A Line-Synchronous Noise Blanker

About Line Noise:

One of the banes of the VLF/LF/MF listener is powerline noise.  Down at these frequencies (several MHz and below) the energies of various electronic switching devices (such as light dimmers, motor controllers, fluorescent lights, etc.) are very strong.  There are several reasons for this:  Harmonics tend to decrease in energy with increasing frequency, and at these low frequencies, their energy hasn't dropped off by much.  Another reason is that the filters built into devices such as light dimmers just aren't effective below several hundred kilohertz owing to the small effect of the capacitive and inductive reactance at these frequencies.  Finally, these devices are typically connected directly to the house wiring and are therefore conducted into equipment and radiated on inside wiring and powerlines.

Efforts to rid yourself of this type of noise may be hard-fought battles, particularly at LF and below:  A filter that works effectively at LF frequencies is likely to be too large to fit within the fixture or enclosure of the device for which it is intended.

There are several other ways to minimize the impact of powerline-related noise on your listening:

Assuming that you have tried all of these but you need more help, then a noise blanker may be for you.

Most of today's radios contain something referred to as a "noise blanker."  Those who have used these noise blankers also know that most of them have only limited effect on powerline-type noise and many of them will also cause large amounts of intermodulation distortion in their operation, obscuring weaker signals.  A few noise blankers (such as the one in the Drake TR-7/R-7 or R-4 lines) do work well, but these blankers (as well as others) are affected by nearby strong signals and furthermore, they don't usually work on but the first "layer" of line noise (more on this later.)

A Brief Analysis of Line Noise:
From top to bottom - Triac noise in a 12 KHz, 2.2 KHz, and 300 Hz filter
Figure 1:
From top to bottom:  Dimmer noise pulses in a 12 kHz, 2.2 kHz, and a 300 Hz filter.  Note the widening of the pulses with progressively narrower filters.  (Division are 2 ms)

Line noise typically occurs on both sides of the power line sinusoid -  that is, it produces pulses at twice the line frequency (we'll assume throughout this page that we are talking about the 60 Hz U.S. power system) at 120 Hz but even so, this noise often has strong 60 Hz components as well.  The result of this is that powerline-related noise produces energy every 60 Hz across the spectrum.  To further complicate matters, AC power is distributed by the utility in three phases, each being 120 degrees from the other phase:  This can seem to "multiply" the number of noise pulses that occur during each cycle of the powerline's waveform.

The noise is produced by the abrupt switching of some device (i.e. a triac in a light dimmer) and this fast switching produces noise pulses extremely rich in harmonics.  The timing of this pulse with respect to the "beginning" of the sinusoid (let's say that this is where it crosses zero on the positive-going portion) can vary, depending upon when the offending device triggers.  This can also be offset in timing if, say, the device is operated on a different phase of a three-phase power system, for example.  This fast switching produces a pulse that is very narrow (a few tens of microseconds or less by the time it reaches your antenna.)

What about filters?  While it is true that with a narrower filter less noise energy will be intercepted, there is another problem:  Narrower filters tend to stretch noise pulses.  The three images to the right demonstrate this:  These images are from a line-triggered oscilloscope connected to the audio output of a Drake TR-7.  Three filter settings are used (top to bottom) with the first one being a 12 kHz filter (only the first IF filters are used) followed by a 2.2 kHz SSB filter and a 300 Hz CW filter.

With the 12 kHz filter (the top image) the vast majority of the noise pulse energy occurs within 1/8 of a division (about 250 microseconds) with a bit of ringing extending through 1 millisecond.  Contrast this with the middle image depicting the same pulse, but through a 2.4 kHz SSB filter.  Notice that the pulses now drag out to longer than 1 millisecond, about 4 or 5 times longer than before.  The bottom image shows the same pulse through a 300 Hz CW filter:  The pulses are actually beginning to run together.
It is true that most receiver noise blankers are placed earlier in the IF of a receiver - before the main bandwidth-determining filters, but even these receivers do have (wider) bandpass filter that tend to stretch out the pulses somewhat.

