Part 2:  More IMDR Adventures

 

Jeff Smith  VE1ZAC   May 15, 2009

 

In part one, I took a stab at measuring the IMDR ( Inter modulation distortion dynamic range)  for my IC7700.  I used a pair of HP8640A generators that I have in my shop and followed the conventional methods of doing this measurement. The numbers I got were around 80 dB.  Several folks sent me notes on the effort, including Adam Farson and some others who alerted me to problems in using 8640’s for measurements like this. The problem is… these generators have oscillator phase noise that is no better than about 80 dBc/ 1 Hz. Since we need signals that are sizeable in the receivers passband to make IMDR measurements, there are problems with the phase noise in the side bands mixing in the pass band and with the receivers local oscillator noise bands and producing  IMDR products.

 

Is this important ? Yes, and no. On the yes side, if we truly want to make a good test that we can use to compare receiver numbers, then we need a test that is not limited by the measuring equipment. In this case I am really only interested in CW signals in close to our operating frequency, and 2 KHz seems to be the popular standard spacing for this. We put two strong signals in our passband that are 2 KHz apart and then listen for a distortion signal that is formed at a point 2 KHz above the highest signal. The difference between the interfering signal strength and the MDS number is the IM dynamic range in dB. The problem is.. we need very low phase noise oscillators to make this test.  On the other hand, we should think about the reasons we do this test. The idea is to give some idea about the rigs ability to deal with strong adjacent signals in the receivers passband. These signals can be of two types. One type is from a strong but clean transmitter that typically has low phase noise when it gets to us. I won’t go into this here, but small amounts of phase noise at the transmitter end are often left behind during the propagation process. As a result, these kinds of signals are ‘Clean’.  There are also plenty of less than high quality signals that could be in our pass band as well. These ones are probably better emulated with oscillators with more phase noise for a real life approach to the test. It is perhaps of note to compare the two, which we will do later on.

 

To make a long story short, my first tests really demonstrated the limits of my workshop oscillators!  The best IMDR numbers I got were in the 80 to 85 dB range, and were clearly phase noise limited.

 

Better Oscillators

 

Several evenings of research provided steaming heaps of information on phase noise! I even dug out my 1974 engineering text “Communications Circuits: Analysis and Design” by Ken Clarke and Don Hess and studied up on issues affecting oscillator issues.  Some great material is available from Wenzel Associates web site. http://www.wenzel.com/ There is also a very good technical tutorial on phase noise at Oscillator phase noise : A tutorial

These folks are experts in low noise oscillators. I also perused my now very dog eared copy of EMRFD by Hayward and Campbell. It was obvious I was destined to build a pair of oscillators for the job. My goals were to use as much of my ample junk box parts as possible, and have a noise figure that was at least 20 dB better than my HP 8640’s. Although, one problem that was apparent from the start was my lack of ability to measure the phase noise. I punted that problem till later. Also, since these oscillators have to be stable and 2 kHz apart, crystal resonators were essential. Better start with that issue. I have a bizarre hodge podge of crystals. That is something that I never throw away. And, after rummaging through the stash for a while I found two crystals that would work. One was 8.9985 MHz and the other was 9.0015 MHz. Their planet of origin was unknown to me, but since the 7700 can be tuned anywhere in the HF spectrum, these would be good target frequencies. ( Actually they were likely recovered from some old transceiver that used these as USB and LSB BFO frequencies around a 9 MHz IF.. that’s my best guess)

 

I narrowed down my oscillator circuit to either the recommended EMFRD one or a simpler one I found at Wenzel’s web site.  Here is a link to it :  Low Distortion Crystal Oscillator This one filled the bill for construction, but no actual numbers for phase noise were provided. However, the source pedigree was excellent, so I decided to give it a shot.

 

Construction

 

I have done some two dozen projects using Eagle Cad ( free version) and my little homebrew CNC Sherline conversion mill. As a result, this one turned into a one evening project.

Here is the EagleCad schematic and the board graphics. I used a one sided board layout but used a double sided PC board to establish a nice ground plane. You can remove some of the copper around component holes with a bigger drill bit to prevent shorts. An effective technique.

 

I used a J309 NFET for the Q1 and NPO capacitors for C1 and C2, since I found some in my capacitor collection. J1 is a jumper for a ferrite bead. L4 is  1.2 uH and the TWEAKER is the primary of a  RF transformer I found with 3 uH primaries. This provides a ready method of adjusting the frequency a little. In practice, it was easy to get these guys to be 2 KHz apart.

Here is the layout graphic: ( for one oscillator.. there are two built onto a common PC board)

 

Testing

 

It fired up first time !  That doesn’t happen all that often for me, so it was pretty rewarding. Nice big signals in the -5 dBm range too. Just what is needed. I elected to go with 9 volt batteries to make it as free of power line noise as possible, and very portable. Each oscillator only draws 5 mA or so. Batteries will be fine for testing receivers.

 

So, we are back to figuring out the noise. I had a look around the laboratory I consult with and discovered a couple of spectrum analyzers that should have the capability to make phase noise measurements. This lab doesn’t do much RF work.. mostly acoustic work. But I did find one nice RF job… an HP 8594A. This unit even has the noise measurement function built in to the marker menu.. a very convenient feature.  Here it is hooked up to one of my oscillators. You can see the noise function running on the screen. The attenuation in practice is about 20 dB so the peak is around 0 dBm.

 

Here is a photograph of the display. Note the noise marker reading and calculation of -117 dBm/ 1 Hz. This winds up actually translating into the noise spec of the spectrum analyzer of about -90 dBc/ 1 Hz.  Drat.. we are limited by the phase noise of the analyzer. But, we can say that our noise spec must be pretty good.

 

I went through this same procedure with two other analyzers. One had a noise spec of -108 dBc/ 1 Hz. In all cases, the measurement produced a number that was limited by the spectrum analyzer. So, it looks like the noise figure is at least as good as -108 dBc/ 1 Hz. If I can locate a better analyzer, I will get a reading on these things, but for now I am going to have to live with that upper range of the noise as the best I can do.

Here is the finished unit:

 

One oscillator has an output of -7 dBm and the other has +1 dBm. An attenuator on the strong output easily brings the two signals to the same amplitude for the IMDR test. The usual hybrid coupler and attenuators are used as in the Part 1 article.

 

Tests with <108 dBc/ 1 Hz oscillators:

MDS of -135 dBm and  IMDR of 98 to 100 dB. ( at 9 MHz, CW, 3 kHz roofing filter, 2 KHz spacing)

 

Previous tests with 80 dBc/ 1 Hz oscillators:

 

MDS -121 dBm and IMDR of 80 dB. ( at 14 MHz, CW, 3 kHz roofing filter, 2 KHz spacing)

 

Conclusions:

 

  1. These are tricky tests, and a host of calibration issues come up. So, let’s say our new IMDR tests are in the 95 to 100 dB range. This is very respectable performance, for clean signals.
  2. I still don’t know the actual noise figures for the oscillators, but I suspect it is beyond what we need for this test.
  3. “Dirty” signals will indeed cause more IM problems in our receiver passband.
  4. The Wenzel folks know their stuff !
  5. You can build low noise oscillators fairly easily.
  6. If you want really good IMDR numbers for dirty signals, you probably need narrower roofing filters and different IF scheme.
  7. While not the definitive answer for IMDR testing, these tests do point to the high quality of the IC7700 receiver.

 

I have heard that the ARRL latest test on the 7700 produced a number in the same range. That is encouraging.

 

VE1ZAC