On The Inside

On The Inside

Wednesday, May 20, 2015

On Filter Response

In a previous installment I showed a filter pulled from a CB radio. I've been battling with the filter impedance thing for some time now and can't seem to get a good method to reliably determine filter impedance. Here's my simple test set with two series pots and a couple of fifty ohm shunts.

So, you're supposed to adjust input and output for the best filter shape. Yeah, that's easier said than done.

I have found the process of trying to determine filter impedance empirically without the aid of expensive test equipment frustrating to say the least. This little diversion started because I wanted to measure the impedance of several filters that are in my junkbox, in particular, this nice little filter that came out of a Sears Roadtalker 40.

The way that this is supposed to work is that you put the filter into the jig, attach the input side to your sweep generator or noise source, the output side goes to your scope or spectrum analyzer, and you adjust the input and output knobs alternately until the ripple in the passband is minimized.

Great, that sounds easy-peasy!

Well, the devil is in the details. First, if we use the noise method, at least with the gear that I have, it is challenging to see the passband ripple in the noise, Ok, no problem, I'll use the sweep method. I have a couple of different sweepable generators, one of them should get the job done, right?

Generator #1

It's a BK-Precision 4040A. I'll say this, choosing the word "precision" as a part of your brand name is a stroke of marketing genius. Unfortunately, I don't think that the engineering department got the memo.  It's not useless, but I'm not really sure for which application the sweep function is intended. There is no sweep output so you cannot use the sweep generator to drive the X input of your scope in XY mode. If you do use the sweep function, it's very touchy and extremely wide, too wide, in fact, for adjusting a crystal filter. It does have a voltage control input which would be awesome if the start and end controls scaled an external ramp input, but  no, it's an unscaled input that is virtually useless without taking the time to put together a scaling jig.  The manual states that the input takes 0-10V ± 1V for a 100:1 frequency change. Great, plus or minus ten percent of the input range, that must be some of that precision.  Ok, enough bagging on the BK, mine was a craigslist find, it still gets some use. Speaking of bagging on the BK, you can order a handy tote! Ok, for real, enough bagging on the BK.

I also have an old HP signal generator, the ubiquitous and somewhat infamous 8601A. I say infamous because they are so well known for blowing their output amplifiers, which are made of unobtanium, that someone has re-engineered a replacement. Unfortunately for me, I think that 1) my output amplifier is blown, and 2) that the $90 cost is too high for me to justify repairing this old girl.

This lovely example is much cleaner looking than mine and also has the RF and aux outs routed to the front like any sane person would expect. Ok, I can see the value of having the cables out of the way for fixed applications, but that wasn't going to work for me.

First things first, a bit of testing. Sweep output hooked up to the X input, scope in XY mode, blanking out put hooked up to the blanking input, check!  I forgot to snap a picture, but, the blanking is reversed, the retrace is bright, and the sweep is dim.

I wired up a quick inverter using a 741 and connected it in line with the blanking signal.

This gave me a relatively nice looking sweep signal and I could get good blanking just by tweaking the intensity control of the scope. 

It's not quite that bright in reality, I used a long exposure on my camera to get a good trace picture. However, you can clearly see that the retrace (the falling edge of the ramp) is much dimmer than the sweep trace. The flat portions at the ends of the sweep cause the sweep to be fixed for a period of time at the start and end of each sweep. This makes it easier to setup the scope and signal generator.

Although the op amp worked well, there is no bipolar power supply that's easy to use in the 8601. I could make a simple zener supply from the +26 and -6 volt supply rails, but there isn't a lot of room in this thing so I decided to simplify the design. The op amp does a proper inversion around the ground point, but strictly speaking, that wasn't necessary, I just needed a change in the polarity of the blanking signal. So I replaced the op amp with a simple inverter made from a PNP transistor so that I could power it directly with the negative supply rail referenced to ground.  While I was at it, I put the generator on the repair bench to move the connectors around to the front. That way I could use the holes in the back for a new blanking output and a switch to disable it when I didn't want to use the sweep generator with my scope. It's a lot less hassle to reach back and flip a switch than it is to move both the scope and the generator just to connect a cable.

