Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com

Updated: 19 April 2008. Initial version
20 April 2008. Added new sections on signals out of K2 IF

Table of Contents:

To_purchase_a_Softrock_Lite_6.2_suitable_to_use_as_a_K2_IF_panadapter
To_purchase_a_Z10000-K2_buffer_amplifier
Introduction
Construction
Schematic
Soundcard_and_Software
Signal_Leakage_from_the_Softrock
Softrock_4915_KHz_Basic_Sensitivity_
Setting_PowerSDR_Parameters
K2_Frequency_Shift_with_Mode and Band Inversion
3rd_Order_Intermodulation_Perfornance_of_Softrock_Lite_6.2
Single_Tone_Overload_
Signals_Out_of_the_K2s_IF_Port_
How_much_isolation_does_the_Z10000-K2_provide_when_installed_in_a_K2

This page provides connection and performance information on using a SoftRock Lite 6.2 as a panadapter for an Elecraft K2 transceiver.

Other pages that are necessary to understand this page are:
Z10000 Buffer Amplifier
Softrock Lite 6.2

To purchase a Softrock Lite 6.2 suitable to use as a K2 IF panadapter:

I've been asked how one acquires a Softrock Lite kit to be used as a panadapter for the K2 or K3. Tony Parks, KB9YIG, "Mr. Softrock" replied to my query as follows:

To purchase a Z10000-K2 buffer amplifier:

Please see my Z10000 buffer amplifier page. The best deal for a K2 is the K2 complete kit, as it includes the Z10000  buffer amplifier, frequency selective parts for the K2 and the internal installation parts to mount the amplifier inside a K2 transceiver.
 

Based on my experience last summer with 30 meter and 40 meter Softrock Lite 6.2 receivers, I knew it is highly desirable to provide audio and antenna isolation and to mount the Softrock in a shielded enclosure. Based on measurements of several transformers made at the same time, Triad SP70  transformers were well suited for this purpose, although at $17 each are on the expensive side. Still, the SP70 provided good frequency response over the range 300 Hz - 100 KHz and is the transformer of choice for my design.

The data is taken with an E-MU 0202 sound card as in the prior work.

The die cast enclosure I used is from Marlin P. Jones & Assoc. http://www.mpja.com/ part number 16286 BX, and is 4.5" x 3.5" x 2.2." Similar boxes are available from Hammond and other suppliers.

In order to float the antenna input, I used a plastic housed BNC connector. The two audio output jacks are floated with a piece of Delrin sheet. (The ground side of the  two conductor shielded cable is connected or lifted if necessary at the EMU-0202 sound card.
 

When I built the Softrock, I placed standard header pins at the input/output pads. This allows me to substitute other Softrock receivers quickly and to work on the units out of the enclosure. One other change I made was to replace the Softrock's 0.1uF output coupling capacitors with 1 uF Tantalum capacitors to improve low frequency  response. This is mostly helpful if you add 600 ohm  terminating resistors to the transformer secondaries.

The photo below shows the general layout of the package. The white BNC connector is intended for PCB mounting and has a plastic body so that the shell is not connected to the enclosure. A short run of RG-178 miniature Teflon coaxial cable connects the antenna to the Softrock receiver.

To reduce stray coupling, the audio wires are twisted. Also, rather than run a common ground from the Softrock module to the transformers, each transformer has a separate ground, with the common point being the Softrock's connector.

Incoming +12V is fed through a 0.01uF feedthrough capacitor and a ferrite bead (multi-aperture core).
 

Schematic

The schematic below shows the current construction. Points for experimentation include terminating the two isolation transformers with 600 ohm resistors and also comparing grounding the shell at the die cast box end as well as at the sound card end.

The E-MU 0202 has balanced input ports, which, when combined with the isolation transformers, does a great deal to reduce ground loops and noise. The audio cables I made up use shielded 3-circuit plugs at both ends and are wired with balanced audio pairs with a shield. The shield is connected at both ends of the cable so that various grounding arrangements can be experimented with by making changes at the Softrock or by operating the "ground lift" slide switches on the E-MU 0202 sound card.

It is, of course, important to keep the audio phase correct. Hence the transformer pin numbers should be followed, or, if changed, the connections should be made so as to preserve the relative phase of the two channels.

Not all sound cards accept a balanced input, in which case this arrangement cannot be used without revision.

Based on last year's tests, I stayed with the E- MU 0202 sound card. It's a 24-bit card that samples as high as 192 ks/sec, and connects with a USB interface.

Larry, N8LP, also uses an E-MU 0202 sound card and provides the following comments concerning it:

Althog the E-MU is the only acceptable card I have found so far for 192 kHz, it's a resource hog. I have an M-Audio Firewire card that uses 1/3 the resources (at 96kHz) than the E-MU 0202, with virtually identical noise and DR. PCI cards also seem to be better. USB is less efficient than either Firewire or PCI, as they use hardware for tasks that USB handles through interrupts and processor calls. The E-MU 0404 PCI card would probably be very good as well.

