Designing a Noise Gate

[hex_void]

Here's the idea:
About three years ago I made a sort of bet with a friend, also a guitarist, that I'll build my own distortion pedal since I didn't like anything that was commercially available at a reasonable price and my budget was too low to buy several pedals (about a truckload) to get the tones I wanted. I also dislike digital effects processors (they're a lot of fun but can't tweak them that well) and I wanted a lot more tonal versatility than any analog distortion so after a few versions I achieved what I was aiming for. Now I'm experimenting with noise reduction, both within the distortion pedal itself (less parts in the path of the signal between the guitar and the stage with the highest gain - actually the gain control you find on all nearly all OD/distortion pedals - less noise at the output, while keeping the same tone) as well as a noise gate which will be on the same board as the distortion, everything being powered by 20Vdc.

I know the basics of a noise gate and have a few schematics I found on the web but there are two drawbacks: They either use a hard to find (for me) voltage controlled amplifier or a JFET as a voltage controlled resistor which has different characteristics than any JFET I can get my hands on (kind of stuck with the BF245C) so it won't work properly - I tried it.

Since the tone controls affect the response of the noise gate I shall sandwich the latter between the distortion and tone controls (in my case - a 6 band EQ with mid scoop switch).

Adjusting the volume using a JFET is fairly easy (0V gate-source voltage turns it on - drain-source resistance decreases to a minimum, -7V gate-source voltage turns it off - high drain-source resistance, but it should not go below that or it gets fried and anything between the two... well you get it).

The hard part is getting the noise gate to let the signal through when it should get through - that's the control circuit's job.

This is my idea of a control circuit: input from distortion ---> 2nd order low-pass filter (reduces some of the high frequency noise) ---> opamp gain stage with adjustable gain (sets the sensitivity) ---> precision rectifier + RC low pass filter to smooth out the rectified signal ---> schmitt trigger with adjustable hysteresis ---> JFET

The JFET attenuator will be sandwiched (damn... I'm starting to get hungry )  between two opamp buffers.

I might also use some of the gated signal (after the noise gate) and add it to the input signal... it may improve things a bit.

Any ideas/suggestions on what else I could do?

I prototyped one on a solderless breadboard a few months back and it worked fairly well taking into account that it was placed after the EQ but I wasn't totally satisfied with it when I started randomly goofing around on the fretboard - it couldn't decide if I was actually playing or not. 

[Mark Hammer]

Here's an idea for the more industrious and diligent of you: a hum-gate for single-coil users.

"Noise gates" work under the principle that unwanted signal is distributed across the entire spectrum.  So, a sidechain/rectifier detects the amplitude of the entire signal (20hz-20khz), and should that overall amplitude fall below some threshold, the entire signal is drastically attenuated.  Because all portions of the spectrum are equally affected, there is traditionally legitimate concern about the turn-on and turn-off affecting the initial attack and/or the natural decay of the signal.

For most guitar players, the brunt of the unwanted signal at the point of initial plug-in from the guitar itself, is likely to be 60hz hum, rather than hiss or other sources (which are more likely to arise from subsequent processing by pedals, especially high-gain devices).  So, when you think about it,  any nise-control to be inserted as the first thing a guitar sees, would ideally focus on EMI.  In other words, it would ideally be an automatic lo-cut filter, that left the bass in when the guitar signal was substantial enough to mask it, but cut the bass out when you stopped playing.  Since it would not affect the top end, it would have very little impact on initial attack, and modest effect on string decay (which does tend to have more bottom than top).

Let's say we had a 2-pole highpass (12db/oct cut) filter, with a 160hz corner frequency (or thereabouts, for calculation purposes).  That would allow for "mistakes" to be made in threshold setting, yet still hear much of the desired part of the guitar signal.  Bass might be trickier, admittedly.  If threshold is set right, that would  trim a big chunk of annoying hum during the quiet bits.  Again, this presumes that such a circuit is the FIRST thing you plug into.  Any noise-control circuit inserted further downstream would be looking for multiple sources of noise, not just hum.

Whatcha think?

[hex_void]

Here's an idea for the more industrious and diligent of you: a hum-gate for single-coil users.

"Noise gates" work under the principle that unwanted signal is distributed across the entire spectrum.  So, a sidechain/rectifier detects the amplitude of the entire signal (20hz-20khz), and should that overall amplitude fall below some threshold, the entire signal is drastically attenuated.  Because all portions of the spectrum are equally affected, there is traditionally legitimate concern about the turn-on and turn-off affecting the initial attack and/or the natural decay of the signal.