All is not lost:  With powerline noise, we have an advantage over random noise in that we know very precisely its repetition rate - 60 or 120 Hz (in the U.S. and a few other countries, at least.)  Furthermore, this noise source is most likely powered from the same power grid as your equipment (even if it isn't on the same phase as you) and thus you have a ready-made reference for the noise-pulse frequency.  What is not known is the precise position and width of the offending pulse.

(By the way, the noise blanker described reduced the above "light dimmer noise" by well over 35 db.)

A few comments on DSP noise reduction:

At first glance, one might think that a modern DSP-based noise reduction circuit would go a long way toward eliminating line noise - but this isn't necessarily the case.

Take Audio-based DSP, for example:  This type of DSP is that which is most commonly found on lower and moderately-priced amateur rigs.  There is a problem, however:  As seen in the example, the noise pulse is often badly distorted and "smeared" by the radio's filters even before they get to the DSP circuit - a fact that makes it extremely difficult to get rid of the noise pulse.

Another type of DSP commonly found is IF-based DSP.  Rather than being placed in the audio path, this DSP circuit is placed in an Intermediate Frequency (IF) stage.  On most radios with IF DSP, this is actually at or just above the audio frequency range, typically in the 10-20 kHz range.  Placing the last IF in this low frequency range reduces the computational power required to process the signals - but still is high enough that the IF bandwidth coming into the DSP filter can be wide enough to prevent excessive distortion of the pulse.  If designed correctly, these types of systems can be very effective in removing power line noise - but in very serious cases, the dynamic range of the DSP hardware and software can be pushed to its limits.

In short:  With audio-based DSP, you will still likely benefit from having an antenna-based noise blanker.  Even if you do have an IF-based noise blanker, an antenna-based blanker may still provide some benefit.

Deleting the offending noise pulse(s)...
Oscillographs showing the holes punched in audio by the blanker.
Figure 2:
The effects of the blanker on a received signal.  The top image shows the "holes" punched in the received signal with one blanking channel, the bottom image shows two active blanking channels.  The blanking widths are exaggerated for clarity.

As can be observed above, simply muting the pulses at audio is not a very good solution due to the stretching of the pulses by the receiver's filters.  If this fact weren't the major concern, there is the problem of the receiver's AGC action attenuating the desired signal:  The noise will set the AGC to levels based on the strongest signal being received (even noise) - but if the desired signal is, say, 15 db below the noise peaks, the AGC will bury that signal 15 db in the audio... amongst the stretched noise pulses...

What we need to do is blank the RF during the noise pulse, starting just the blanking action before it gets to any  of the receiver's filters:  Since the noise pulse itself is (usually) of very short duration, we don't need to blank it for very long.  The most obvious place to do this is at the antenna connection and this may be done several ways.  One type of blanking gate described here uses PIN diodes in an attenuator arrangement and this provides attenuation that is related to the current applied to its diodes and is fairly "linear" even when only partial attenuation is occurring, which is important if you want to minimize the amount of intermodulation that the noise blanker will cause.  The use of MPN3404 PIN diodes is intentional here:  While this circuit will work with ordinary 1N914/1N4148-type diodes, you can expect performance to be worse as garden-variety diodes produce more nonlinarity when "partially" turned off.  As mentioned below, another type of blanking gate that reportedly works somewhat better at HF uses a diode-ring mixer.

Most important to intermodulation reduction is careful control of the slew rate of the blanking pulse:  If one were to simply employ a switch that turned the antenna on and off, the switch itself would cause significant harmonics and the blanker itself would cause what sounds like line noise.  If one (relatively) slowly turns the attenuation on and off during the blanking, you can greatly reduce the possibility of intermodulation and reduce the bandwidth and intensity of the "blanking sidebands" that are necessarily created during blanking (we are amplitude modulating our received signals with the blanking pulse, remember...)

As it turns out, while there is typically one major noise pulse causing most of the buzzing, but there may be several other noise pulses occurring during different times -
that is, if you blank just one set of the offending pulses, there may be another set that you are now able to hear that also obscures the desired signal and for this reason, you may wish to have the ability to set up individual blanking pulses for the various "layers" of noise.

