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

E-mail: Jack.Smith@cliftonlaboratories.com
 

Shortcuts to sections of this page

Introduction
Building_the_kits
Initial_Problems_with_CPU_Loading_and_Zero_IF_Spike__&_EMU_0202_Soundcard
Triad_Audio_Transformer_Frequency_Response
Triad_MIL-T-27E-SP68_Audio_Transformer_Performance
Triad_SP-70_Transformer_Phase_and_Amplitude_Consistency
Bourns_LM-NP-1001-B1_600_Ohm_Audio_Transformer_
Radio_Shack_Audio_Isolation_Transformer_P/N_273-1374
Antenna_Isolation_Transformer
Rocky_Software
FlexRadio_PowerSDR_Software
M0KGK_DSP_Decoder_Software
Two_Software_Defined_Radio_Programs_from_I2PHD
Radiation_from_the_Softrock_Receiver
Third_Order_Intermodulation_Intercept_(IP3)_and_MDS_(minimum_discernable_signal)
Receiver_Input_Impedance_and_VSWR
Frequency_Stability
Conclusion

Revisions:
11 Feb 2010. Added measurements of phase and amplitude comparison of 4 Triad SP-70 transformers
02 November 2009. Added frequency stability section.
13 July 2008. Pete, N4ZR, provided a Radio Shack PN 273-1374 transformer which I've evaluated.
21 August 2007. Added information about using E-MU 0202 with Windows Vista.
19 August 2007. Added a note about E-MU 0202 drivers only available for Windows XP.
28 July 2007. Added measurement data for Bourns LM-NP-1001-B1 600Ω : 600Ω isolation transformer.
25 July 2007. Added measurement data for Triad MIL-T-27E-SP68 transformer
23 July 2007. Added details for broadband antenna isolation transformer.
22 July 2007. Added frequency response plot for Triad audio transformer.
11 July 2007. Quick comments on two software defined radio programs from I2PHD, Winrad and SDRadio appear below.
09 July 2007. Intermodulation performance analysis completed on Softrock Lite 6.2 30 meter band receiver. Details in the IP3 section below.
07 July 2007. I've installed M0KGK's SDR Decoder software. With my SX260, it's almost impossible to run due to CPU consumption. Details below.
06 July 2007. I've completely rewritten this page to reflect my experience with an E-MU 0202 sound card. My SX260 computer is still CPU-bound, but there's a world of improvement with this sound card compared with the built-in SoundMAX device.

The kits are not projects for beginners. The PCB is approximately 1.5" x 1.5" and in that 2.25 square inches per side are a lot of parts. Resistors are vertically mounted to save space and the two inductive components you must wind are on 0.25" cores. Parts are a mix of surface mount and through-hole mount and are on both sides of the board. The photos below show one of my completed boards.

I added the right angle headers to permit swapping boards and for easier experimenting. I also mounted the crystal in a socket for the same reason. If I were doing it again, I would use straight headers, not right angle headers.
 

The Builder's Notes do not provide "Heathkit type" step-by-step, "solder R1," instructions. Rather they are of the form "install the bottom surface mount 0.1 uF capacitors." This is not a problem for an experienced builder but may be inadequate if this is your first kit. The Builder's Notes could benefit from proofreading and editing, but these flaws are not critical.

I found the surface mount parts easier to install than the through hole parts, by the way. This is partly a matter of having the right tools and partly a matter of experience. If you have never built a kit with surface mount parts, you might wish to start with one of the inexpensive Norcal dummy loads. This Norcal kit has 46 surface mount parts (44 resistors, one capacitor and one diode) and costs only $7.50, so by the time you finish building it, you'll have plenty of surface mount practice.

I made two mistakes with the first receiver, installing one resistor lead in the wrong hole and not making a good solder joint on one IC.  The resistor body is indicated by a silk screened circle around the pad and the associated hole for the free lead is shown by a short line between the two pads. In several places the lines are not easy to see, and there are other pads spaced the same distance as the correct one. I wound up inserting the free end in the wrong hole, and caught the mistake when I found the pad already used when it came time to install the correct part. The soldering error was U5, the TLV2462 op-amp, pin 3 or 5. Although I examined the solder joints with a 10X stereo microscope, I did not find the bad connection until troubleshooting the receiver. Since I re-soldered both pins 3 and 5 at the same time, I don't know which one was bad.

The builder has to wind two toroidal cores. One is an inductor used in a low pass filter on the antenna input and the second is a three-winding transformer to connect the antenna to the mixer IC. In the two kits I built, both are wound on T25 (0.25 inch diameter) cores, with no. 30 AWG wire. (Receivers for some bands use larger cores.) The receivers I built required 36 (30 meter receiver, 36T on T25-6 core) and 38 turns (40 meter receiver, 38T on T25-2 core). This is tedious to wind on a small core and in fact significantly exceeds Micrometals' single layer specification of 24 turns for No. 30 wire on a T25 core. For 38 turns, Micrometals' reference book calls for No. 34 AWG wire, which allows up to 41 turns in a single layer. (Micrometals is the main manufacturer of powdered iron cores such as used in this kit.) I found not possible to wind the specified number of turns in a single layer, so some turns are in a second layer. Multi-layer toroidal coils have increased distributed capacitance compared with a single layer coil of identical inductance, which in this application would show up as degraded high frequency rejection in the low pass filter.

