Design Notes, FST4W, SDR, Turn Island Systems, WSPR

SDR Dynamic Range, Filters, and Broadcast Band Transmitters

January 14, 2025 paul Leave a comment

As we push the envelope of SDR bandwidth with receivers like the RX-888, In addition to sample rate and aliasing concerns we also also need to look at the dynamic range issues. With traditional “analog” receivers the dynamic range is usually dominated by the IMD (intermodulation distortion) of the front-end preamp and the following mixer stage. Techniques to optimize dynamic range include pre-filtering and AGC (automatic gain control), which can be quite effective when used in relatively narrow-band operation.

But AGC and filtering aren’t so easy in a SDR that is receiving the entire spectrum from (say) 10 KHz to 60 MHz. There is a huge range of signal levels, especially when you are close to a megawatt-level broadcast transmitter. While the 16-bit ADC (analog to digital converter) in the RX-888, with it’s 96dB range, is clearly superior to the 12 and 8-bit ADCs used in many other SDRs, the real-world range of RF signals can easily exceed this.

By the way, the effective dynamic range of an ADC is often several dB better than is suggested by the number of bits, due to what is essentially noise dithering and oversampling.

In some cases the programmable attenuation provided by the SDR can be adjusted dynamically for the best use of the available dynamic range. In addition, there is always the possibility of digital overload and/or digital gain control via numerical scaling within the numerous digital signal processing stages of the software. These are subjects for a different discussion. Here, I’m talking about filters.

While some amount of ADC over-range clipping is tolerable, we usually try to minimize this by inserting attenuation or otherwise setting the gain before the ADC sufficient to avoid significant overloads. We also use a moderate amount of frequency-shaped filtering to reduce input levels at lower frequencies where the general atmospheric noise establishes a higher noise floor. The Turn Island Systems SDR Filter and Filter-Preamp are examples of this type of filtering:

This filter has a steep 30 MHz anti-aliasing low-pass filter, and a high-pass “shelf” filter that attenuates the frequencies below around 10 MHz, flattening out below 1 MHz to a “shelf” attenuation of about 28 dB. This shelf filter roughly compensates for the change in the general atmospheric noise level, and attenuates the USA AM BCB (broadcast band, 540 – 1700 KHz) , helping reduce the chance of ADC overload.

A similar filter/preamp (under development) places the LPF corner frequency at 60 MHz, allowing reception of the 6-meter ham band (requiring a RX-888 SDR sample rate of 130 MHz). This filter also has the stopband notches placed to optimize attenuation of the 88 – 108 MHz FM broadcast band, which would otherwise be aliased down to 22 – 42 MHz in frequency. While the RX-888 does have a built-in 60 MHz LPF, it’s characteristics are pretty marginal (see the links at the end of this post for more details).

While the shelf filter reduces the chance of AM BCB overload, strong signals elsewhere can still cause problems. If these are in the ham bands then simple filtering is not going to help. But some people are using tunable notch-filters to attenuate specific signals, or ham-band bandpass filters to attenuate out-of-band signals. Here is an eight-band bandpass filter I am currently working on (I’m calling it the BPF8):

The BPF8 is actually a bank of three-section bandpass filters connected in parallel. Each filter has this design:

The (simulated) overall response looks like this, covering the 80, 40, 30, 20, 18, 15, 12, and 10 meter ham bands. These filters are tuned for the WSPR sub-bands:

BPF8 frequency response (linear frequency plot)

Will this help with SDR dynamic range? That depends on the interference — if the out of band signals aren’t strong enough to create problems then this filter will be unnecessary.

If reception between the filter bands is still desired, a broadband shelf response can be selected (via a jumper), which limits the maximum attenuation — here, to 30 dB:

BPF8 frequency response, -30dB shelf (linear frequency plot)

In the USA, most of the BCB signals are below that 1700 KHz frequency. But in Europe there are many powerful broadcast transmitters scattered over the HF frequencies, and these can cause big dynamic range problems. Here are the shortwave BCB frequency bands superimposed on the above multiband filter response (some of these bands are only lightly or regionally in use):

As you can see, some of these SW BCB frequencies are extremely close to the ham bands, and so in those cases the filter will probably not help much, while other SW BCB signals will be significantly attenuated. The BPF8 will probably include jumper options to optionally disable a selected filter section. Is the BPF8 worth a try? We shall see.