Design philosophy:

There are several schools of thought when designing noise blankers:  "Normal" noise blankers (such as those found in receivers) have fast-acting blanking gates - a necessity because these blankers are typically "asynchronous," operating by first detecting the rising edge of the pulse, and then blanking it before much of its energy finds its way into the receiver.  Clearly, if you are going to blank a pulse that you have just detected, you need to act quickly.

With a synchronous noise blanker such as the one described here, you have the luxury of knowing precisely when a pulse is going to appear and because of this, you could use a fast-acting blanking gate or you could use a slower one.  What's the difference, then?  Which one is better?

The answer isn't necessarily a simple one and each has its own set of trade-offs.  About fast-acting blanking gates:

About "shaped pulse" blanking gates like this one: In either scheme, blanking gate design is very important.  Whether the design is for a "fast acting" or a "shaped pulse" blanker, the device that actually does the blanking of the RF should be resistant to intermod.  This is arguably more important when a "shaped pulse" blanker is used, as there is a significant period of the blanking pulse where some manner of "variable attenuation" is desired - a point during which the PIN diodes (in this design) may be prone to generating IMD in the presence of strong signals.

The use of a doubly-balanced diode-ring mixer (used as a variable attenuator) can be used to mitigate this problem as their intrinsic balance reduces the likelihood if intermodulation distortion.  The problem with using a diode ring mixer instead of the PIN diodes is that their cost is much higher, and they have a very definite low-end frequency response which may make them unsuitable if very low LF or VLF reception is desired unless the proper device is used.  If you only intend to use this blanker on, say 160 meters and/or the AM broadcast band, you may wish to consider using a DBM - see the details here.

Description of the circuit:
Figure 3:
The schematic of the unit:  Click on the image for a larger version.

What we need is a device that will produce a blanking pulse that has two adjustable parameters:  Position and width.  The position of this pulse is relative (in this case) to the zero crossings of the 60 Hz line frequency (which occur at a rate of 120 Hz) and the width is, well, width...

Schematic versions:

Click on the diagram for the schematic.  When printing, select LANDSCAPE mode so that it will best-fit on the page.
A view of the proto board containing the pulse and blanking generators
Figure 4:
A view of the proto board containing the pulse and blanking generators.

(Click on the picture for  a larger view)

The blanker is powered by an 18 volt AC power cube and the voltage is filtered and isolated (at RF frequencies) with a bifilar-wound inductor removed from a defunct computer switching power supply.  This inductor has several millihenries of self-inductance and it prevents noise that might be conducted from the power cube (from the power line) into the receiver itself.

This power is then half-wave rectified and filtered and applied first to a 7818 regulator for a stable 18 volt DC supply (for the blanking drivers and for powering the receive antenna on the coaxial cable) and then filtered and regulated to 5 volts for the timing logic:  The capacitor and series resistor across the diode are used to suppress any "hash" that might be produced by its nonlinearities.

A portion of the "pre-rectified AC" is extracted, filtered, AC coupled (to remove any DC bias) and applied to U2C which is wired as a comparator to produce a 60 Hz 18 volt square wave.  This square wave is resistively divided down to approximately 5 volts and applied to U3B, a 74HC86 which buffers the 60 Hz square wave and applies it to U3A which is wired as an edge-detector.  The resistor/capacitor combination on its input cause this section to produce a narrow pulse on each transition of the 60 Hz square wave, thus producing a 120 Hz pulse train which is buffered by U3D and is made available for the blanking-pulse generators.  The remaining section (U3C) is simply tied to U3D so that it's input isn't floating and is therefore unused. (Yes, I should have tied it to to either the ground or the 5 volt supply, but I was lazy...)

The 120 Hz pulse train is applied to U1B, a 74HC123, a dual one-shot timer and is the "position" timer and has a period that is adjustable from a few microseconds to the entire duration of the 120 Hz pulse repetition interval to allow blanking to occur at any point in the period of the pulse train.  The output of this is first section is sent to U1A.