Here's the little inverter, built on a bit of perfboard, ready to be tucked into  the back.  The green wire is the inverted blanking output and the other three wireds are input, ground, and negative power. On the right you can see it tuked into the top of the 8601; it's the yellow blob just to the left of the transformer.

The original outputs on the back came with a nice label plate that just flipped over and added my own labels.

 Here's the output of the ramp, the original blanking signal, and the inverted blanking signal. You can see that the blanking signal is now inverted.  Yes, I should have kept the PNP in its linear region so that I preserved the nice ramps on the blanking signal, but it's not that critical and I was able to just throw the switch together with resistors that were just lying on the bench, literally. I have no idea what's causing the small spurs, I thought that it might have been my circuit, so I took it out, but they didn't go away. In any case, it's a blanking signal, it's hardly critical.

With blanking fixed and the jacks moved around to the front of the generator, I installed it back onto my bench ready to do some more testing. The output of the generator isn't quite according to hoyle. It looks reasonably clean on the scope, but, it appears to be low and the highest output position doesn't increase the signal at all. I suspect, as I stated earlier, my output amp has problems, but, for now, it's feeding a fairly clean and adjustable signal to the output, so, I'll proceed anyway.

My first attempt used my Boonton microwattmeter as a detector. It's old and crusty, but it works. I have no idea how accurate it is, however, as I don't have anything that is precise enough and reliable enough with which to check its calibration.  In any case, it's too slow to be useful in this context. To get a reasonable trace you have to run the sweep speed faster than the Boonton can reasonably follow.

I switched gears and grabbed an RF probe that I hacked together some time back for some reason. This is not a work of art by any means and was constructed without much thought, somewhat blindly, from an old handbook.  
The resistor tacked onto the 1/2 watt was not in the original build. After reading more detail I learned that the value of the resistor was based on the 10 megaohm input impedance of most VTVMs of the day. My scope has a one megaohm input impedance so I tacked a resistor in parallel to create the approximately correct voltage divider circuit. It helped, but I also read that these are really designed for a relatively low impedance drive circuit and I suspected that it was loading down the filter and my scope was already at its highest sensitivity. Further, it did not give a log output which would give me a better picture of the filter's response.  Finally, I never liked the housing choice for a probe, it was made to be quick and I would prefer to have an RF probe in a more probe like enclosure. So, I removed the simple circuit. Upon doing so I found a cold solder connection from the diode to ground, that might have been the cause of some of the nose, no matter, I was sticking to my plan.

I built a log detector using an MC13136, which is a receiver on a chip. Yes, it's a bit of a waste, but it's a through hole part that I have a few of which meant that I could whip it up quickly on proto-board.

The TO-92 part to the right is an LM78L05 which provides a regulated five volts for the log amp. To the right you can see a 1N4004 in series with the positive power terminal. This serves to drop the input voltage a small amount and to provide reverse polarity protection, something that might happen with such a casual power input. The two resistors at the top right are 100 ohm resistors in parallel, so, yes, this has a 50 ohm input impedance, but, it also has a wide range.

Does this yield an improvement? Yes, it does, but, it's still fiddly. At least now I can see something of a filter shape and adjusting the input and output pots does not completely kill my signal. It is not really effective at coming close to the impedance of higher impedance filters. I have one with a known impedance of 3.9k and as best as I can get with this setup is, larger than a few hundred ohms. 

My best result was with a Yaesu filter believed to have a 200 ohm impedance, my adjustment showed about 150 ohms. As for the filter that is the subject of this post, it remains a source of frustration.

What I've concluded however, is that old analaog equipment with voltage controlled sweep is not ideal for this kind of measurement. The controls themselves must interact; to get a wider image on the scope you either have to adjust the sweep, or, the X gain of the scope, either choice moves the bandpass in the image on the scope.  The precision required here is really only in two dimensions; first, you need precision in frequency, and second, you need precision in measurement of the resistors. The latter is independent of the test fixture, while the former is much more easily controlled in the digital domain.

The goal is to be able to repeat these measurements easily;  in the next installation I will move forward with a DIY simple scalar network analyzer based on the AD9850.