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

E-mail: Jack.Smith@cliftonlaboratories.com
 

Elecraft's K3 transceiver, when equipped with the optional KXV3 Interface Option board, provides an isolated, wideband, sample of its 8.215 MHz IF. The sample is brought out to a BNC connector on the K3's rear panel. The sample is from the K3's main receiver.

If you prefer not to read  the detail, you may skip directly to my recommendations.

Table of Contents of this Page
How_the_K3_Provides_the_IF_Sample
What_then_is_the_transfer_gain_of_the_K3_
Softrock__Lite_6.2
Z90_Panadapter_
K3_IF_Sample_Isolation
Unwanted_Signals_Out_of_K3
Leakage_to_KRX3,_2nd_Receiver_
Is_this_a_real_problem
Alternative_to_increasing_K3_IF_sample_port_output
Recommendations
Note_on_Using_a_Z90_With_K3

Created 16 August 2008
Revised 18 August 2008 to add KRX3 information and revise alternatives
Revised 20 August 2008 added GM0ELP E-mail and link
 

How the K3 Provides the IF Sample

The illustration below provides an overview of how the IF sample is derived. Since the sample is taken before the K3's narrowband "roofing" crystal filters, the sample's bandwidth is limited only by the K3's input bandpass filters.

So far, so good. However, Elecraft's designers and I disagree on the desired IF sample "transfer gain" of the sample port. The term "transfer gain" is appropriate, rather than just "gain" because the IF sample is at a different frequency than the input signal. Based upon typical specifications for military and commercial receivers, I believe the transfer gain should be approximately 0 dB. In other words when the K3 is set to 7 MHz, 1 uV signal at 7 MHz  into the antenna connector should yield close to 1 uV at 8215 KHz out of the IF sample port, measured into a 50 ohm instrument. I wouldn't be too concerned with a dB or two one way or the other, but the target should be 1:1, or 0 dB transfer gain.

In fact, it's not uncommon for military and commercial receivers to have a significant positive transfer gain at the IF output sample,  thereby allowing an external panadapter with relatively poor sensitivity to be used. The theory is apparently that it's easier to reduce the panadapter's sensitivity with the gain control or via an attenuator than to have to add an outboard amplifier.

Watkins Johnson, for example, provides the following specification for the WJ-8718 series HF receivers:

Cubic's R-2411 receiver has both a wideband and narrowband IF output port. The wideband port, suitable for use with a panadapter, has approximately 10 dB transfer gain. The 455 KHz sample, like the WJ-8718, is post filtering and post AGC, and is intended for output demodulators.

I measured the K3's transfer gain at 7 MHz. The K3's input signal is from an HP8657A signal  generator set to -60.0 dbm. The 8215 KHz output level is measured with an Advantest R3463 spectrum analyzer. Checking  the HP8657A signal  generator's output with the R3463 spectrum analyzer showed -59.0 dBm. The data below is based upon input signal level of -59.0 dBm.

In the most normal operating conditions, therefore, the 8215 KHz IF sample is between 7 and 17 dB below the level of the signal at the antenna port, not the desired 0 dB to positive gain.

Is this a problem? My measurements suggest the answer is "yes."

I looked at two potential panadapters; a Softrock and my Z90. (Telepostinc's LP-PAN has a built-in amplifier to overcome the transfer gain problem.)

I've previously covered the Softrock 6.2 at:
Softrock Lite 6.2
Using Softrock as a Panadapter for the K2

For the tests discussed below, I used the Softrock 6.2 described at my pages linked above. The Softrock receiver is packaged in a shielded enclosure, with 600:600 ohm wideband isolation transformers on both audio output channels. An E-MU 0202 USB-connected sound card running at 192 ks/s is used in these tests and a Dell laptop computer, with a dual core Pentium processor and 4 GB RAM. The signal source is an HP8657A synthesized signal generator, with a GPS disciplined 10 MHz external time base. Flex Radio's 1.12.1 software is used in these measurements. The K3 is set to the 7 MHz band, with the preamplifier off.

As a starting point, we note that a -100 dBm (2.2 µV) signal at 8215 KHz fed directly into the Softrock's antenna port is easily observable, being 15 dB or so above the noise level.

Adding an external amplifier (a Z10000-U) between the K3's IF output port and the Softrock significantly improves the displayed signal. The Z10000-U employed is set for 10 db gain and has a 50 ohm input impedance, obtained by adding a 49.9 surface mount resistor to the Z10000-U's input pad. If intended to be used only with the K3, the 49.9 ohm resistor could be dispensed with and  the Z10000-U's gain reduced to the 6 - 8 dB range. (As discussed below, the K3's IF output port is not 50 ohms.)