My shack PC is 2.8G P4 with 1GB RAM, and with just PowerSDR-IF Stage and LP-Bridge (my app that links to Scott, WU2X's PowerSDR-IF Stage) running, I see about 35-40% CPU usage. With the M-Audio Firewire Audiophile (recommended)  or PCI Audiophile 192 (not recommended) cards, the CPU load drops to 8-9%. I think you'll find that some of the load you attributed to the programs or graphics is actually due to the E-MU 0202's USB links.

The E-MU 0404 PCI card isn't expensive at $85; it's the E-MU 0404 USB card that is. I've had no reports on the the PCI version, but I think Bill, W4ZV might try one.  I may try it too, but you have to be very careful about  sound card specs. They almost never tell you the response or noise levels out to 90 kHz for the 192 kHz cards, and many don't actually sample at 192 kHz in "record", just playback. Some also filter the input down to 40 kHz, and even the ones that are truly wideband are noisier at the extremes... some as much as 25dB. They also play games with names, like car companies, so similar names don't mean similar designs.

[At his  request, I've edited Larry's comments for grammar and to remove some duplication.]

I ran the tests with a Dell Dell Precision M6300 laptop. This machine has an Intel Core 2 Duo T9300 processor running at 2.50GHz and 6M L2 Cache, Dual Core. The main memory is 4 GB of 633 MHz RAM.

I tried all the available compatible software in last year's tests. I installed the most  recent versions of Rocky, PowerSDR, M0KGK's software and the two programs by I2PHD. Although all the programs have improved to some degree, in general I found concerns and problems not much different than when I looked at them last year.

Rocky could be the best program available if it would accept 192 Kb/s sampling. At the moment, the only program I consider acceptable is PowerSDR. It works with 192 Kb/s sampling and  the E-MU 0202 sound card. I experience a periodic sound dropout when listening to the demodulated output, but I assume this is an artifact associated with my particular installation.
 

A major problem, inherent in its design, is that  the Softrock's local oscillator leaks out of its antenna port. Since the oscillator is a square wave, odd order harmonics are also present, in addition to other switching noise from the receiver's high speed digital dividers. The consequence of this leakage is that I consider it inadvisable to connect a Softrock to a receiver's IF circuitry without a buffer amplifier presenting high reverse isolation—it's not good design practice to inject extraneous signals into a receiver's IF chain. The Z10000 amplifier I've designed does exactly that, with > 80 dB reverse isolation.

The spectrum analyzer image below, made with an Advantest R3463 instrument, shows the output signals present on the Softrock's antenna port over the range 0...100 MHz.
 

The Softrock Lite 6.2 receiver, in its normal operating mode, is reasonably sensitive and is certainly adequate for either casual operating or to be used as a panadapter. However, the K2 Softrock version operates a bit differently, at less sensitivity.

In the normal mode, the Softrock uses a crystal oscillator at either 4X or 8X the desired center frequency. In the K2 case, the center frequency is offset from the receiver's IF frequency to help reduce the leakage problem, and the crystal supplied results in a center frequency around 4898 KHz, about 17 KHz from the K2's 4915 KHz nominal IF frequency.

Hence, we would expect the K2 Softrock version to use a 19.592 MHz or a 39.184 MHz crystal for divide factors of 4 or 8 respectively.

Unfortunately, these are not off-the-shelf crystal frequencies, and as a cost savings measure, the K2 Softrock version uses a stock 13.0625 MHz crystal and "harmonic sampling." I'm not going to go through the details, but a sampling detector such as used in the Softrock responds to its fundamental switching frequency and all odd harmonics. The odd order harmonics are reduced in sensitivity (in dB) by 20 log(1/N) where N is the harmonic order. Hence the 3rd harmonic response will be 20 log(1/3) or 9.5 dB down from the fundamental frequency response in the best case. In practice another dB or two loss is usually observed.

Will this reduced sensitivity be a problem?

The image below shows the K2 version Softrock's response to a -100 dBm level signal at 4920.5 KHz from an HP 8657A signal generator input to the antenna port. The signal is about one division to the right of the red vertical line. The pip about a half division to the left of the red line is the 0 Hz response hum and noise. (If you compare the hum and noise with this setup to that I ran into last year, you will see a major improvement. The SP70 transformers combined with better shielding and filtering make a big difference.) I calibrated the PowerSDR software's gain, so the noise level and signal level readings are accurate.

This image suggests that near the center response, the minimum discernible signal will be around -100 to -105 dBm. (Averaging is enabled in this graph which also improves the signal to noise ratio.)
 

Connecting the K2 version Softrock to a K2 through an internally installed Z10000 makes a significant difference in the Softrock's sensitivity, however.