Since U1A is edge-triggered, it responds only to the rising edge of the pulse from U1B - the end of the timing period.  The section of U1A operates exactly like the U1B section, except that its timing range is restricted to approximately one-third of the interval of the 120 Hz pulse train.  This section functions as the "width" generator for the blanking pulse.

The finished blanking pulse appears on pin 4 of U1A and it is at this point that this blanker generator may be "diode-ORed" with the outputs of other pulse generators.  If it is desired that more than one blanking pulse is needed (at least two are recommended) then the next section would be connected at this point.  For enabling/disabling a blanking generator, a switch (which may be mounted on the rear of the "width" control) is placed on pin 3 of U1A to enable/disable the timer.  For a single section blanker (or for the first of several blanking channels) this may be simply tied to +5V.

U2D, an Op-Amp section, is used as a comparator and it takes the diode-ORed input from the pulse generator(s) (using a 10k pullup resistor) and generates inverted 18 volt blanking pulses.  This pulse is then shaped by U2A which is wired as a 3rd order lowpass filter.  A 100k resistor limits the maximum output to prevent clipping of the amplifier at the 18 volt rail and thus prevents distortion (and the ensuing harmonics) from the blanking pulse.  A switchable 0.018 microfarad capacitor allows selection of a lower slew rate to allow the blanker to be used at lower LF frequencies (more on this later.)

The filtered output of the blanking generator is applied to the PIN diodes through a resistor (for current limiting) and across a capacitor and through an inductor (for RF decoupling.)  When the voltage is high, current flows through the two PIN diodes, turning them on.  When the voltage is off, the diodes do not conduct and high isolation is obtained.

Surrounding the PIN diode network are other inductors and capacitors that allow for power passing and insertion to allow active antennas to be powered by the blanker's power supply and/or the receiver.

Important Notes relating to previous versions of the schematic:

The PIN diode attenuator section of the noise blanker
Figure 5:
The PIN diode attenuator section of the noise blanker.  (Note that there are 2 extra coils shown - these were originally used in the prototype, but bypassed - but not removed - in the "final" version.)

(Click on the picture for a larger view)

Using a Doubly-Balanced Diode-Ring Mixer ("DBM") as a blanking gate in lieu of PIN diodes:

If you do not intend to use the noise blanker at very low frequencies, it is possible to replace the PIN diode blanking gate with a diode-ring doubly-balanced mixer of either homebrew or commercial manufacture.  An example if where you may wish to do this is if you were to use this blanker exclusively on, say, 160 meters.

There are many suitable diode-ring mixers that may be used - the selection depending upon the desired frequency range.  An inexpensive and readily-available diode-ring mixer is the venerable SBL-1 by MiniCircuits.  This mixer is rated for use from 1 to 500 MHz (although it will work just fine across the entire AM broadcast band) and is typically sold for well under $10.00 U.S. and may be found for sale using an internet search.

Alternatively, you can construct your very own diode-ring mixer fairly easily using two trifilar-wound ferrite toroids and four matched diodes.  Details how diode-ring mixers may be constructed can be found in the ARRL Radio Amateur's Handbook as well as by doing some internet searches.  A few salient points on constructing a diode-ring mixer, however:

How it works:

First of all:  Yes, the "output" of the mixer is the "LO" port.  Because the mixer is being used as a variable attenuator, it isn't wired exactly the same way as you would wire a mixer.

When there is no current on the "IF" port of the DBM (this is the port that has direct DC connection to the four internal diodes) all of the diodes are turned OFF.  As the current increases, however, the diode starts to conduct, allowing RF to flow through.  Owing to the symmetric nature of the circuit, most of the diode's nonlinearity is canceled out internally, reducing the potential for intermodulation distortion.

The signal from the blanking gate goes from about 0 volts (if you use an LM324 or another op-amp that can swing to its negative supply rail) to about 15 volts. The 3.3k resistor limits the maximum current to about 5 mA and, importantly, the 51 ohm resistor (a 47 ohm resistor could be used) and the 0.01 uF capacitor set the terminating impedance at the IF port (at RF frequencies) at 51 ohms for best performance.