The transformer is easier to wind as it has fewer turns.
 

The 30 meter oscillator runs at 40.5 MHz (remember, it is divided by four to get two 90° phase shifted waveforms to drive the mixer in quadrature) , whilst the 40 meter oscillator runs at 28.2 MHz. The 2N3906 did not provide enough voltage swing to toggle the 74HC74 divider, so I replaced it with an MPSH81 transistor. The MPSH81 part has Ft = 600 MHz, while the 2N3906 Ft is 250 MHz. (The MPSH81 has a different pin-out, so it's not a direct swap—you have to exchange the base and emitter pins.) Sure enough, the MPSH81 improved drive to the 74HC74 divider and it started toggling as it should.

From comments on the Softrock Yahoo group, I've noted other builders have experienced the same problem, leading me to believe it is a design error, with some receivers working and others failing, depending on the parameters of the particular 2N3906 supplied.

My initial work with the 40 meter Softrock receiver and the Dell SX260 was troublesome in several aspects:

• Very high CPU resource loading, particularly with Rocky software
• Large "zero IF" spike at the center frequency with both Rocky and PowerSDR

Based on recommendations from Joe, K9HDE, I made several changes in my setup:

• Purchased a new sound card. Since the SX260 limits me to external USB cards, my research showed the E-MU 0202 to have decent specifications for noise and sample rate.
• Packaged the receiver in an enclosure with isolated antenna input and isolated audio output.
• Investigated and discovered why CPU resource consumption is a problem with the SX260 computer.

Since the E-MU 0202 sound card has optional balanced audio input, I set up the audio isolation transformers for balanced output.

The E-MU 0202 supports sampling speeds through 192 ks/s.

Note that the E-MU 0202 does not have drivers for Windows 2000, nor are Windows Vista drivers released yet. W2K will not be supported according to comments I've found.

There is an "official" work around to use E-MU 0202's Windows XP drivers with Vista at http://www.emu.com/support/files/storage/E-MU%20USB%20Interfaces%20on%20Vista.pdf.

Real Vista drivers for the E-MU 0202 are scheduled for 3Q 2007. http://www.emu.com/support/vista.asp
 

The Triad 5-58X 600Ω to 600Ω audio transformers I use to isolate the Softrock's audio output have quite decent audio response, with a 3 dB bandwidth from 158 Hz to 74.83 KHz when measured with 600Ω source and termination. Unfortunately, these transformers are no longer in production, I believe.

The data is taken with an HP3562A Dynamic Signal Analyzer, with tracking generator.

In the first plot, below, I've added a 560 ohm series resistor to the tracking generator's 50Ω output to bring it up to approximately 600 ohms. Likewise, a 620 ohm resistor is applied across the transformer's output, in shunt across the HP 3562A's high impedance input, so the transformer sees approximately 600Ω on both the source and termination.

This measurement protocol duplicates how a 600Ω to 600Ω audio transformer is normally used; both the driving source and destination receiver have 600Ω impedances. But, is this the case when used to isolate a Softrock receiver? And, if it isn't how does that alter the transformer's response?
 

Hence, we expect C16 and C17 to roll off low frequency response when a 600 Ω isolation transformer is used. C16 and C17's size is not as important when the Softrock works into a high impedance termination, such as directly into a sound card. In this case, the 0.1μF capacitor's reactance is small with respect to the sound card's input impedance, even below 1 KHz, and hence will not cause appreciable low frequency roll-off until we get well below 100 Hz.
 

If you are interested in how to analyze and model a real transformer, read Midcom's Tech Note 82—Tips for Transformer Modeling, written by Dave LeVasseur and available at http://www.midcom-inc.com/Tech/pdf/tn82.pdf. Dave's technical note explains, in simple language, leakage reactance, magnetization reactance and other nuts and bolts of how a transformer looks to the outside world. Of course, even more complex models of transformer behavior are possible, and necessary in certain circumstances, but you can go a long way with the simple model developed in Tech Note 82.

To simulate the Softrock, I ran a second frequency sweep with the same transformer, but fed with a 0.1μF series capacitor, with the 50:600 Ω matching resistor removed. Hence, the driving impedance in the following plot is 50 Ω in series with a 0.1μF capacitor. Although not the nearly zero ohms of the Softrock's op-amp stage, it's close enough for the purpose. I also removed the 620 Ω termination resistance, so the transformer's secondary is loaded only with the 3562A's 1MΩ input impedance, plus, of course, its input capacitance and the capacitance of a 3' length of coaxial cable test lead.