The BPF8 prototype board is densely populated with 24 toroid core inductors, used for their low-loss and self-shielding characteristics. The filter capacitors are on the underside of the board. This will be another “hard to build” design, because of all the hand-wound and tuned toroids, but I hope the results will be worth it.

For more details of the RX-888, and about shelf filters, look at KA7OEI’s blog for some excellent in-depth discussion of these issues:

measuring-signal-dynamics-of-rx-888-mk2

frequency-response-of-rx-888

revisiting-limited-attenuation-high-pass-filter

International broadcast bands:

International_broadcast_bands (Wikipedia)

Beacon Blaster, FST4W, WSPR

FST4W-120 on Ten Bands Simultaneously

November 22, 2024 paul Leave a comment

As you may know, over at Turn Island Systems (TIS) we have developed the WSPRSONDE-8 (WS-8), a multi-channel transmitter used in ionospheric propagation measurements. This transmitter has eight channels, each covering 160 to 6 meters (and all frequencies in-between), each channel capable of transmitting at the same time. Typically some or all of these outputs are combined to feed a single multi-band antenna, using one of the TIS filter-combiner boxes. The WS-8 and earlier incarnations are currently transmitting at locations on multiple continents, in support of the HamSci ionospheric study programs.

TIS has a test site near Occidental, California where I have a number of antennas and now two WS-8 transmitters. One antenna is the EFHW-8010 from myantennas.com, and this is being driven by one WS-8 through a prototype nine-band Filter-Combiner. The frequencies are 80/40/30/20/17/15/12/10 meters.

The second WS-8 is driving two dipoles, one for 160 meters, and the other for 6 meters.

Here is the new setup:

The little boxes on top of the WS-8 stack are (left to right):

160 meter Low Pass filter — This is a L/C/L/C filter, with a few dB of peaking. I used T50-2 toroids for the inductors. The input series inductor reduces the loading on the WS8 power amplifier (the strong third harmonic sees a rather high impedance). Harmonic suppression is better than -50 dBc.
6 meter Low Pass Filter. This is a simple L/C/L TEE filter, again using toroids (T50-6).
9-Band Filter-Combiner. There is a description of this on the Turn Island Systems website.
Clock Distribution Buffer (a predecessor to the TIS-126). This is taking the output of a Bodnar GPSDO and sending it to the two WS8 units.

If you care to listen to these transmitters, here are the frequencies (“tone 0″ of the FST4W 4FSK):

WS8 #1 (EFHW-8010): 
Channel 1, frequency: 3,570,135.000000 Hz 
Channel 2, frequency: 7,040,134.926829 Hz 
Channel 3, frequency: 10,140,234.959350 Hz 
Channel 4, frequency: 14,097,134.959350 Hz 
Channel 5, frequency: 18,106,134.878049 Hz 
Channel 6, frequency: 21,096,134.878049 Hz 
Channel 7, frequency: 24,926,134.869041 Hz 
Channel 8, frequency: 28,126,134.634146 Hz 
 
WS8 #2: 
Channel 1, frequency: 1,838,134.982578 Hz
Channel 2, frequency: 50,294,534.290017 Hz

Uncategorized

Aluminum Circuit Boards

October 15, 2024 paul Leave a comment

I’ve been running a lot of designs through JLCPCB for both bare and assembled circuit boards, and I noticed that they were delivering aluminum substrate boards for quite reasonable priced. These have been used for front panels, but the thermal characteristics are great for higher-power devices such as LEDs. Interesting…

I’ve been using regular dummy loads in the testing of my WSPRSONDE (https://turnislandsystems.com/the-wsprsonde-v2-is-here/), but for easier testing I wanted a way to combine all eight channels outputs onto one 50 Ohm port. So I decided to try an aluminum substrate dummy load / attenuator / combiner:

40 dB is a good attenuation value when testing 1W signals: the combines eight outputs of the WSPRSONDE outputs the total is under 0 dBm, which is a safe level for my test equipment. So the circuit looks like this:

First, a single 40dB attenuator (actually 39 dB):

Combining eight of these stages to feed a single output:

Note that the input port-to-input port attenuation is about 80 dB, which is good since I don’t want any port interaction.

The layout looks like this:

These aluminum boards provide only a top copper layer, with no through-hole capability. This makes ground plane connectivity a challenge. Here, I am relying on the edge-mount SMA connector ground pins to provide a bridge between otherwise isolated grounds. I also have three zero-Ohm resistors bridging across the horizontal trace where the eight attenuators are connected. As it turned out these jumpers were not needed, since the ground-current flows don’t normally cross this zone.