The image below shows a -110 dBm signal (10 dB weaker than in the plot above) at 7 MHz into a K2 receiver with the preamp on. The Z10000 buffer amplifier has its gain set to be net 0 dB, i.e., 1 uV into the antenna input port at, say, 7 MHz, provides 1 uV into a 50 ohm load at the Z10000's IF sample output.

In the image below, the -110 dBm signal is the pip in the green area. The stronger pip to the right is the K2's 4915 KHz BFO leakage. (To read more about BFO leakage in the K2 read my page http://www.cliftonlaboratories.com/k2_interface.htm)

I've also re-calibrated the PowerDSR amplitude to be accurate, i.e., the left scale accurately reads in dBm at the K2's antenna jack, with the K2's preamp engaged. Note that the noise level near the center is -130 dBm, and the -110 dBm signal is thus about 20 dB above the noise floor. (Averaging is on in this image.)
 

My sense is that the gain is about right; the inherent noise in the panadapter display makes it difficult to detect signals at the noise floor anyway.

But, it's easy to provide additional  gain in the Z10000 buffer amplifier by changing R907 to a smaller value. The Z10000 instructions provide information on gain versus R907 value. If R907 is reduced from the recommended 2.2K increased  gain will result. Based on these measurements, however, at most I would look for an additional 6 db gain. Starting points for R907 are between 220 and 150 ohms. Before changing R907, however, I would use the stock value for a few days to see whether increased gain is really necessary.
 

I used the following PowerSDR parameters with my E-MU 0202 sound card. These parameters are dependent upon the sound card, of course and may not be the same for other cards.
 

The receive rejection phase and  gain, of course, must be adjusted for each specific Softrock and transformer. In addition, adjusting for best image rejection at one frequency will not provide nearly the same rejection even a few KHz away. At its best, I nulled the image for 80 dB suppression, but at 10 KHz difference the null was 60 dB and at 20 KHz difference it was 50 dB. This is a case where the multi-point automatic image nulling used in Rocky is far superior to the one-point calibration in PowerSDR.

Note center shows 0.000, with offsets from the tuned center frequency, plus and minus. The pip just to the right of the center red line is the K2's BFO leakage.

I've previously covered in detail how the K2's BFO and passband center shifts between different modes at http://www.cliftonlaboratories.com/faq.htm. In addition to shifting the BFO, on some bands the K2's IF spectrum is inverted, which is also discussed at the same page.

This means that if you set the PowerSDR software up so that signals above the frequency to which you are tuned appear to the right of center (the normal relationship in a panadapter) then on some bands this relationship will be reversed and frequencies that are higher will appear to the left of center. This is a consequence of how the K2 is designed and is explained in the FAQ page referenced above.  One possible solution would be to define two PowerSDR configurations, one for bands without inversion and one for bands with inversion.

While I had the test gear set up, I also measured the 3rd order intermodulation intercept of the Softrock Lite 6.2 K2 version receiver.

The image below shows the two tones at a level of -10 dBm out of the combiner. At this level, there are no 3rd order intermodulation products visible at the noise floor, which is 82 dB below the level of a single tone. This places the 3rd order intercept of the test setup > +31 dbm. (The other artifacts seen in the image are generated within the spectrum analyzer, as the guaranteed spurious free  range is 70 db.)
 

As the above PowerSDR plot illustrates, the Softrock's output has signals other than 3rd order (and the image frequencies, of course.) This lead to applying a single tone to the K2 version Softrock's input to see when the visible overload point was reached. (As explained in my earlier Softrock experiments, distortion arises when the output op-amp is driven into clipping.)

With -30 dBm input, the spectrum looks clean.

Signals Out of the K2's IF Port

In addition to signals out of the Softrock that may be coupled  back into the K2, in some cases we may be concerned with unwanted signals coupling out of the K2's IF tap that go back into the Softrock or whatever other panadapter is connected to the K2.

The Z10000-K2 has some frequency shaping in it, with both high and low pass filters. However, these filters are not complex enough  to completely suppress all unwanted signals resident in  the K2's signal chain.

The image below is taken with the K2 set for reception at 7 MHz. The output is from a Z10000-K2 buffer amplifier into an Advantest R3463 spectrum analyzer.

The strongest signal is at 12.1 MHz with a signal level of -72 dBm. The 4915 KHz signal at Marker 3 is the K2's BFO leakage.
 

To determine how much isolation a Z10000-K2 buffer amplifier provides, I attached an HP8657A signal generator to my K2's IF sample output port. With the 8657A set to approximately 4915 KHz and the K2 in CW mode, I increased the generator's level until I could hear a weak CW beat tone. At this point, the signal generator's output was -10 dBm.

This is quite  remarkable isolation, and at this level of suppression details such as coaxial cable routing, etc. become important. I don't, therefore, warranty that you will see the same level of isolation when installed in your K2, but rather regard this value as typical.