The resulting plot, below, certainly looks different than the first plot. As expected, the 0.1μF series capacitor rolls off the low frequency response quite a bit below about 700 Hz. But more objectionable is the resonance around 850 Hz where the 0.1μF resonates with the transformer's inductance causing a 10 dB response peak, up through three or four KHz. (If you read Tech Note 82, you should understand why this peak occurs.)

There's also a bit of resonance showing at the upper frequency end.

Whether your particular isolation transformer will have a similar resonance with the Softrock's output coupling capacitors depends on the transformer, and your sound card's input impedance. My E-MU 0202 card has either a 1.5KΩ or 1MΩ input impedance, depending on which amplifier is selected. Presumably, operating my E-MU 0202 with the 1MΩ amplifier risks a peaked frequency response, although the exact location and range of the peak will not necessarily match the plot below.

The low frequency roll-off is still present, of course, but the peak has vanished at both the high and low end.

For example, if the objective is simply to characterize your particular installation, I would use an RF signal generator with your normal operating software and see how the signal amplitude varies as you move in frequency. Or, if you use Rocky, the built-in amplitude versus frequency plot can give you an idea of the overall frequency response of your setup.

Using test equipment as I've done, lets me understand better why things behave the way they do, and, when combined with system measurements in situ, allows fixes or adjustments to be made on a logical basis.
 

Kees, K5BCQ, loaned me a Triad MIL-T-27E-SP68 audio transformer to measure, as a possible audio isolation transformer to use with a Softrock receiver. These transformers are available in a variety of impedances, identified by the SPxx suffix. The SP68  transformer supplied by Kees has a 10KΩ design impedance primary and dual 2.5KΩ secondaries, or 10KΩ : 10KΩ when the secondary windings are in series. Allied Radio stocks some MIL-T-27E parts, at $18.70 each.

The transformer has, when properly terminated, a good frequency response, up to 100 KHz, but you must ensure it is  terminated with the 10K design impedance to avoid a nasty high frequency resonance. Some sound cards will present the required 10KΩ impedance to the transformer, whilst others, such as the E-MU 0202 unit I use, would require auxiliary terminating resistors, as the card's input impedance is 1 MΩ. Driving this transformer from a low impedance source, such as the 50 Ω signal generator in the 3562A, or from a Soundcard's line out port, is not a problem and impedance matching is not required at that end of the circuit. And, it should not be matched at the source end via series resistance, for example, as the result will be increased end-to-end voltage loss.

First, the transformer's key parameters:

SPICE Model. The diagram below provides a model of the MIL-T-27E-SP68 transformer, in a representation of the HP3562A Dynamic Signal Analyzer test circuit (50 Ω generator output source; 1 MΩ input impedance).
 

All data is taken over the range 10 Hz - 100 KHz, log horizontal axis. The vertical axis varies from plot to plot, with the specific value shown on the plot. (The plot below is 2.0 dB/division.)

50Ω generator source / 1 MΩ termination.

Note the peaking  response, with about a 10 dB peak at 100 KHz. This is due to a combination of the transformer's distributed capacitance, stray capacitance in the test setup and the 3562A's input and source capacitance. The frequency and magnitude of the resonance will differ from setup to setup, as can be seen below. This transformer performs best when terminated with the 10KΩ design value.

Terminating the transformer with its 10 KΩ design value knocks the resonance peak down to negligible levels. At 100 KHz, under these test conditions, the response is down about 1 dB from mid-band values.

Terminating the transformer with its 10 KΩ design value for both source and load, should produce the response the designer intended. Indeed, the measured response easily meets the specified ±2 dB from 300 Hz to 100 KHz.  The measured data shows the ±2 dB range as about 100 Hz to > 100 KHz.

To demonstrate the usefulness of SPICE modeling, the plot below shows the predicted response for the transformer under the test condition 50 Ω source / 1 MΩ termination, as in the first measured plot shown above. The simulation shows generally good agreement with measured data, except that the actual low frequency response is a bit better than predicted by the model.

In February 2010, a controversy flared up on the Softrock mailing list over the bad effects of isolating audio transformers. Claims were made that transformers have too much phase shift and amplitude shift.

Among the best transformers for the purpose of breaking  ground loops, i.e., isolation, of the Softrock are Triad's SP-70 parts. These are the 600:600 ohm versions of the SP-68 parts examined above. I have provided distortion and frequency response measurements for the SP-70 at Non-Linear Transformer Behavior.

I have four SP-70 transformers in my junk box and measured the phase and amplitude response of all four. The test setup is an HP 3562A Dynamic Signal Analyzer driving the transformer under test with its internal signal source, set at 5 mV PP, over the frequency range 100 Hz - 100 KHz. This low signal level is typical of the output of a Softrock when exposed to normal radio signal levels. The 3562A is run with leveling enabled to hold a constant drive level. The source drive impedance is 50 ohms, and the SP-70 was terminated with a 10K 5% carbon film resistor. This simulates a typical sound card impedance. The 3562A has a 1 Meg input impedance which is negligible compared with the 10K load resistor. Both the 3562A's reference channel (used for leveling) and the transformer under test are connected to a Mini-Circuits resistive power splitter, model ZFRSC-2050, usable from DC to 2 GHz, through identical length RG-223 double shielded coaxial cables.

Before measuring the phase shift associated with the SP-70 transformers, I established the residual error in the instrumentation, bypassing the transformer under test. The plot below demonstrates the typical phase error is within ±80 milli-degrees of 0 over almost the full range, with the residual increasing to around -120 milli-degrees at 100 KHz. Note the horizontal axis is log frequency.

Above 1 KHz, the total differential phase shift is on the order of 0.4 degrees. And above 2 KHz, the difference is, for all practical purposes, zero. (Vertical scale is 2.5 degrees per division.)

Although there is a  total phase shift between 1 kHz and 100 KHz of about 12 degrees, as I understand it, the current crop of software used with Softrock receivers is capable of correcting for overall phase shift via a curve fitting algorithm, originally (to the best of my knowledge) introduced in Rocky software.

The data shows that the differential amplitude above 200 Hz is very small and is essentially immeasurable above 400 Hz. Again, this degree of shift can easily be compensated for by the receiving software.

The LP-PAN uses a pair of SP-70 transformers to isolate the audio output, with hundreds of units in the field. My understanding is that there are no complaints of inadequate image rejection.

Crispino, I5XWW, has supplied me with some Bourns LM-NP-1001-B1 600Ω : 600Ω transformers to evaluate. Data sheet at http://www.bourns.com/PDFs/LMNPLP.pdf  My conclusion is that these transformers will work if you have the correct terminating impedance—not 600 ohms, but something closer to 10K. Their high sensitivity to terminating impedance, however, requires considerable experimentation with your particular sound card before providing acceptable bandwidth. And, depending on your sound card's input impedance, they may not be acceptable at all. Hence, I do recommend these transformers as "plug and play."

From the data sheet, the transformers in the LM-NP line are intended for interfacing to the public switched telephone network, or similar low-fidelity applications. The transformers are small, intended for PCB mount, although the leads are long enough to permit soldering wire extensions.

There are 10 transformers in this series, and the data presented is only for the LM-NP-1001-B1 model.
 

The vertical axis varies from plot to plot, with the specific value shown on the plot. (The plot below is 3.0 dB/division.) The horizontal axis shows the sweep range, with a log frequency basis.

600 Ω source and 600 Ω load

With this test condition, the specification sheet environment should be duplicated. The performance meets the data sheet specification of +0/-2.5 dB, 200 Hz to 10 KHz. In fact, the -2.5 dB point is around 15-16 KHz.

50 Ω & 0.1 μF Series Capacitor Source, 1600 Ω Termination

This combination shows roll off at both the low and high frequency end of the spectrum, showing a 3 dB response from about 700 Hz to 30 KHz.

Since the purpose of the analysis is to determine the suitability of the transformer to isolate the tip and ring audio outputs of a SoftRock receiver, I made measurements with the transformer connected to a SoftRock Lite 6.2 receiver for the 30-meter band (center frequency 10.125 MHz.). The test setup is shown in the block diagram below. The custom software steps the SG-100 function generator in frequency and records the audio output level. Resistive terminations were added on the transformer output for the various test conditions studied. The SG-100 is set at an output level of 10 mV PP.
 

The first test swept the receiver over the range ±125 KHz from center, with a direct connection from the receiver's tip and ring outputs to the HP3456A voltmeter. The purpose of this test is to obtain a baseline performance picture of the receiver, operating without a transformer.

As the data shows, the receiver's 3 dB bandwidth is approximately 100 KHz on either side of the center frequency. There is some asymmetry in the response, and difference in response level between the tip and ring audio channels. These imperfections would be calibrated out in the associated software in normal use. The spike/dip at zero offset corresponds to the normal "DC" spike and associated 60/120 Hz noise around zero offset.

The plot below shows the ring audio output when coupled through an LM-NP-1001-B1 transformer, with various terminating resistances.

For reference purposes, the plot also depicts the receiver's response with no transformer.

The  data shows the transformer's response beyond 50 KHz or so is highly sensitive to terminating impedance, with relatively good performance provided only around 10 KΩ. The data taken at 1 MΩ, which is the impedance of my E-MU 0202 sound card, shows severe resonance effects around 110 KHz from center. The lower values of terminating resistors dampen the resonance, but result in high frequency attenuation. Only at around 10 KΩ is the response reasonably flat.

The data also shows that for a restricted frequency range, up to 50 KHz or so, the transformer is much less sensitive to termination impedance, so long as it exceeds about 5 KΩ.

The final plot shows the response with the LM-NP-1001-B1 transformer,  terminated with 1500 ohms, expanded to ±5 KHz from center. The data clearly shows the low frequency rolloff caused by the transformer and the SoftRock's series 0.1 μF blocking capacitor. Fortunately, less than 1 KHz is severely (more than 10 dB) attenuated near zero Hz.

If your requirement is only 24 KHz bandwidth, then the terminating impedance is far less critical and almost any reasonable value > 5.6 KΩ will work. Even with 48 KHz bandwidth, the terminating impedance is not overly sensitive. However, to obtain acceptable results over a 96 KHz bandwidth, the terminating impedance must be carefully selected.

If the sound card's impedance is higher than 10 KΩ, then adding a parallel resistor of suitable value should provide the necessary 10 KΩ impedance. For example, the E-MU 0202 sound card, when in high impedance mode, has an input impedance of 1 MΩ. For this card, in this mode, a 10 KΩ resistor on the LM-NP-1001-B1's output winding should prove satisifactory. On the other hand, if the E-MU 0202 card is operated in low Z mode, the input impedance is 1500 Ω. In order to let the LM-NP-1001-B1 transformer "see" 10 KΩ in this case, a series resistor of 8.5 KΩ is required. Although providing the desired terminating impedance to the LM-NP-1001-B1 transformer, the voltage divider effect will reduce the audio into the E-MU 0202 sound card by about16 dB, a major loss in sensitivity, which renders the LM-NP-1001-B1 transformer unacceptable in this case.

Whether the LM-NP-1001-B1 transformer will work for you depends on the sound card impedance. If greater than 10 KΩ, it should be relatively easy to make it work. If much less than 10 KΩ, the extra loss resulting from resistive matching likely will produce poor results. I can recommend the LM-NP-1001-B1 transformer only if you are comfortable making performance measurements verifying its operation, and, of course, if you have the necessary test equipment. You should be able to get a good view of the transformer's performance with the band scope feature in the standard software, plus either an RF signal generator or a broadband noise generator.

Pete, N4ZR, supplied a Radio Shack 600:600 ohm audio isolation transformer to evaluate its suitability to isolate a Softrock receiver.

The transformer is small, about the size of a sugar cube, or, for that matter, the Triad SP-xx series of military spec audio transformers. It is encased within green heat shrink tubing, with the windings brought out through short (3" or 75 mm) wire leads. As is the norm with Radio Shack components, specifications are on the minimal side:

• Coupling
• 600-900 ohms
• 300Hz to 5kHz response
• 100 megohms insulation resistance at 250VDC

The frequency response specifications are not encouraging, but measured data shows the transformer is much better than these specifications suggest.

The 273-1374 transformer has a list price of $3.99 and should be available in most Radio Shack stores, or  via Radio Shack's web ordering service at www.radioshack.com.

Rather than run tests on the 273-1374  transformer as a stand-alone device, I decided to look at its performance with an updated version of the test setup used for some of the 2006 tests. The signal generator steps, in 1 KHz increments, from -125 KHz to +125 KHz from the Softrock's center frequency of 8192 KHz. The tip audio output is read by an Agilent 34410A digital voltmeter. Both the VP8191A and 34410A are  controlled via a GPIB bus with a Prologix controller.
 

Several points of interest emerge from the plot.

• The zero reference point is the direct connection at 8191 KHz, representing a 1 KHz offset audio tone. Because the transformers do not have exactly a 1:1 winding ratio, the output of the two transformers is slightly greater than when directly connected to the voltmeter. This difference is negligible for the intended purpose.
• All connection methods have artifacts around zero, discussed in more detail later.
• Although the SP70 transformer is better performing than the 273-1374, the difference is modest at best, amounting to about 0.5 dB at 125 KHz offset.
• The Softrock receiver has a 3 dB bandwidth of ±100 KHz. This results from the low pass filter implemented in the op-amp output stage.

We note, moreover, some artifacts in the direct connection as well. This could result from hum and noise on the VP8191A signal generator.

The peaked response for both transformer connection likely is a product of series resonance between the transformer's winding inductance and the series blocking capacitor in the Softrock. (I've modified this particular Softrock receiver by replacing the stock 0.1 µF output capacitor with 1.0 µF units for better low frequency response.) The resonance is negligible more than 500 Hz from center.
 

Considering the 273-1372's modest cost, it's worth considering as an isolation transformer.

I found it also moderately helpful to double-isolate the antenna from the Softrock receiver with an RF isolation transformer. The Softrock's design provides isolation as the input transformer's primary floats with respect to ground. Hence the antenna isolation transformer described here provides a second isolation stage. I can't say it made a huge difference, but it may have helped a bit. I made multiple changes in one test (not a good idea but I didn't want to spend all day tinkering), with one of the changes being the antenna transformer. I saw improvement, but can't accurately apportion the benefit to the various simultaneous changes.

An isolation transformer can take many forms, and the one I used is based on the parts I happen to have at hand. Still, it works reasonably well, and should be easily duplicated if desired.

The transformer consists of 6 bifilar turns wound on a 61 material binocular core, Fair-Rite part number 2861000202. (Fair-Rite calls these "multi-aperture cores.")  This core is available from many sources, including Ocean States Electronics, part number BN61-202. http://www.oselectronics.com/. As of late July 2007, the core costs $0.65 from Ocean State.