Here is the hand-assembled board. I use four 200 Ohm 2W resistors in parallel for each 50 Ohm load:

The back of the board is pure aluminum, and would be a good place to put a heat sink. At my <10W power levels, the heat sink isn’t necessary:

The dielectric constant of the insulation layer between the copper layer and the aluminum is not specified, but rumor has it that the capacitance has been measured at 0.56 pF / square mm. This does provide enough capacitance to the aluminum substrate for it to act as a ground plane at higher RF frequencies. I decided to not make a proper RF characterization board, but instead just try my dummy load design, using skinny traces so the characteristic microstrip impedance wasn’t *too* low. Since my frequency range of interest is between 1 and 50 MHz a sloppy design is unlikely to cause problems.

And it worked quite well! Here is a NanoVNA plot, showing the attenuation and return loss from 1 to 100 MHz. This is definitely good enough for my purposes.

The thermal characteristics were quite acceptable too. Here is the board, only half-built, with close to 2W into each of the four ports. The resistors are showing about a +50 deg c temperature rise, which is well under the rating.

So, a successful result, and I now have a useful bit of test gear. It’s not a lab-grade GHz attenuator, but it certainly meets my needs.

Design Notes

A Whole Bunch Of New Designs

April 12, 2024 paul Leave a comment

I’ve been busy. Here is the latest batch of designs that have been sent out for fab. Some are simple, some more complicated, and some are updates of earlier designs. KiCad and JLCPCB sure make this stuff easy!

Look for these designs to show up on the Turn Island Systems website.

RX-888 External Clock Interface Kit:

This is a fairly simple kit that provides termination, appropriate attenuation, and DC isolation for the external clock interface to the RX888 SDR. This includes the interface card, and a replacement back-panel for the RX888. The complete kit will also include a short U.FL jumper cable and a piece of thermal foam gasket that is used to reduce the RX888 internal temperature rise.

That SMA plug shown on the right side of the board is not present when this board is installed inside the RX888, but can be soldered on when the board is used externally.

Updated Clock Distribution Buffer:

The existing five-output clock buffer has proven quite useful for distributing reference clocks in multiple-SDR receiver systems. Since I am running out of that unit, I have sent this upgraded design for fabrication and assembly. This new design now provides six outputs, each with jumper-selectable AC/DC coupling on both the signal and ground connections. The AC coupling won’t make any difference if you have RF ground-loop issues, but can really help with DC/power supply ground issues (as we often see with USB-powered devices). This new board also provides a power/clock indicator LED.

AC-Coupled SMA Adaptor:

This tiny board provides capacitor-coupling for both signal and ground. While it’s simple, it can be very useful when dealing with grounding issues. Note that the commonly-available DC-block adaptors only block the signal pin, and still leave the grounds connected.

Filter/Preamp V2:

This is my second try with the filter-preamp. I have rearranged the low-pass filter to (I hope) reduce blow-by, and have upgraded the amplifier to use a Minicircuits MAV-11BSM+ device. This should improve the intermodulation distortion and provide a good noise figure. I have also moved the amplifier to the output of the low-pass filter, to reduce potential out-of-band overload issues.

The SMA jack and plug positions provide the option for the in-line jack/plug and right-angle jack/jack configurations.

I may eventually replace the MAV-11 amplifier with a high-speed current-feedback amplifier, which should result in excellent distortion numbers, reasonable noise figure, and lower power consumption. But for now the current design should be a good performer, and I want to verify the performance of the new filter arrangement.

Design Notes

Swiss Army Knife vs Can Opener?

January 13, 2024 paul Leave a comment

In my last entry I described a feature and option-laden filter/preamp board, comparing it to a Swiss Army Knife. Great, but sometimes you just want a can opener…

After some serious discussion about how such a design would be used, and the typical requirements, I’ve decided that a major simplification is in order. (And, of course, a new feature)

The consensus was that there was no need for bypassable filters, so the stages are now:

18 dB shelf filter (-9 dB @ 6 MHz)
17 dB preamp
30 MHz elliptic low-pass filter (or 60 MHz as a different assembly)

This board is powered by an external +12V DC supply, and draws about 20 mA. There is an active-filter stage that should clean up most supply noise. The new feature is simple low-loss reverse-polarity protection for the power input.