I designed the Z90 to integrate with an Elecraft K2 using a Z10000-K2 buffer amplifier to provide an IF sample at net 0 dB transfer gain. Accordingly, I expected additional gain to be necessary when using a Z90 with a K3.

With similar test conditions, except the input signal is -96 dBm, not -100 dBm, a Z90 connected directly to the K3 shows a barely perceptible trace.

First, we see background noise, which was below the first image's sensitivity. The -96 dBm signal is seen centered on the Z90's display. albeit with some noise fluctuations.

Of course, turning the K3's preamplifier on increases the IF sample signal by approximately 10 dB. However, operating the K3 with the preamplifier on may or may not be desirable, depending upon the frequency band and signal levels in general.

The K3 Owner's Manual describes the IF output as follows:

In particular, Elecraft notes that Additional isolation circuitry may also be  required.

Isolation can work in either direction; to keep signals that might leak from the panadapter back into the K3 or to keep signals other than the 8215 KHz IF from getting into the panadapter's input.

We'll first look at isolation from the outside world into the K3 via the IF sample port.

To evaluate the port isolation, I used the following test setup. Set the K3 to 7.000 MHz, AGC turned off, preamplifier and attenuator turned off. CW mode, bandwidth set for 600 Hz, passband centered on 600 Hz.. Connect an HP8657A signal generator to the K3's antenna port and a true RMS voltmeter to the Line Out audio output (I used an HP 3400A). Set the signal generator to 7.000 MHz, touch up the K3's tuning if necessary to center the signal in the passband and increase the signal generator's output (starting at -140 dBm) until the signal plus noise as observed on the true RMS voltmeter increases to +20 dB over the reading observed with the signal generator off.

Remove the signal generator from the K3's antenna port and connect a 50 ohm termination on the antenna port. Set the signal generator to 8215 KHz and connect it to the K3's IF output port. Increase the signal generator level until the signal is easily audible. Using the K3's spot or frequency tuning display option, adjust the signal generator's frequency if necessary to be center tuned, as with the reference signal. Adjust the signal generator's level until the same true RMS voltmeter reading is observed as with the direct input case.

The IF sample port isolation is the difference in signal generator level between the two connection methods.

I found the isolation to be 67.8 dB. I don't believe my measurement equipment and technique is accurate to 0.1 dB, so I'll call it 68 dB in round numbers.

Is this enough isolation? It's difficult to impossible to answer this in the abstract. If we look at a couple of specific examples, however, we can sense whether it is or is not.

First, the Z90 is designed so that the strongest leakage signal out of the signal input port is less than -80 dBm. When combined with 68 dB internal isolation, any unwanted signal is going to be at a level below normal measurement ability.

The Softrock Lite receiver is a different story, however.

The image below shows the local oscillator leakage from a 40 meter Softrock Lite receiver at around -39 dBm. The Softrock Lite used as a K3 panadapter has a 8192 KHz local oscillator, with leakage levels similar to those of the 40 meter unit.

When combined with 68 dB isolation, the resultant signal will be -107 dBm. If the Softrock Lite were operating at 8215 KHz, this level of signal would be quite objectionable in the K3's IF. However, the Softrock's local oscillator is approximately 23 KHz offset at 8192 KHz, which is outside even the K3's widest FM roofing filter.

There is always a concern when unwanted signals are injected back into a receiver's IF chain. In this case, however, I believe the risk of problem is relatively low, considering the signal levels involved.

The main caveat to this statement is that the measured 68 dB isolation is between the IF sample port and the receiver's IF chain. The isolation between the IF sample port and other portions of the K3's circuitry are not known and are not, in general, easily measured.

If a Z10000-U buffer amplifier is installed to pick up the K3's gain shortfall, of course, the extra isolation provided by the Z10000-U will knock even the Softrock Lite's signal leakage down into the noise level.

The second component of isolation is knowing what signals in addition to the 8215 KHz IF sample come out of the IF sample port.

To determine this, I connected an Advantest spectrum analyzer to the IF sample port and observed emitted signals over the range 0 - 50 MHz, with the K3 set to the bottom edge of the band for all amateur bands between 1.8 and 50 MHz.

The image below is typical of my measurements. The K3's local oscillator produces a reasonably strong leakage signal, typically in the -50 to -60 dBm range. The local oscillator is 8.215 MHz above the tuned frequency, except on 50 MHz where it is below the tuned frequency.

Rather than present a series of these images, one for each band, I've made a composite image by overlaying all measurements onto a single plot.