Note:  Because the attenuation of the DBM (and the PIN diode, for that matter) is somewhat proportional to the diode current (over portions of the curve) it is imperative that the diode current go to zero in order to maximize the amount of attenuation during the "off" portion of the blanking pulse.  In the case of the DBM, it also must not go negative, or else the signal will re-appear, but in the opposite phase.  It is also important that the diode current just reach zero at the bottom of the blanking pulse in order to minimize the intensity of the sidebands imparted on received signals by the blanking action:  Built according to the schematic, you will be pretty close to optimum, but if you are of the sort to fiddle and tinker...

A few comments:

Construction notes:

The prototype unit (shown in the pictures) is a three-channel synchronous noise blanker.  Although two channels usually suffice, there are (rare) instances where all three are required.  
Front panel of the Synchronous Noise Blanker
Figure 6:
Front panel view of the Synchronous noise blanker.

(Click on the picture for a larger view)

The front panel shows six potentiometers:  One for position, and another for width for each of the three channels.  Channels 2 and 3 (the middle and the one on the right, respectively) have switches mounted the rear of their width control pots and these switches are used to disable their respective channels.  They are placed on the width control so that the channel may be disabled without disturbing the pulse position settings.  Each of these pots is wired such that counterclockwise rotation corresponds to the shorter timing interval for the associated timer (i.e. "earlier" blanking pulse for the position control, or "narrower" blanking pulse for the width control.)
Rear view of the front panel showing the wiring of the potentiometers
Figure 7:
Rear view of the front panel showing the wiring of the potentiometers.

(Click on the picture for a larger view)

The PIN diode attenuator is constructed on a piece of circuit board material and is mounted using the antenna connectors (RCA connectors, in this case.)  This provides a good "RF" ground, mechanical rigidity, and minimizes extraneous RF pickup or radiation.  1.5 and 2.2 microfarad ceramic capacitors (precise values are not critical) are used for DC decoupling and these values are sufficiently large enough to allow for minimal attenuation at VLF frequencies.  If such large-value ceramic capacitors are not available, electrolytic capacitors may be used (observe polarity!)  If response at such low frequencies (below 10 kHz) is not needed, lower-value capacitors may be used.  It is recommended that electrolytic capacitors (if used) are paralleled with some 0.1 uF ceramic capacitors to minimize losses that electrolytics often exhibit at higher frequencies.

  With my receiver (a Drake TR-7) I have an "LF Interface box" that puts power on the LF antenna input connector to power an active antenna.  Diodes are employed to steer power from the receiver to the antenna and from the noise blanker.  A switch is used to enable/disable this power - useful if using a passive antenna.
A view of the rear panel of the synchronous noise blanker
Figure 8:
A view of the rear panel of the synchronous noise blanker.  A banana jack (left) is for monitoring the blanking pulse with an oscilloscope.

(Click on the picture for a larger view)

It is recommended that the bodies of the potentiometers and switches be grounded to reduce "hand effects" and to protect the CMOS circuitry from static discharges.   It is also strongly recommended that the timing capacitors on U1A and U1B (the 0.068 and 0.022 uF units) be (relatively) temperature-stable devices such as Mylar or polyester:  Ceramic disk types in this capacitance range are not stable with temperature and you will be plagued with temperature-related drift of the position and width adjustments if you use them.

Finally, it is necessary to choose op amps that are capable of swinging all the way to the negative power supply rail.  The LM324 specified in the schematic is capable of doing this - but many op amps are not.  If the op amp that you use cannot go all the way to the negative rail (in this case, ground) it is possible that the PIN diodes in the blanker gate can never be turned completely off and blanking performance will suffer.  In other words, just use the LM324!