Here is the (simulated) gain of this unit:

And the schematic:

Actually, the two filters can be bypassed at assembly time as a load-option. Perhaps this will let this board also be used as an amplifier-only design.

I have sent this design out for PCB fabrication and ordered the components — I can’t wait to see how it all comes together!

Beacon Blaster, Design Notes

A Combined Multi-Section Filter and Preamp

January 7, 2024 paul Leave a comment

Many SDR receivers, especially those designed for full HF-spectrum use, have needed various filters and preamps between the antenna and the SDR. High-power AM and FM broadcast stations can overload the receiver front-end, so these are filtered out. The characteristics of atmospheric background noise are such that a “shelf” (or “shelving”) filter which attenuates the lower-frequency range can also be useful in optimizing the SDR dynamic range.

Finally, the SDR sensitivity / input noise-floor at the higher end of the spectrum can be marginal. Add the inherent losses of all these filters and a low-noise, high dynamic-range preamp becomes very useful.

So I am designing the “Swiss Army Knife” of SDR front-end modules:

No, there aren’t four SMA connectors, only one pair will be installed, giving an “in-line” version, and a “right-angle” version that can fit on a small die-cast enclosure. There are four sections to this board:

AM Broadcast Band filter
15dB Shelf filter
13db gain preamplifier
FM Broadcast Band filter

Each of the filters can be bypassed using pin headers and jumpers. The amplifier components can be omitted during assembly and a jumper used to bypass this section.

The FM band filter allows for 6-meter reception, but many SDR receivers are clocked at 66 MHz or so, and require a lower corner frequency on the low-pass filter. This filter can optionally be built with a corner frequency at 30 MHz.

Amplifier:

Amplifier simulated with 2N5109

The amplifier is a simple wide-band zero-inductor design (found in the November 1984 edition of Ham Radio magazine, page 100: https://www.worldradiohistory.com/Archive-DX/Ham%20Radio/80s/Ham-Radio-198411.pdf). It’s become difficult to find the classic 2N5109 transistor, so I am using the more modern BFU90Q, which has similar performance. But I am simulating with the 2N5109 because that’s the model I could find.

The amplifier draws about 20 mA from a +12V power source, and my board includes a simple but effective filter that eliminates most noise and ripple on the DC source.

Here are the three filter sections:

From left to right the first filter is the AM broadcast filter,. This starts attenuating below 3 MHz, and is at least -50dB down at the top of the AM band.

Next is the 15 dB shelving filter. In a previous filter I had two 10 dB shelf filters in series, with each of them bypassable. This limited-size board requires a compromise, so there is just a single filter section. The inclusion of the AM BCB filter should mitigate the effects of the simpler shelf filter configuration.

The right-hand filter is a elliptic low-pass design, that cuts off at 60 MHz. The attenuation in the FM broadcast band, and above, is approximately 80 dB. This is also an excellent anti-aliasing filter for the SDR.

With the shelf filter bypassed the filter response from 3 to 60 MHz is essentially flat:

With all filters and amplifier enabled, the overall gain at 60 MHz is about +10 dB. The input (blue trace) is -60 dBm, and the output (green trace) is -50dBm :

And here is the schematic for the whole thing:

The simulated performance shown above should be reasonably close to the actual thing. At these frequencies the inductors are the largest source of simulation errors, but I am using Coilcraft inductors and their provided simulation models. These have proven to be remarkably accurate in previous designs.

But this is a tight little board, and layout parasitics may be an issue. You will see that the (unshielded solenoid) inductors have been oriented to reduce coupling, and (which you can’t see) the ground planes have been relieved underneath the most sensitive tuned circuits. I do hope that the amplifier won’t turn into an oscillator, but the design and layout do give me some optimism. I plan to have a working prototype in about a month.

Like the Swiss Army Knife, this design may not be the best tool for every job, but it should be useful in SDR receiver systems.

Acknowledgement: I would like to thank Clint Turner / KA7OEI for his invaluable advice on filters, amplifiers, and optimizing SDR performance. His excellent blog: https://ka7oei.blogspot.com/2020/08/revisiting-limited-attenuation-high.html

Beacon Blaster, FST4W

HamSCI, AGU, WSPRSONDE-8

December 16, 2023 paul Leave a comment

Good friend Rob Robinett (developer of wsprdaemon), along with leaders of the HamSCI Citizen Science Project, are at the American Geophysical Union (AGU) annual meeting in San Francisco, showing how WSPR is being used to study the behavior of the ionosphere. One of the posters being exhibited features the WSPRSONDE-8 multi-channel transmitter currently in development at Turn Island Systems:

See more WSPRSONDE-8 information at TurnIslandSystems.com

Beacon Blaster, FST4W, WSPR

Rant: WSPR and FST4W Timing — Why???

October 10, 2023 paul Leave a comment

In this post I complain about the symbol-rate timing of the various FST4W rates. I am adding WSPR and other FST4W rates to the Beacon Blaster ( which currently generates FST4W-120, see https://turnislandsystems.com/ for details) and am getting very annoyed with the timing variations.

So let’s start with WSPR. The symbol rate is approximately 1.4648 baud (4FSK symbols per second), or exactly 12,000 Hz / 8192. The FSK shift is the reciprocal of this, or about 0.6827 Hz.

OK, I have no complaints about this, the numbers are easy to work with and “8192″ is a nice power of two. This makes it easy to use a timer-interrupt to run a digital filter at (say) 64x the symbol rate. But really, any values for that 12,000 / 8192 fraction would be workable if WSPR were the only concern.

Now, look at the FST4W rates (from the wsjtx source code):

 if(mode=="FST4" or mode=="FST4W") { //FST4, FST4W
   if(trPeriod==15) txt=1.0 + 160*720/12000.0;
   if(trPeriod==30) txt=1.0 + 160*1680/12000.0;
   if(trPeriod==60) txt=1.0 + 160*3888/12000.0;
   if(trPeriod==120) txt=1.0 + 160*8200/12000.0;
   if(trPeriod==300) txt=1.0 + 160*21504/12000.0;
   if(trPeriod==900) txt=1.0 + 160*66560/12000.0;
   if(trPeriod==1800) txt=1.0 + 160*134400/12000.0;
 }

I assume “txt” means “transmit time”. “160″ is the number of symbols in a message. So the symbol rates are:

FST4-15: 12,000 / 720
FST4-30: 12,000 / 1,680
FST4-60: 12,000 / 3,888
FST4W-120: 12,000 / 8,200
FST4W-300: 12,000 / 21,504
FST4W-900: 12,000 / 66,530
FST4W-1800: 12,000 / 134,400

For the four FST4W rates, the divisors are 8200, 21504, 66540, and 134400. The only common factor is 2. And at that common rate of 12,000 / 2 there is no hope of programming the clock-generator chips, especially since the different symbol times drift past each other.

Since the symbol-rate accuracy is critical to the kind of propagation measurements we are using FST4W for, I am stuck running different timer interrupt rates for each mode and speed, and only running one mode at a time. Why didn’t they use something like this instead?

CXC4-15: 12,000 / 512
CXC4-30: 12,000 / 1,024
CXC4-60: 12,000 / 4,096
CXC4W-120: 12,000 / 8,192
CXC4W-300: 12,000 / 20.480
CXC4W-900: 12,000 / 61.440
CXC4W-1800: 12,000 / 122,880

Sure, in most modes the gaps between transmissions become a bit larger, but if this is an issue then pick some other numbers. But at least have them related by a larger common factor!

*END OF RANT*

Design Notes, WSPR

More Stuff at Turn Island Systems

July 20, 2023 paul Leave a comment

I’ve actually been quite busy with Ham projects, and unfortunately have been shamefully neglecting my blogging duties. So, here are a couple of posts at the Turn Island Systems site.

Design Notes, WSPR

Latest Batch of Boards

May 4, 2023 paul Leave a comment

I’ve just received another batch of QDX External Reference Interface boards. More info at Turn Island Systems

Also, I just sent out boards for a four-channel RF Combiner / Splitter, which takes four 1W (or lower) signals (80, 40, 20, and 10 meters), and combined them into a single output for feeding a multiband antenna. These provide at least 20dB port-port isolation, and less than 2dB loss. If all works as planned, these will be used for Ionospheric research.

And finally, here’s a 1W (approximately) power amplifier, using 74ACT04 digital integrated circuits to drive a wideband transformer. This takes a 3.3V digital input between 3 and 30 MHz, and produces a 1W square wave output. External filtering will be required, and the combiner shown above should provide adequate filtering.

There are “better” ways to make a 1W amplifier, but this is quick and easy, and requires no tuned circuits. I can’t wait to test these designs!

WB6CXC