To wind the transformer, cut two lengths of wire, 11.5 inches long. I used #30 AWG "wire wrap" wire, one length with red insulation and one with blue, to help distinguish the two windings. You can use almost any wire that will physically fit. No. 28 magnet wire would be a good substitute.

After cutting the two wires, twist them together, at about two twists per inch. This is not a critical value, so don't worry if it's not exactly 2 twists per inch.

Wind the twisted wire through the core, six turns total. Remember—one turn requires the wire to go through both holes. Leave about 1" excess wire at the start. When finished winding, trim any excess wire to the length needed in your installation.

If you have test equipment, each winding should measure approximately 12.5 μH at 2.5 MHz. (Q measured at 220, HP 4342A Q-meter.)

The plot below shows the transformer's return loss when terminated with a precision 50 ohm load. It is less than 10 dB over the range 1.8 MHz - 29.7 MHz.
"

Finally, we look at the transformer's insertion loss. Over the range 1.8 - 29.7 MHz, it is less than 0.5 dB, and it's closer to 0.25 dB between 3.5 MHz - 10 MHz.
 

The Softrock is a direct conversion receiver, but with I and Q outputs, which permits image rejection when used with the correct hardware or software audio combining.

Most Softrock users connect their receiver to a PC's soundcard and use one of several software programs available to provide standard receiver functions, such as tuning, filter bandwidth, spectral display and the like. I've started with Rocky, written by Alex, VE3NEA, and available without charge at http://www.dxatlas.com/rocky/. 

I like Rocky. It has, in my view, a near perfect balance of features and user interface. However, it has a few rough edges compared with PowerSDR.

I won't duplicate Alex's instructions, but will point out the main features. The software's spectrum analysis tuning mode is shown below. (There's also a great waterfall display option--see below.). The pip at the center is the receiver's DC output, and there is a band a couple of KHz on either side of it with increased noise as seen below. The 40 meter receiver center frequency is 7056 KHz, and with 48 KHz sampling rate, you can tune 24 KHz either side of center. (Since the receiver has I & Q samples, the sample rate equals the bandwidth.) The image below shows two SSB signals above the center and several CW signals below center.

To tune, you can place the mouse on the display and click. Or you can use the keyboard's arrow keys to tune up or down, or if your mouse has a wheel, use it to tune.

The main Rocky screen is shown below. It's a panadpter view.
 

Waterfall image. I've reduced the size, but in the original you can easily read the Morse. The large waterfall is fast moving, and the smaller waterfall to the right is slower, to provide a better view of activity over the space of several minutes.

After Rocky ran for a while, I measured as much as 70 dB unwanted signal suppression. And because the phase and amplitude corrections are applied on a frequency-related basis, this level can be achieved across the full band.

I found two major issues with Rocky. Some (perhaps all) are unique to my particular combination of computer and sound card. Still, they are real for me.

CPU Resource consumption
Rocky has two major activities that are computationally intensive. First is the DSP-related code, filtering, detecting and the like. The second activity is graphic-related; to update the panadapter display many times a second is a graphically intensive process.

To make a long story short, Rocky's DSP code component runs relatively efficient on the SX260, consuming less than 10% of the CPU resources. However, the way Dell designed the SX260 makes graphics rather inefficient. Rather than install dedicated memory for the graphics processor, Dell instead reserves 32 MB of main memory for the graphic processor. This decision saves some money, but it is a major bottleneck when a program—such as Rocky—frequently updates the screen. When using the built-in sound card, running Rocky in anything other than a small window causes severe conflicts with other programs need for CPU cycles. The memory bottleneck problem is aggravated by running the SX260 in 1600 x 1200 resolution, its highest resolution mode.

As the data below shows, using an external USB card adds to the CPU requirements. (These figures are unlikely to apply to your computer if it has a separate graphics card.)
 

Rocky supports 96 Ks/s with some sound cards (Delta 44), but this mode does not work with my E-MU 0202, even when I manually set the card to match Rocky's setup parameter. I understand this is a known issue, in that not all 96 Ks/s cards are supported with Rocky.

The image below shows PowerSDR running on my SX-260 with the E-MU 0202 sound card, in 192 Ks/s mode.  At 192 Ks/s, PowerSDR displays (and will tune) over a 192 KHz range, ±96 KHz from the 7056 KHz center frequency.
 

For example, to adjust the I & Q channels for phase and level balance to null the image requires a two-step process. With a signal generator input, you run an automatic nulling process. Then, you manually fine tune the level and phase settings.
 

I may well be missing something in the calibration process, such as repeating the calibration every 10 KHz. As I say, the program and set up is far from simple and the documentation only discusses image calibration at a single frequency.
 

M0KGK's software looks interesting, but has serious compatibility problems with my Dell SX260.

Even resized to as small as is reasonably feasible, it still consumes a lot of CPU resources.

48 Ks/s is on the edge of being usable, so long as no other programs are running. Any program that runs in addition to M0KGK grabs too many CPU cycles, causing the audio to break up and tuning to become jerky. The two higher speeds are unworkable, even with no other programs running.