That's the Noise Gate I'm looking for.
I play mostly death metal / thrash metal so obviously I'm not using single coil pickups but fairly high output humbuckers (they do their job, the only two cases where they fail are pretty extreme: CRT monitor deflection coil EMI at less than 3m and 230V mains cables at less than 20cm) and the distortion I made/use is high gain, high enough so the TL082 low noise opamps' noise (along with some resistors' noise) is audible since it gets amplified along with the guitar signal (which is then hard clipped with two 1N4148 diodes... no more technical details here, everything is sorted out and works perfectly). Because the whole signal is hard clipped it gets limited to about 1.1Vpp...1.3Vpp while at the opamp output it's amplitude is at least ten times higher, but the noise is (*checks oscilloscope just to be sure*) the same, about 30mVpp with the gain cranked up to the max.

Basically even with the input shorted the circuit's noise will be amplified and you get that hissing sound - mostly higher frequency stuff (you can hear it here http://www.youtube.com/watch?v=C16zkacL4lQ&feature=player_detailpage#t=50s no advertising intended, just an example of the hissing noise) - that's why I want to use a low-pass filter for the control circuit, I don't want the gate to be triggered by the noise it should eliminate, that would make the gate completely useless. While playing it won't be noticeable because it's low compared to the useful sound.

Now, the noise is not a problem when I'm playing at home, on a 15 watt practice amp set on clean channel and usually cranked to less than 1/4 of its max volume, but it becomes annoying on higher power amps at higher volumes. If you're asking why I don't just buy a noise gate it's because it's about 50 times cheaper to build one into my distortion pedal + I can tweak it to work better with my distortion than any commercial unit will, and believe me, I've tested a few of them... not so impressed.

[Mark Hammer]

It runs contrary to the catechism here, but maybe just get yourself a ToneCore Uber-Metal.  Decent high-gain type distortion, with lots of flexibility, AND a built in noise gate!  They can be found fairly cheaply these days.  The only drawback is that they really ARE "heavy metal".  The chassis weighs a bloody ton.

[Ben N]

... and/or (for the single coils) put a bandpass filter in the sidechain, so that signal in the offending band is in effect boosted and the overall threshold can be lowered. Ergo, quicker response to note attack and less cutoff of decay.

Are other noise sources in the signal chain concentrated at particular frequencies or distributed like white or pink noise? If they are concentrated, then a customizable noise gate might offer multiple filter bands to be optimized to the end-user's rig.

As to location in the signal chain, Mark, wouldn't the best thing to feed the control input with (buffered) raw guitar signal, but put the gate itself as far down the chain as possible--in the fx loop of an amp, e.g.? The idea being, again, to set the threshold to be as sensitive as possible, but to still gate out noise from as many sources as possible.

[Mark Hammer]

That is the very strategy behind all the various noise gates with send/receive loops.  Just keep in mind that, in that scenario, they are deployed to tackle ALL forms of accumulated noise, which could include LFO ticking, BBD clock whine, hiss, hum, etc.   What you have to contend with at the end of the signal chain, is different than what is present at the start.

My own philosophy is that the gripes people have against noise-control devices stem from sticking them in one spot only.  Try to deal with noise ONLY at the start of the signal chain, and you fail to address all thse other noise sources that come from the pedalboard.  Stick it at the end and you have to be brutal in setting the threshold to take on all the various noise souorces.  Neither of those are satisfactory.

Personally, I think the optimum arrangement is a two-station one.  First get rid of any EMI/hum entering the first pedal, without affecting anything else, so that it doesn't get amplified over your pedal chain, and at the end of the chain, trim the accumulated hiss, whine, ticking with a noise filter.  Whether a single sidechain, using the envelope detected at the start of the pedal chain, is the one to use to control both noise-control circuits, well that's an empirical question that I have no answer to at the moment.

[hex_void]

It runs contrary to the catechism here, but maybe just get yourself a ToneCore Uber-Metal.  Decent high-gain type distortion, with lots of flexibility, AND a built in noise gate!  They can be found fairly cheaply these days.  The only drawback is that they really ARE "heavy metal".  The chassis weighs a bloody ton.

128 Euros (including shipping, since it's not available in my country) pretty much defeats the purpose of having my own distortion + I don't really like the tone, too dirty, mine sounds more like the EH Metal Muff, but it's way more versatile, with some settings I can get close to tube OD sound, but for high gain settings I still need a noise gate. Also, for that money I can build at least two of mine and get both lead and rhythm tones.

That is the very strategy behind all the various noise gates with send/receive loops.  Just keep in mind that, in that scenario, they are deployed to tackle ALL forms of accumulated noise, which could include LFO ticking, BBD clock whine, hiss, hum, etc.   What you have to contend with at the end of the signal chain, is different than what is present at the start.