Obtaining MPN3404 PIN diodes:
Several people have reported difficulty in finding MPN3404 PIN diodes.  While not ubiquitous, there are a number of places where they may be obtained in small quantity:

Operation of the unit:

Most of the time, operation of just one blanking channel will eliminate most of the noise.  Even if two (or three) blanking channels are ultimately required, you start out with just one.  The noise blanker is set up thusly:

  1. Turn off the receiver's own noise blanker.  Since the idea is to null the noise, you don't want the radio's blanker to make it harder to hear the noise while you are doing your adjustments.
  2. Make sure that only one of the synchronous noise blanker's channels is on.
  3. Set the receiver's bandwidth to a fairly wide setting (usually an SSB or AM filter.)  This will allow the noise to be heard more clearly as you null it out.  If your receiver has it, you might find it easier to fine-tune adjustments with the receiver set to AM as it will be easier to hear the "buzz" go away as you null it out.
  4. Set the receiver's AGC to SLOW, or turn it OFF (adjusting the RF gain to set the signal to a comfortable level.)  This will allow easier nulling of the noise.
  5. Set the width control to to about mid-rotation.
  6. Slowly, carefully rotate the position control.  A definite decrease in the powerline noise (the "buzz") should be noted at some position - unless you just happen to have two noise sources of different timing and similar strength.  Keep an eye on your S-meter as you do this, as the AGC might simply compensate for reduction in the amplitude of one noise source and make another one more audible!
  7. Adjust the width control to narrow the pulse slightly.  Readjust the position control to re-null the noise.  Repeat the previous steps.  If you adjust the width control too narrow, a deep null cannot be obtained while adjusting the width control too wide can result in needless incidental AM modulation of desired signals (more on this later...) - but remember that some pulses (especially those from switching power supplies) may simply be "wide" by nature.
  8. At this point it may become obvious that there is more than one noise pulse and a second (or third!) blanker channel may be activated.  It is adjusted in the same way as the first.  It may be advantageous to "touch up" the first (or second) channel a time or two to achieve maximum noise reduction.  Remember:  Set the "width" control to the minimum required to achieve the best blanking!
  9. If you are trying to listen to VLF signals (i.e. those below 30 kHz) it is possible that the switching of the PIN diodes themselves may cause some audible harmonics at the switching rate.  At (and below) these frequencies, one would set the slew rate switch to the slow position.  Note that changing the switch setting will usually necessitate readjustment of all width and position controls on all enabled channels.
  10. The AGC, filter, mode, and noise blanker settings on the receiver may be returned to their normal positions.
After using the blanker a few times you should get the "feel" of how it operates and you should be able to quickly set up the position and width controls for best noise elimination.

Operational notes and comments:

Here are a few comments about a synchronous noise blanker - what the blanker will and will not do:

The noise blanker in action:

I have received several emails over the years about the operation of the noise blanker and how effective it may be.  The latter question is very difficult to answer, as it all depends on the qualities of the QRM with which you are being afflicted.  To demonstrate some of the various properties, here are some audio clips with analyses:

Again, remember that not all noise sources produce "blankable" QRM and even when they do, you can hear some of the artifacts that the blanking action creates in the received signals.  Also note that at different frequencies, you may hear different noise sources which require "re-tuning" of the blanker to optimize noise suppression.  Finally, note that this blanker was designed primarily for operation at LF and low MF frequencies (e.g. longwave) where the noise pulses tend to be quite distinct and quite nearby:  It has been observed that similarly-created noise at HF (160 meters, 80 meters and particularly above) is often "smeared" and a bit less-distinct, making it much more difficult to blank!

Other pages at this site:

The CT MedFER Beacon page - This page describes a PSK31 MedFER beacon.

The CT LowFER Beacon Archive Some pictures/info about the "CT" LowFER beacon of the late 1980's.  (Includes QSLs and sounds from some other beacons of the time.)

"QRSS and you..." - Using absurdly low-speed CW for "communications"

Using your computer to ambush unsuspecting NDBs - A brief description of how Spectran may be used when trying to receive NDB

Any comments or questions?  Send an email!

This page copyright 1999-2015 by Clint Turner, KA7OEI and last updated on 20150415

Since 12/2010: