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

E-mail: Jack.Smith@cliftonlaboratories.com
 

 

To search within the Clifton Laboratories site, enter your search term below.
 

 


Home
Up
Updates
Current Products
Prior Products - no longer available
Documents
Book
Software Updates
Softrock Lite 6.2
Adventures in Electronics and Radio
Elecraft K2 and K3 Transceivers

 

 

Jackson Harbor Press LF Converter Review

Written 27 August 2008

Last week, I purchased two LF Converter kits from Jackson Harbor Press, run by Chuck Olson, WB9KZY.

Jackson Harbor describes the kit's function:

The LF converter takes the band of frequencies from approximately 10 KHz to 300 KHz up to the 75, 30 or 20 meter ham bands (4 to 4.3, 10 to 10.3 or 14 to 14.3 MHz bands). The converter can be easily changed to another ham band (or another 300 KHz range) by changing the crystal. A socket is provided to make this easy to accomplish.

The kit is inexpensive at $14.00 plus shipping at $3.00. (Chuck shipped both my kits for one $3.00 shipping charge.)

The kit includes a small printed circuit board and all electronic parts. No printed documentation is provided; the purchaser downloads the manual, schematic and parts layout from Jackson Harbor Press's web site.

I've prepared a block diagram of the converter below. The kit includes both a 4 MHz and 10 MHz crystal so the user has the option of up-converting 0-300 KHz to the 4-4.3 MHz band or 10-10.3 MHz band. (Actually, as we'll see, the coverage extends beyond the 300 KHz filter cutoff.)

Since there is no output bandpass filter, the up-converted spectrum is mirrored about the output frequency. If using the 10 MHz oscillator, for example, a 100 KHz input signal appears at 10.1 MHz and also at 9.9 MHz. If you have a ham-band only receiver, such as an Elecraft K2, you probably will wish to use the 10 MHz crystal. You can then just mentally ignore the receiver's leading "10" digits and  read  the input frequency directly. (The kit includes a trimmer capacitor so you can net the converter's oscillator to be spot on.)

If your amateur band receiver doesn't have the 10 MHz band, you could use the 4 MHz crystal and tune the mirror spectrum 4.0 - 3.5 MHz, but you'll have to mentally invert the tuning, i.e., a 100 KHz input signal will appear at 3900 KHz, and tuning lower corresponds to higher input frequencies. Or, of course, you could substitute another frequency crystal in the kit.
 


The NE602 mixer (also known as the SA602, and its close relative, the NE612/SA612) is popular in battery powered amateur radio receivers, as it provides a balanced mixer, crystal or LC oscillator functionality, with about 15 dB net gain, with very low power consumption. As might be expected, "there ain't no such thing as a free lunch" and the downside to this popular chip is that it's intermodulation performance and gain compression figures are nothing to write home about. Philips Semiconductor's SA602A  data sheet specifies a  typical 3rd order intercept of -13 dBm. Since the converter may be exposed to very strong AM broadcast signals in the 530-1700 KHz band, the input low pass filter is very important for clean VLF/LF reception. I'll have more to say about the low pass filter later in this review.

Each kit took about 30 minutes to assemble. The printed circuit board is double sided, but does not have solder mask and silk screening. The pad and trace spacing is sufficiently large that lack of solder mask is not an issue. The kit is certainly suitable for a beginner; all through hole parts and, if you build the stock low pass filter, even the two inductors are supplied wound. No special tools are necessary.

The photo to the right shows the assembled board, but with different low pass filter inductors than supplied with the kit. I also added the 0.1" spaced header pins to provide an easy way to make connections to the converter.
 


I built one converter for Ron, K8AQC, to use in chasing low frequency interference around his neighborhood, in conjunction with a battery powered short wave receiver.

Since his converter must be battery powered, I replaced the 78L08 8V linear regulator with a 78L06, a 6 V part, as a 78L08 requires 10.5V input to be in regulation. In retrospect, even the 78L06, with a minimum input voltage of 8.5V is a bit marginal operating with a 9V battery that isn't fresh. A 5V regulator (7V minimum input) might have been a better choice. The NE602 works down to 4.5V, but the lower the supply voltage the worse the intermodulation performance.