My own philosophy is that the gripes people have against noise-control devices stem from sticking them in one spot only.  Try to deal with noise ONLY at the start of the signal chain, and you fail to address all thse other noise sources that come from the pedalboard.  Stick it at the end and you have to be brutal in setting the threshold to take on all the various noise sources.  Neither of those are satisfactory.

Personally, I think the optimum arrangement is a two-station one.  First get rid of any EMI/hum entering the first pedal, without affecting anything else, so that it doesn't get amplified over your pedal chain, and at the end of the chain, trim the accumulated hiss, whine, ticking with a noise filter.  Whether a single sidechain, using the envelope detected at the start of the pedal chain, is the one to use to control both noise-control circuits, well that's an empirical question that I have no answer to at the moment.

I agree on that. I can trigger the second gate with the control signal from the first one, that's the easy part when I'll complete my pedalboard. But I still have to get the noise gate to work properly. So, is my initial idea good or do I have to take anything else into consideration?

[Mark Hammer]

Fair enough.  Money IS a deterrent.

If you are using a FET as the control element, my sense is that it should be placed somewhat early, or midway, in the circuit, rather than at the output; the reason being that, whatever JFET you use, it will have limits on how big a signal it can handle before distorting in its own way.  So, sticking it just ahead of an output volume pot, such that it is being asked to negotiate a 3-4V P-P signal, may be asking for trouble.

From what I can make out of your hypothetical/prototype circuit, it seems like you are on the right path.  But as always, it takes a peek at a schematic to have a better sense of how it all fits together.

[Keppy]

Whether a single sidechain, using the envelope detected at the start of the pedal chain, is the one to use to control both noise-control circuits, well that's an empirical question that I have no answer to at the moment.

I'd say the answer is yes. Reason being, noise gate designs do not respond instantaneously, and affect the beginning of notes once they're clamped down in ways that are generally audible if you listen closely (in my experience at least). If you place a noise gate earlier in the chain and then a second one further down, you've doubled the lag in response unless you trigger them from the same signal.

The compromised transient response of one noise gate is one I'm willing to make. Two, not so much. As always, YMMV.

[WaveshapeIllusions]

If your noise is around 30 mV P-P, there's a pretty simple way to cut it down. Put two antiparellal diodes (like you would use for standard symmetrical clipping) in the singal path, preferrably after one of the gain stages. Something like a 1N4148 or 1N4001 would do, or really any standard silicon diode. Make sure both sides of the diodes are biased to the same voltage of course.

What that will do is limit any signal below the diode's forward voltage, so it will block noise. The downside (could be an upside) is that you will get an effect much like crossover distortion. Not too many pedals use it, the only one I know is the Boss HM2. It's an interesting sound, with the added effect of lowering noise.

With regards to a triggered noise gate like what you are working on, I've found it best to feed the sidechain with the input and gate at the output. You don't have to worry about false opening and thresholds since the clean guitar opens it. And you don't get any extra circuit noise, since it's all cut off until opened.

Another idea is to EQ the sidechain. Some filtering between the input and the rectifier can do wonders for the response. If the noise is primarily in a certain frequency band, cutting it can help ensure it won't open the gate. Also, boosting the guitar's range (or cutting everything else) can help out. You could even boost a few frequencies so that each note triggers the gate equally.

Hope this helps.

[defaced]

I had the opportunity to play a Marshall KK JCM800 amp a few weeks ago which has a noise gate in it.  It  performed very well for tightening up a high gain amp. The schematic if freely available.  It might be worth a look, I haven't dug into it, so I can't offer any other comment than just a pointer to check it out. 

[amptramp]

RCA had a circuit they called the "Magic Monitor" that they used from 1940 in my A33 console radio to the VRA141 console radio/phono unit that was designed to cut noise from the record player.  It used two tubes, either 6SQ7 and 6SK7 for the early unit or 6AV6 and 6BA6 for the later unit.  It took in an input and fed it to the output through a 1 meg resistor paralleled by a 330 pF capacitor.  The treble part of the signal went through the 6SQ7/6AV6 triode amplifier coupled to the diode of the same tube via a 180 pF capacitor.  This provided a negative DC output based on the high frequency audio on a record.  This went through two RC lowpass stages then to a reactance tube (the 6SK7/6BA6) that was capacitively coupled to the output.  The idea here was to look for the amplitude at high frequencies.  If it was low, there would be little bias on the reactance stage and it would appear fully capacitive.  If it was high, the reactance stage would decrease the effective capacitance across the output of the unit as a lowpass filter with the 1 megohm resistor from input to output.  This acted as a dynamic tone control that reduced the treble at lower volumes.