I don't believe this is  totally a graphics issue, as I've made the window as small as the size I use for Rocky and smaller than PowerSDR runs in. (The CPU data is for this small window size.) In the process of resizing KGKSDR to make the panadapter window disappear, I got into a cascading error message problem which required closing the program with Task Manager.

Like Rocky, KGKSDR does not change the E-MU 0202 card's sample rate. I have to manually adjust it using the E-MU 0202 control application.  On the plus side, KGKSDR supports sample rates  through 192 Ks/s.

KGKSDR's calibration window is shown below. Three display screens are used to see the state of calibration, amplitude, phase and number of samples versus frequency offset.
 

At Aldo's ( IW2DZX) suggestion, I tried two SDR programs from Alberto, I2PHD. Alberto has written many useful program for amateur radio, with Winrad and SDRadio his two SDR projects. The programs are available for download at http://www.winrad.org/ along with Alberto's other software.

Winrad's main screen is shown below. It has about every display known to man, all running simultaneously. A panadapter-type display, waterfall display and graphical displays of the receiver section bandwidth (you change bandwidth by grabbing the selectivity curve with the mouse and moving it) and a zoomed waterfall view of the signal within the receive section bandpass.

Winrad works with my E-MU 0202 sound card at 48, 86 and 192 ks/s.

• The screen is way too busy with graphic displays and consequently consumes a great deal of CPU resources from my Dell SX260. With the panadapter and waterfall display running in the slowest refresh mode, the total CPU resource consumption was 75%, with Winrad grabbing 65%.  I also found a discrepancy between CPU resources reported by Winrad's status indicator and those reported by Windows in Task Manager. For example, Winrad shows 48% CPU resources whilst Task Manager shows Winrad consuming 70% CPU resources.
• I would find Winrad easier to use if "clutter control" were available, so that unwanted displays could be disabled and hidden. This would simplify the display and also reduce graphics and CPU loading.
• I like the grab and move bandwidth adjustment, but there's no need to show it all the time.
• I & Q balance (amplitude and phase) requires manual adjustment. The adjustment is made only for a single frequency and consequently although I could achieve 80 dB+ image rejection at one frequency, this could not be maintained over the full frequency range.
• Support for ASIO drivers and the E-MU 0202 at all sample speeds is good and smooth.

Like Winrad, SDRadio has a single frequency manual I&Q balance setting.
 

I should add that both Winrad and SDRadio work well in terms of audio quality and the like. My issues relate mostly to user interface concerns. To a large extent this is a matter of personal preference and my desire for a clean, uncluttered display may be considered foolish by those who like to see all the possible options visible at all times.

The main spurious is at the Softrock's local oscillator frequency, 7.056 KHz. It's level is -39 dBm, a hefty signal. A second weaker spurious can be seen at 8.015 MHz.

The crystal oscillator at 28 MHZ also leaks through the antenna port, along with the second and third harmonic of the divide-by-four LO.

I've made IP3 and MDS measurements on the 10 MHz Softrock Lite 6.2 receiver. To avoid sound card issues, I've made the measurements with other test equipment here at Clifton Laboratories.

The signal generation part of the setup is conventional and I've used it to make many IP3 measurements. The signal generator, pad and hybrid combiner setup is capable of measuring IP3 figures in the +30 dBm or so range.

The Softrock's output feeds an HP 3562A Dynamic Signal Analyzer. A DSA is a combined 0-100 KHz spectrum analyzer and tracking generator, in addition to many other things. It's a digital box, with A/D converter and dedicated hardware for signal processing and the spectrum analysis portion is done via fast Fourier transform technology, similar to the way Rocky, PowerSDR and other programs generate their panadapter display screens.

The advantage of using the 3562A instead of a computer and sound card is that it avoids sound card and computer noise issues. By the standards of sound card and computer ratings, the 3562A seems distinctly under powered, but it has excellent noise performance, with a noise floor over the range 1 KHz - 100 KHz of -116 dBv, or 1.6 μV. That's for the entire band, and for a 1 KHz span within the band, the noise level is in the -130 dBV range, or about 0.3 uV. The instrument has a spurious (IMD and harmonics) range > 80 dB. Considering the 3562A is an early 1980's product, it specifications are even more impressive.
 

The image below shows the Softrock's output spectrum over the range 10 KHz - 50 KHz. The two signal generator tones at 10150 and 10160 KHz (shifted down to 25 and 35 KHz, respectively, by the receiver's 10125 KHz local oscillator) are centered. The stronger intermodulation product is -57.36 dB with respect to the two test signals.  This capture is with the single tone input to the receiver of -11 dBm, or two tone input of -5 dBm, safely below the op-amp clipping point.

The resulting IP3, with respect to the single tone input level is:

IP3 = -11 dBm + 57/36 dB/2 = +17.7 dBm.

Although I've shown one decimal place in the IP3 figure, I certainly don't imply the numbers are accurate to 0.1 dBm! I don't have a good way to quantify the error, but I suspect the IP3 measurement error in in the order of ±1 to2 dB.