I built this unit into a Radio Shack plastic case. (The case has both a plastic and metal top, and I used the aluminum top.

It has a BNC input connector paralleled with an 8-32 threaded stud for a telescoping antenna. The output is also a BNC and the unit has a power switch and power on LED.

I did not use Jackson Harbor's standard input low pass filter in K8AQC's unit. The reason for that is the filter components supplied with the kit are designed for 50 ohm input and filter output impedance.

At VLF and LF, a short whip antenna can be modeled as a few pF capacitance in series with a full size antenna. At, say, 60 KHz, therefore, the short whip antenna has an output impedance in the neighborhood of 500KΩ capacitive reactance. Operating into a 50 ohm impedance, a loss on the order of 80 dB can be expected.  One normally uses an JFET source follower to efficiently match a short receiving whip at VLF/LF frequencies to a 50 ohm receiver or converter, but in Ron's case such a front end would cause far more problems than it would solve, as he lives in an area with AM broadcast signal strengths exceeding 1 volt/meter, an exceptionally strong signal. Hence, no JFET source follower front end.

As a compromise filter, however, I designed a 4.8KΩ low pass filter of similar cutoff frequency to the filter supplied with the kit. Why 4.8KΩ? It turns out that is the input impedance of the NE602. There's still a mismatch loss of 40 dB or so between the whip and the filter, but 40 dB is better than 80 dB. And, of course, the filter's attenuation versus frequency characteristics will vary wildly with over the VLF/LF range as the source impedance presented to it by the whip antenna changes.
 


The second converter I built for my use around the shack and to use in working on my active antenna project. I housed it in a small Bud utility box. Input and output are BNC connectors and the DC power is brought in through a feed through capacitor. The board is mounted with 4-40 aluminum standoffs, 3/8" tall. Since my converter will be used with an external power supply, I built it with the standard 78L08 8V regulator.

The filter inductors are wound on FairRite FT50-61 ferrite cores.


How does it work? Considering the $14.00 price, it works very well indeed. (It can be made to work a bit better if you are willing to do some extra work, as we'll discuss below.)

The spectrum analyzer image below is captured with my Advantest R3463 spectrum analyzer and shows the range 0-100 KHz as seen when connected to my home brew active antenna. The double blip at 24-25 KHz are two US Navy stations, Cutler ME at 24 KHz and one on North Dakota at 25.2 KHz. The stations are a bit close to  resolve into individual signals at this resolution bandwidth. WWVB at 60 KHz can be seen, along with the Canadian Navy station CFH at 73.6 KHz. Between 90 and 100 KHz are LORAN-C navigation signals, centered at 100 KHz so only the lower half of the signal and modulation sidebands are visbile.

The image below shows the same 0-100 KHz slice of the spectrum, but this time shifted up to the range 10.0 - 10.1 MHz with the Jackson Harbor converter. It is connected to the same active antenna used in the direct spectrum analyzer sweep above. (The two images were captured 2 minutes apart in time, so propagation,  to the extent it changes from day to night at these frequencies is not an issue.)

In general, the direct and converter output spectrum plots look identical. A few comments should be made, however:

  • My R3463 spectrum analyzer has a minimum calibrated frequency of 9 KHz, as below that the input DC blocking capacitor rolls off. You can see this along with the local oscillator feedthrough in the first half graticule division.
  • My active antenna rolls off below 10 KHz as well, and that rolloff is quite visible in the converter output, although it is not visible in the direct connection.
  • Signals  through the converter are about 10 dB stronger than for the direct connection case, representing the NE602's gain.
  • I live in an area thankfully free of strong AM broadcast band signals and I found no  trace of AM broadcast intermodulation interference in the converter's output.

I also spent some time tuning around using an Elecraft K2 transceiver in the 10 MHz band to listen to the converter output. With my prototype active antenna, WWVB was quite audible at my northern Virginia location, with both the high power and low power (-17 dB) portions being easily heard. I did not hear AM broadcast band intermodulation products, but my location fortunately has no extraordinarily strong AM broadcast signals. (For  testing in areas of high AM broadcast signals, I rely upon Ron, K8AQC, who lives within the 1V/meter contour of a couple of AM stations.

 

Low Pass Filters

I've mentioned the converter's low pass filter input several times for a reason. Before going further, however, I wish to clearly state that the low pass filter supplied with the kit is perfectly serviceable and there's no reason for the average builder to make changes in it. I can't resist, however, a chance to tweak the filter and make things slightly better. And, the Converter serves as a example to discuss filter design and component selection.

The figure below illustrates the low pass filter section of  the Converter. Although L1 and L2 are not specified by inductance, 8 turns on an FT37-43 core corresponds to 22.4 µH.

Chuck Olsen, in an exchange of E-mail messages says the input low pass filter is a half-wave design, with a 300 KHz cutoff frequency and 50 ohm input/output impedance. Chuck used a half-wave design as it is said to be more forgiving of component tolerances and impedance mismatches than more exotic filters.
 

Whether a filter works as designed depends on several things. Assuming the components are within a percent or two of the design value, the most important remaining factors are (a) the impedances presented to the filter on both the input and output side and (b) losses in the inductors, i.e., their Q factors. Other considerations include parasitic capacitance in the inductors and parasitic inductance in the capacitor as well as unwanted coupling due to the filter's physical construction and layout.

From what I can determine, all NE602/SA602 data sheets say the device's input impedance is 1.5 KΩ. That figure is repeated in more amateur radio articles than I cared to read through.

The problem is, it's not correct.

Although the data sheet is not completely clear, it seems that the input impedance quoted is for balanced input, i.e., the signal is applied differentially between pins 1 and 2. The Converter grounds pin 2 and applies a single ended signal at Pin 1. (This also throws away 6 dB of gain and also likely degrades the 3rd order intermodulation intercept, but adding a balanced input to the Converter that operates properly at several kiloohms impedance over a frequency range 10 KHz - 500 KHz is not trivial.) To better understand the NE602's input impedance, I built one kit without an input low pass filter and measured the input impedance using an HP 87510A vector network analyzer.

The figure below shows a measured input impedance of 4.8 KΩ essentially pure resistive over the range 25 KHz to 500 KHz. The fluctuations below 25 KHz are in the test setup, as the 87510A is operating outside its recommended frequency range for this particular setup.


Looking around the internet to find confirmation for my measurement, I found an incidental comment by Wes, W7ZOI, at http://w7zoi.net/imtc.pdf:

Regarding the NE602 input impedance: (10Aug07)

Often we see the NE602 input modeled as ideal transistors (which means infinite beta) with 1.5K resistors from each input to “ground.” From what I’ve been able to tell, the circuit is really just a differential pair with a total bias current of about 2 mA (1 mA per transistor.) The equivalent emitter résistance is then re=26/Ie(mA), or 26 Ohms. If the beta is 100, this would put the resistance looking to the base at 2.6 K. If we consider frequencies near or just above the beta cutoff, the net impedance drops down to 1.5K, justifying the traditional model.

But there is one more detail: The base-to-base input impedance of a bipolar differential pair is twice that of the single transistor. This is the result of the common emitter connection that is between the two devices. This point would be a virtual ground if the bases are driven differentially. It is a floating point with half the input voltage if one of the bases is bypass to ground, which is the usual situation with a Gilbert Cell. Hence, the input impedance is around 3K paralleled with a couple of pF of capacitance. This is just what I measured several years ago when I had daily access to a wideband network analyzer.

This prompted an exchange of E-mail messages with Wes, in which he noted:

The 3 K or more value makes sense with regard to the part having a diff amp input stage and standing a current of 2 mA.    Glad to see you came to the same conclusion.

It's difficult to know what the driving impedance of the converter's input will be. My prototype active antenna, for example, has an output impedance very close to 50 ohms over the frequency range 10 KHz - 30 MHz. A random wire antenna will likely to range between tens of ohms and hundreds of thousands of ohms reactive impedance over the range 10 KHz - 30 MHz. (It's important to consider  the filter's response over the full frequency range, not just the 10 KHz - 300 KHz range as the filter's job is to reject signals above 300 KHz or so, including those at 4 or 10 MHz  that otherwise might leak through into the converter output. To some degree, these leakage signals are suppressed by the NE602's mixer balance, but most of the suppression must come from the input low pass filter.)

As a sweeping generality, filters that are not driven by a source and terminated in a load matching the design impedances have "lumpy" responses, with peaks and dips not seen when the filter is properly terminated.

To obtain a quick look at the converter's response, I made a series of spot attenuation measurements with an HP 8657A signal generator source (50 ohms). The filter output is based upon reading the converter output signal level with an Advantest R3463 spectrum analyzer. When a 50 ohm source, such as the 8657A signal generator, is operated into a different impedance  than 50 ohms, the delivered signal level will not match the generator level setting. Into an open circuit, for example, the delivered voltage will be twice the indicated value. Since the filter input impedance varies, I've normalized the plot to 0 dB at 100 KHz to provide a better comparison.

The red plot is the output when the converter is built according to the instructions, with the supplied parts. The green plot adds a 49.9 ohm, 1% resistor across the input filter's output (C6 in the schematic) thereby providing both a 50 ohm source and termination.

Several things are immediately apparent from this plot:

  • When terminated into the 4800 ohm impedance of the NE602, there's a clear peak in the filter response, broadly centered around 300 KHz;
  • Without the correct termination load, the filter gives up 5 dB or more rejection in the lower end of the AM broadcast spectrum; and
  • When terminated with 50 ohms, the filter shows passband tilt with no clear transition between passband and stopband. The passband tilt is generally evidence of low inductor Q.
The kit is supplied with two wound inductors, each consisting of a few turns of no. 30 Kynar "wirewrap" type wire on a small binocular core of FairRite type 43 material.

I've written at http://www.cliftonlaboratories.com/estimating_q_of_ferrite_cores.htm about Type 43 material and using it in  tuned circuits. In general, Type 43 material has quite low Q and has an inductance that varies with frequency. For most filter designs, even one of these problems is bad news and both together are even worse. It's tempting to use Type 43 material in filters and tuned circuits because it has a high effective permeability and thus permits relatively high inductance in a compact form factor. However, one must be fully cognizant of the problems as well as the benefits. (Note that the newest Type 43 material formulation has significantly reduced losses [higher Q] and may prove more suitable for filters and tuned circuits. Care must be exercised, however, before designing it into a circuit.)

Using the HP87510A vector network analyzer, I looked at the inductance and Q of the filter inductors provided with the Converter kit. A few spot values show the problem.

Freq (KHz) L (uH) Q
10 39.8 3.7
50 29.7 3.2
100 26.6 3.5
500 21.2 4.6
1000 20.0 4.7

The inductance varies by 2:1 over the frequency range of interest and the Q never breaks 5, an exceptionally low value for a low loss filter. Looked at over the full frequency range of interest, up to 30 MHz, an even greater variation in inductance and lower Q will be seen.

Considering these problems, why does the stock filter work as well as it does? First, this is a fairly forgiving application, with only five frequency selective components. Second, to some degree the low Q benefits the filter. The inductors can be thought of (and modeled as) as a high Q inductor in series with a resistor, with the resistor value increasing with frequency and the inductance decreasing with frequency. This transforms the filter into a mixed LC and RC low pass filter. The main drawback is that the RC aspects of the filter cause increased loss everyplace, including the passband—hence the tilted passband region seen in the plot above.

One easy modification is to replace the two supplied inductors with higher Q parts. For low frequency work, I like FairRite Type 61 material. It has a moderate relative permeability (125) that is stable with frequency over a reasonable frequency range. Equally useful, up to 10 MHz or so it has quite low loss compared with Type 43 material.

I wound two 22.8 µH (nominal) inductors (18 turns, No. 22 AWG solid magnet wire) on FT50-61 cores, and measured their inductance and Q at 2.5 MHz using an HP4342A Q-meter. At 2.5 MHz, the measured Q values were 185 and 168.

 

Sample No.

L (uH) Q
  1 20.5 185
  2 21.4 168

At 1 KHz, the measured inductance was within 5% of the 2.5 MHz values.

Using two FT50-61 based inductors makes the frequency response look more like a filter should, as seen in the plot below. Both data sets are taken with 49.9 ohm terminating resistor across C6 so that the filter is correctly terminated. Note the passband is flat, within a half dB or so and there's a clear transition from passband to stopband.

The two filter's stopband differences above 600 KHz are negligible. However, the increased loss in the stock inductors does provide extra attenuation in the range 400 - 600 KHz, albeit at the price of increased passband loss.

It's possible to squeeze improved stopband performance from the filter if one uses a different set of design criteria. To see the possible difference, I designed and built a Chebychev low pass filter. In order to provide a "head to head' comparison of the filter types, I built three low pass filter prototypes on a breadboard:
  • Stock filter with the binocular 43 material cores
  • Stock filter with the high Q FT50-61 cores
  • Chebychev filter, 400 KHz cutoff, 3 dB passband ripple. It also uses FT50-61 inductors wound with Litz wire.

Pictured to the right is the stock filter design with FT50-61 cores. As an experiment, I wound these inductors with Litz wire to improve the low frequency Q. The improvement over solid No. 22 AWG magnet wire is small, however.

I used the excellent filter software AADE Filter Design, available at http://www.aade.com/filter.htm to design the Chebychev prototype.
 

Using the HP87510 vector network analyzer, I swept all three filters over the range 10 KHz - 10 MHz. The horizontal axis is log scale in these plots, and I set markers for three passband frequencies and the lower and upper AM broadcast band limits. Building the filters on the prototype board generally yields better results than a typical printed circuit board design because the ground plane is larger. In addition, the HP87510A provides better source and load terminations.

First, the stock filter with stock components.

Next, the stock filter design, but with the low Q inductors replaced with high Q parts, FT50-61 core, wound with Litz wire. Note the clean transition from passband to stopband, as well as how flat the passband is up to the 300 KHz 3 dB down cutoff frequency.

Finally, the Chebyshev filter.

Superimposing all three filters onto a single plot:

If  3 db ripple in the passband is acceptable, the Chebychev filter pays significant benefits in the stopband. If a flat passband is more important, then the stock filter design with FT50-61 inductors is a good choice.

To repeat my comment at the outset, in almost all applications for which the Jackson Harbor Press  converter is used, the stock filter, built with the supplied components will prove satisfactory. However, some filter performance improvement is possible with either (a) better quality inductors and (b) a Chebychev style filter design.


To show the effectiveness of the converter low pass filter (with high Q FT50-61 inductors but otherwise stock filter design), I've made two spectrum analyzer captures. Both use my prototype active antenna with a 50 ohm output impedance, so your results with a random wire antenna, for example, may differ. I've also modified my converter by adding a 49.9 ohm, 1% terminating resistor across C6.

The first spectrum analyzer capture is made with my Advantest R3463 connected directly to the active antenna. Note the many strong AM broadcast signals between 530 and 1700 KHz. 

The capture below shows the converter output over the same input range 0 - 2 MHz. The lower end of the broadcast band (the blue dot marker is at WMAL, 630 KHz) is not suppressed as well as the higher frequencies, of course. Still, the difference is quite noticeable, particularly when you consider that the converter has about 10 dB gain.

I also designed and built a Chebychev band-reject filter for the AM broadcast band, to be discussed in a future web page. Its effectiveness in knocking down AM broadcast signals can be assessed with the spectrum analyzer image below, showing the results of connecting my active antenna to the spectrum analyzer through the band-reject filter.