[hex_void]

Fair enough.  Money IS a deterrent.

If you are using a FET as the control element, my sense is that it should be placed somewhat early, or midway, in the circuit, rather than at the output; the reason being that, whatever JFET you use, it will have limits on how big a signal it can handle before distorting in its own way.  So, sticking it just ahead of an output volume pot, such that it is being asked to negotiate a 3-4V P-P signal, may be asking for trouble.

From what I can make out of your hypothetical/prototype circuit, it seems like you are on the right path.  But as always, it takes a peek at a schematic to have a better sense of how it all fits together.

The max input signal to the gate won't go over 1Vpp. Here's the schematic. I'll add a bandpass filter later. http://www.aronnelson.com/gallery/main.php/v/Schematics-etc/noise_gate.jpg.html

If your noise is around 30 mV P-P, there's a pretty simple way to cut it down. Put two antiparellal diodes (like you would use for standard symmetrical clipping) in the singal path, preferrably after one of the gain stages. Something like a 1N4148 or 1N4001 would do, or really any standard silicon diode. Make sure both sides of the diodes are biased to the same voltage of course.

What that will do is limit any signal below the diode's forward voltage, so it will block noise. The downside (could be an upside) is that you will get an effect much like crossover distortion. Not too many pedals use it, the only one I know is the Boss HM2. It's an interesting sound, with the added effect of lowering noise.

With regards to a triggered noise gate like what you are working on, I've found it best to feed the sidechain with the input and gate at the output. You don't have to worry about false opening and thresholds since the clean guitar opens it. And you don't get any extra circuit noise, since it's all cut off until opened.

Another idea is to EQ the sidechain. Some filtering between the input and the rectifier can do wonders for the response. If the noise is primarily in a certain frequency band, cutting it can help ensure it won't open the gate. Also, boosting the guitar's range (or cutting everything else) can help out. You could even boost a few frequencies so that each note triggers the gate equally.

Hope this helps.

I tried that, but I have my tone already set. It still sounds good but it gets a bit too dirty and I won't compromise the tone I have without the added diodes. Also, it doesn't work very well, stuff gets weird when single notes fade out - nasty crossover distortion with some pops in some cases, this goes against the reason I've built the pedal - versatility, so I'm sticking with a noise gate.

I had the opportunity to play a Marshall KK JCM800 amp a few weeks ago which has a noise gate in it.  It  performed very well for tightening up a high gain amp. The schematic if freely available.  It might be worth a look, I haven't dug into it, so I can't offer any other comment than just a pointer to check it out.  

Found it and checked it out. It uses the THAT 4301 voltage controlled amplifier. I initially wanted to use a VCA instead of a JFET but nothing was readily available.

RCA had a circuit they called the "Magic Monitor" that they used from 1940 in my A33 console radio to the VRA141 console radio/phono unit that was designed to cut noise from the record player.  It used two tubes, either 6SQ7 and 6SK7 for the early unit or 6AV6 and 6BA6 for the later unit.  It took in an input and fed it to the output through a 1 meg resistor paralleled by a 330 pF capacitor.  The treble part of the signal went through the 6SQ7/6AV6 triode amplifier coupled to the diode of the same tube via a 180 pF capacitor.  This provided a negative DC output based on the high frequency audio on a record.  This went through two RC lowpass stages then to a reactance tube (the 6SK7/6BA6) that was capacitively coupled to the output.  The idea here was to look for the amplitude at high frequencies.  If it was low, there would be little bias on the reactance stage and it would appear fully capacitive.  If it was high, the reactance stage would decrease the effective capacitance across the output of the unit as a lowpass filter with the 1 megohm resistor from input to output.  This acted as a dynamic tone control that reduced the treble at lower volumes.

Hmmm... I think this could be implemented with RL filters using gyrators with the inductance controlled by a JFET (although linearity would be a bit of an issue) and a bunch of opamps to make sure the high frequencies are reduced when input level is lower. That would be an amplitude-controlled low-pass filter - lower the amplitude and it decreases the filter's frequency but that would also change the tone which is exactly what I don't want.

[Thecomedian]

Here's an idea for the more industrious and diligent of you: a hum-gate for single-coil users.

"Noise gates" work under the principle that unwanted signal is distributed across the entire spectrum.  So, a sidechain/rectifier detects the amplitude of the entire signal (20hz-20khz), and should that overall amplitude fall below some threshold, the entire signal is drastically attenuated.  Because all portions of the spectrum are equally affected, there is traditionally legitimate concern about the turn-on and turn-off affecting the initial attack and/or the natural decay of the signal.

