useful for understanding the sampling method used in the N2PK meter:  http://www.g3ynh.info/zdocs/bridges/magdiff/part1.html )

Lewallen, Roy, W7EL, "A Simple and Accurate QRP Directional Wattmeter," QST, Feb, 1990, PDF.  Interesting variant of the Bruene coupler.  Roy uses two transformers for the voltage sample in lieu of capacitor voltage dividers.

White, Ian, G3SEK, "Inside a Directional Wattmeter," RadCom, Sept., 2002, PDF.  Discussion and a bit of analysis of Bruene coupler.  Includes Bruene's phase-relationship diagrams.

http://www.g3ynh.info/zdocs/bridges/reflectom/part1.html  interesting analysis

http://www.g3ynh.info/circuits/Diode_det.pdf  Diode detectors -- includes some info on shunt detectors, which is what Bruene's design uses.


Other references of generally interest:

http://www.g3ynh.info/zdocs/bridges/Xformers/part_1.html  great discussion on current-transformers for directional coupler applications

http://www.g3ynh.info/zdocs/bridges/Xformers/part_2.html Part 2 of current-transformers

http://www.g3ynh.info/zdocs/bridges/Xformers/part_3.html  And part 3, the last part, of current-transformers

http://www.g3ynh.info/zdocs/bridges/index.html  Indexes numerous topics.  Lots of great info to be found here!

http://www.richtek.com/assets/AppNote/AN008_EN/AN008_EN.jsp  Common-Mode choke model


Final Caveats:

As always, I might have made a mistake in my equations, assumptions, or interpretations.  If you see anything you believe to be in error, or if anything is confusing, please feel free to contact me.