To see whether  this is a problem with the Softrock receiver or my sound card / computer / power supply / etc., I looked at the Softrock's output over the range 0 - 1 KHz with the 3562A. As the figure below shows, there are five distinct spectral lines seen, 60 Hz and its harmonics through 300 Hz.

I'm seeing much more low frequency crud using the E-MU 0202 sound card than found when using the 3562A, even though I've added isolation transformers to the I and Q output, and an antenna isolation transformer. The 3562A data shows the level to which I should aspire, should I have sufficient time to devote to the matter.

Over the range 10100-10150 KHz, the Softrock's input return loss is around -17 dB, corresponding to a VSWR of 1.3. This value is certainly acceptable.

The data is taken with an HP 8752B Vector Network Analyzer.
 

I recently (as of early November 2009) used a 7 MHz Softrock Lite I receiver to look at noise and hum sidebands on a signal generator and found that with long averaging times on the audio spectrum analyzer connected to the Softrock's audio output, some evidence of frequency drift was present.

The amount of drift was small, on the order of 1 Hz over the space of 15 or 20 minutes, but in order to determine whether the drift originated in the signal generator or in the Softrock, I measured the Softrock's oscillator stability with the results below.

The Softrock board is in a small aluminum minibox, with the upper section removed so the board is open to random air currents in my basement shop. The data shows the effects of temperature variation in my basement as the furnace cycles off and on. With my particular Softrock, increasing  temperature causes a drop in oscillator  frequency, and as the plot shows, when the furnace runs, there's a rather fast drop in oscillator frequency of between 3 and 4 Hz. When the furnace shuts off, the ambient temperature slowly drops and the Softrock's frequency climbs until the  thermostat kicks the furnace back on.

I didn't measure the temperature excursion in the shop, but my sense is that it's around 3 to 4 degrees F. Call it 2 degrees C for a rough estimate. The Softrock oscillator changes about 4 Hz over the  temperature cycle, or 2 Hz/°C. At 7 MHz, therefore, the temperature coefficient is around 2 Hz/7 MHz, or 0.3 PPM/°C. For an inexpensive microprocessor crystal with no temperature compensation, this isn't terribly bad.

It is interesting that with a high quality time base, it's possible to measure the basement  temperature to a fraction of a degree through the Softrock oscillator frequency. In fact, Hewlett Packard used a similar approach many years ago in a precision thermometer, where the sensing element was a crystal oscillator with known frequency versus temperature characteristics.

It would, of course, be possible to improve the Softrock's crystal oscillator should higher stability be needed. For example, by either a constant temperature oven or a thermistor-based compensating network, or a combination of the two. Or, a suitable external TCXO might be substituted for the Softrock's oscillator. On a long term basis, of course, inexpensive microprocessor crystals will drift downward due to contamination and outgassing which shifts the crystal's resonant frequency lower as the contaminants add mass to the quartz plate. This process is accelerated when the temperature is increased, so while adding an oven arrangement to an inexpensive crystal oscillator may improve short term stability, it harms the long term stability by increasing downward drift. High quality timebase crystals are in hermetically sealed holders constructed from material with minimum outgassing problems.

There's no doubt that a Softrock Lite 6.2 receiver wins the performance per dollar contest by a long measure. For $10 (including postage!) you get a single band receiver that, when used with a PC and suitable software, has all the bells and whistles that one could want. And, the software is free, thanks to the dedicated efforts of some very talented hams. The Softrock's IP3 and minimum discernable signal performance is more than respectable.

However, many built-in sound cards will require replacement to obtain acceptable performance from the Softrock. And, an enclosure, isolation transformers and other bits and pieces will be necessary to obtain better performance. Depending on how well your computer is equipped and your junk box is stocked, these add-ons can add anywhere from $50 to $200 to the $10 Softrock receiver. I paid $125, for example, for the E-MU 0202 USB sound card. The parts required for the enclosure and transformers were on hand, but probably would run another $30 if purchased new.

For my particular computer and sound card arrangement, significant CPU resources are required to run the necessary SDR software. As I've said earlier, this is a function of how Dell designed the SX260's video memory, and how it bottlenecks graphic intensive programs such as those used for SDR reception. A computer with a normal graphics card should do much better.

Of the programs I've tried, PowerSDR offers the best support for sound cards operating at  96 or 192 Ks/s.

If your taste runs to a graphically "fully featured" program with simultaneous displays of about every possible signal, Winrad may be your desired program. I find it cluttered but that's personal opinion. It does well with my E-MU 0202 sound card at 192 Ks/s.

I would not recommend Softrock as a beginner project, nor would I recommend it to anyone without at least an above average understanding of computers. It is nowhere near the level of "open the box, plug it in and start copying signals."

Antenna leakage makes questionable using a Softrock 6.2 as a panadapter without better shielding and a buffer amplifier permitting at least 60 to 70 dB net isolation between the receiver's IF and the Softrock's antenna port. The AD8007-based buffer amplifier I designed (the Z10000) provides this level of isolation, at least up through 5 MHz.