For most guitar players, the brunt of the unwanted signal at the point of initial plug-in from the guitar itself, is likely to be 60hz hum, rather than hiss or other sources (which are more likely to arise from subsequent processing by pedals, especially high-gain devices).  So, when you think about it,  any nise-control to be inserted as the first thing a guitar sees, would ideally focus on EMI.  In other words, it would ideally be an automatic lo-cut filter, that left the bass in when the guitar signal was substantial enough to mask it, but cut the bass out when you stopped playing.  Since it would not affect the top end, it would have very little impact on initial attack, and modest effect on string decay (which does tend to have more bottom than top).

Let's say we had a 2-pole highpass (12db/oct cut) filter, with a 160hz corner frequency (or thereabouts, for calculation purposes).  That would allow for "mistakes" to be made in threshold setting, yet still hear much of the desired part of the guitar signal.  Bass might be trickier, admittedly.  If threshold is set right, that would  trim a big chunk of annoying hum during the quiet bits.  Again, this presumes that such a circuit is the FIRST thing you plug into.  Any noise-control circuit inserted further downstream would be looking for multiple sources of noise, not just hum.

Whatcha think?

Funny that this just conjured the image of such a device to cut out other ringing strings on a bass guitar, so a player doesn't require skill to mute the other strings so they dont ring. I wonder if someone's already created such a filter/stomp.

[hex_void]

Last few days I had the opportunity to fiddle with the gear of a buddy of mine who plays guitar in a thrash metal band. He uses an EVH 5150III 50W amp on high gain setting, a NS-2 noise gate along some other effects not relevant here. After checking the noise gate's schematic I used its effects loop along with the amp's effects loop in such a way that the gate which was placed in the amp's effects loop and was triggered by the clean guitar signal. Setup was like this:

guitar ---> noise gate input
noise gate send (basically the noise gate input buffered) ---> amp input
amp send ---> noise gate return
noise gate output ---> amp return

Basically, the part of the NS-2 that handles the volume is placed between the preamp and power amp and triggered by the signal going into the preamp.

I'm going to implement the gate in a similar way into my distortion pedal. The only drawback is that I'll have to use an input buffer. Should I use a BJT (BC550C), JFET (BF245C) or opamp (1/2 TL082)? Power supply is 20V.

[Keppy]

I'd say use the opamp if you're already using the other half, otherwise use the FET. They'll both have higher input impedance than the NPN (depending on your other component choices of course), which is USUALLY desirable in a buffer.

[hex_void]

I'd say use the opamp if you're already using the other half, otherwise use the FET. They'll both have higher input impedance than the NPN (depending on your other component choices of course), which is USUALLY desirable in a buffer.

I suppose the opamp is more tolerant of power supply noise than the JFET buffer, isn't it?

[hex_void]

Over the last few days I experimented with the sort of setup I last mentioned and it works, everything was wired up on a solderless breadboard. I might actually use 2 JFETs instead of 1, each with it's own capacitor, so I get lower Rds therefore greater signal attenuation. I still have to tweak the circuit so it works better.

[hex_void]

I have another idea.
The gate works fine but kicks in too abruptly. This means that it suddenly kills the sustain. Sucks for longer notes.
I have a way around this but I don't know how to implement it yet. I need two decay modes - fast and slow. If I mute the strings decay should be fast - pretty much how the gate operates now. If I let the note ring the decay should be slow until the signal drops under a certain level.

To put it in a simpler way:
> note is played - gate lets the signal pass without any attenuation - gate is open
> note starts fading out - signal is gradually attenuated - slow decay
> note has almost completely faded out - signal is muted - fast decay - gate closes
> if I play the note then mute it - fast decay - gate closes

How to do this?

[amptramp]

I have another idea.
The gate works fine but kicks in too abruptly. This means that it suddenly kills the sustain. Sucks for longer notes.
I have a way around this but I don't know how to implement it yet. I need two decay modes - fast and slow. If I mute the strings decay should be fast - pretty much how the gate operates now. If I let the note ring the decay should be slow until the signal drops under a certain level.

To put it in a simpler way:
> note is played - gate lets the signal pass without any attenuation - gate is open
> note starts fading out - signal is gradually attenuated - slow decay
> note has almost completely faded out - signal is muted - fast decay - gate closes
> if I play the note then mute it - fast decay - gate closes

How to do this?

This sounds like fun.  You have to get an envelope of the signal which comes from rectifying it and storing the values with a short time constant then using a slower differentiator to determine how fast the signal is decaying.  Then you set a threshold depending on how fast the signal is decreasing.  This is a take-home exam.  You have until tomorrow to come back with a complete schematic.