The result is quite impressive. Whether these local oscillator leakage signals will cause your panadapter a problem is difficult to say. We do know, for example, that the Softrock type receivers have known spurious responses at odd harmonics, e.g., at 3 x 8215 KHz, at 5 x 8215 KHz, etc. These responses have reduced sensitivity by a factor of 20 * log(n) where n is the harmonic number, and the Softrock has some bandpass filtering in the front end.

I also looked at the K3's IF output after installing a Z10010-K3 bandpass filter.

When running the same 1.8 to 50 MHz analysis, all the plots with the filter in place look the same—nothing to be seen. All the local oscillator leakage is reduced below the spectrum analyzer's noise floor, as shown in the plot below.

If the K3 is equipped with the KRX3, second receiver option, another possible leakage or isolation concern is from the IF sample port back to the KRX3's IF.

To measure this, I used the same approach as with the main receiver, but listening and measuring the KRX3's audio output for comparison signal levels.

First, the good news. With an extremely strong signal, 0 dBm, at 8215 KHz, pumped into the K3's IF output sample port, there is no trace of direct 8215 IF leakage into the KRX3's IF chain. This is excellent isolation.

The not-so-good news is that I found a new spurious path with about 90 dB isolation. I believe, however, that this may be more of a theoretical issue than a real one for most operators.

This may be a bit complicated, so please bear with me.

The figure below shows a simplified view of the K3 with the second receiver. The black arrows show the normal signal flow when the main  receiver is tuned to 7000 KHz and  the sub-receiver to 7010 KHz.

If a strong signal at 8205 KHz is injected into the IF sample port, there is sufficient leakage backward through the K3's post-mixer amplifier, balanced mixer and antenna coupler to produce a clearly audible signal into the sub receiver. This signal path is shown in red in the image below.

I've used these specific frequencies as an example, but in general for any frequency that the main and sub  receivers are  tuned to, it is possible to generate an on-frequency spurious signal into the sub receiver by appropriate selection of an 8 MHz range signal injected back into the IF sample port.

Mathematically, Fd = Fb - Fc.  Fb = Fa + 8215 KHz (for most bands; in some cases Fb = Fa - 8215)

Thus, for a given frequency injected into the IF sample port, as you tune the main receiver, the spurious frequency changes. In the case of an 8205 KHz signal, for example, the spurious will move with the main receiver tuning, always being 10 KHz above the main receiver dial frequency. (These comments ignore for simplicity the fact that the K3's IF frequency is not precisely 8215 KHz.)

If the K3 is not equipped with the KRX3 sub-receiver, the same mixing phenomena still occurs, of course.

Either with or without the KRX3, the spurious will be emitted from the K3's antenna port associated with the main receiver.

The spectrum analyzer plot below shows the signal leaking out of the K3's antenna port when a strong 8204 KHz signal is injected into the K3's IF sample port. The transfer coupling is -89 dB in  this case.
 

As the image below,  taken with the same  test conditions but with the K3's preamp on, the spurious signal is reduced another 16 dB.
 

So, is this a real problem or one that is of only theoretical concern? In all but highly unusual circumstances, I believe it's unlikely to be a problem for the following reasons:

• In order to be detectible in the KRX3 sub receiver, the signal injected into the IF sample port must be -20 dBm or stronger. The worst case signal leaking out of a Softrock receiver used as a panadapter is around -40 dBm. This provides 20 dB margin which should be adequate.
• The prevailing coupling path seems to be through the K3's receiving antenna splitter. If the problem is observed, switching the KRX3 to the external receiver input port drops the spurious below the noise level with a 0 dBm level signal injected into the IF sample port.
• In order for the signal to be heard,  three frequencies have to coincide, which reduces the statistical chance of seeing the problem.

The most likely way this spurious problem will occur would be for a strong signal from a nearby  transmitter such as multi-band simultaneous operations, to be coupled into the IF port through, for example, a long poorly shielded cable connected to a panadapter, or the panadapter itself being unshielded. If actually observed, correcting the shielding problem should ameliorate the interference. Or, an 8215 KHz bandpass filter could be added to the K3's IF sample output port.

Looking at the K3 circuit fragment below, we see why the IF sample has a significant amplitude drop.

Q10, a J310 JFET, is configured as a source follower. Its gate is driven by a voltage divider comprising R89 and R8. This voltage divider presents Q10's gate with a 15 dB loss compared with the signal at J77, pin 1.

Secondly, JFETs are notoriously poor as source followers when feeding a low impedance load. An FET's output impedance is 1/g_m,  where g_m is determined at the FET's quiescent point. (g_m is also known as g_fs in some data sheets) The J310 has a quoted gm as between 8000 and 18000 µS according to Fairchild Semiconductor's data sheet, but even that spread of numbers is only valid for a particular idle current.

If we take the midpoint of the data sheet numbers g_m is 12,000 microsiemens so 1/g_m is about 83 ohms. Connected to a 50 ohm load, therefore, an additional loss of 8 dB is seen. (I measured the output impedance of my K3's IF sample port as 74 ohms, quite close to the 83 ohm midpoint estimate.)

Hence, the total loss between Pin 1, J77 and the signal level seen in a 50 ohm measuring instrument can be estimated as 23 dB.

We're about  4 dB short of the measured loss. (The block diagram notes 5 dB loss in the noise blanker. However,  the IF sample point is connected ahead of the noise blanker attenuator.) I've estimated the mixer loss at 6 dB in the absence of data in the Operator's Manual. I'm puzzled as to source of the extra 4 db loss.

It should be possible  to increase Q10's output by altering the R89/R8 voltage divider, preferably by increasing R8. If R8 is, for example, increased to 4.7K, the voltage divider loss decreases from 15 db to 6 dB, picking up 9 dB gain.

Larry, N8LP, at Telepostinc has done work in this  regard and has a recommended component change for R8 (to 15K) and R9 (to 49.9) that improves the  transfer gain by nearly 10 dB. The details may be found at http://groups.yahoo.com/group/LP-PAN/files/K3_Buffer_Mod/

If you make this modification, I suggest measuring the voltage across the new R9. Q10's drain current is set by a combination of the voltage drop across R9 and the individual characteristics of the J310 part in your K3. FETs have a notoriously wide part-to-part spread in parameters and it's possible that the J310 device in your K3 has parameters sufficiently far from the mean that Q10's power dissipation limits will be reached or exceeded. The surface mount J310 has a maximum power dissipation rating of 350 mw, and for reliability a safe operating value is 200 mw or so. Q10's current can be easily determined by measuring the voltage drop across R9. If changed to 49.9 ohms, Q10's drain current Id is 1000*Vs/50 (in milliamperes) where Vs is  the voltage measured from ground to Q10's source pin. To calculate the power dissipated in Q10, measure its drain voltage. The power is then (Vd-Vs)*Id in milliwatts, where Id is in milliamperes.

An example. After replacing R9 with a 49.9 ohm resistor, the following voltage readings are found: Vs = 2 volts, Vd = 12 volts. The idle current through Q10 is thus 2 / 49.9 = 40 mA. The power dissipated in Q10 is (12-2) * 40 = 400 milliwatts. This exceeds the J310's maximum permissible power dissipation and would not be a good design practice.
 

1. To increase the IF sample signal level consider an external Z10000-U broadband buffer amplifier. Build the Z10000-U for 6 dB to 8 dB gain, with the normal high Z input.
2. An alternative is to modify R8 and R9 as described above. If  this approach is taken, verify that Q10's maximum power dissipation limit is not exceeded, after being detrated according to good engineering practices.
3. The Z10000-U will also reduce signal leakage back into the K3 into the noise level when using a Softrock
4. These steps are not necessary with an LP-PAN panadapter as it has a  built-in isolation amplifier.
5. If you wish to eliminate the K3's local oscillator output signals, consider a Z10010-K3 bandpass filter. If you use both a Z10000-U and Z10010-K3 filter, the Z10000-U's gain should be increased by 3 to 4 dB to compensate for the filter's loss.
6. Depending upon your requirements for display sensitivity, it may not be necessary to use an amplifier or filter at all or to modify the K3's IF sample components.

The Z90 or Z91 has an 8215 KHz stock IF frequency selection. In running tests recently, however, it appears undesirable to use the stock 8215 KHz IF frequency selection with the K3. The reason is that the Z90's sweep oscillator exhibits greater spurious responses than desirable at 8215 KHz.

I recommend defining a new "optional" frequency of 24215 KHz and using  that with the K3. The Z90's IF is 8.000 MHz and when set for 24215 KHz the Z90's sweep oscillator is centered at 16215 KHz. This frequency also provides a response at 8215, i.e., 16215 - 8215 = 8000 KHz. When running at 16215 KHz, the Z90's sweep oscillator exhibits far fewer spurious responses.

In addition, in certain bands, the Z90's dynamic range can be extended if a Z10010-K3 bandpass filter is used. Some amateur bands have strong shortwave broadcast signals a few hundred KHz above or below the amateur band. The Z10010-K3 bandpass filter will reduce the strength of these signals when entering the Z90's input stage. (By design, the Z90 has no frequency selective filters on the input stage.)