Power-on JFET mute - works perfectly, could parts be reduced?

[composition4]

Hi all,

I'm in the final phase of developing a circuit with a number of different outputs, powered by a 9v single supply. Everything is great however there is a horrendous power-on thump on the outputs that I couldn't get rid of using the usual tricks.  I spent a couple of nights and came up with a circuit to mute for the first 100ms after power-up using J109s, and it works perfectly. Part of the design goal was that if power was cut then reapplied suddenly, it should still mute perfectly on reapplication of power. I also didn't want to use mechanical relays, even though the whole circuit would have been much simpler if I did.

You can tack on as many J109s to mute as many outputs as you want - each should be wired to shunt the signal to ground as the VERY END of the output, after the output capacitor. The source impedance of each output should be around 100ohm minimum, which reduces even a really bad thump to a minor (negligible) noise, or if the thump needs to be virtually inaudible somewhere above ~2k.

I know the -9v generator oscillator is in the audio band (about 6-7kHz, intentionally so because of FCC), however the current is miniscule so it doesn't feed through to the audio at all, even with sloppy breadboard wiring and a very high gain audio circuit. I had a spare inverter from elsewhere in the circuit so used that as the oscillator.

So I thought I'd share it, and with that ask a question - can anyone see a way to reduce the parts count whilst keep the same functionality?

[PRR]

Seems like there should be some resistance in the Q2 collector. As I see it, when Q2 comes ON, the JFET Gate diodes conduct heavily and "infinite" current flows. That can't be good. And even 1mA ought to be ample.

You *might* lose one BJT by inverting the pulse-maker?


Do P-type JFETs avoid need for a negative rail? (Too late to think.)

> there is a horrendous power-on thump on the outputs that I couldn't get rid of using the usual tricks

Maybe that is the real problem you need help on?

[composition4]

Thanks Paul

Good point on the inversion and removal of the NPN, works fine albeit D1 should be reversed from your picture. The reason I put D1 in was to prevent the pulse to the first transistor base going negative.

Interesting point on the collector resistor. I (naively?) assumed the "infinite" current would be only for a split nanosecond until the JFET gate capacitances charged up so it wouldn't be an issue? Maybe it's good practice to put in that 10k anyway.

You're right about fixing the problem of the thumping rather than "band-aid" it, however I tried for days to remove thump with pulldown resistors and all sorts but I just could not get it to disappear. I couldn't get my head around it, I tried down to a 100k pulldown on the right hand side of C76, but it seemed to me that when power was applied, the near instantaneous VREF from the opamp charged the left side of C76 too quickly and the change in voltage passed through the uncharged cap to the amp?  I feel like I am missing some critical detail and it seems like the detail must be something basic I shouldn't be overlooking, but for the life of me I can't see it.

[composition4]

Oh and in regards to P channel JFETs, their on resistance is too high in general (100 ohms plus it seems) compared to N channel, so it wouldn't work for me unless I increased the source resistance a lot, which I didn't want to do - a couple of the outputs I have need to be fairly low impedance.

[Rob Strand]

HiFi power amplifiers often have simple mute circuits using a couple of transistors.   They often use zeners to make a clear transition between and off.      This helps stop relay based circuits chattering but for a JFET  it will help slowing down the gate signals using an extra cap to get a very soft unmute.

These types of circuits are the sort of thing,

https://www.head-fi.org/threads/power-up-pop.116075/

Have a poke around at some HiFi amplifier service manuals.   Watch out as some circuits rely on rectifying AC from the transformer to work.  You don't want those ones.

For DC circuits its a good idea to put a backwards diode across the timing resistor.   What this does is resets the timing cap quickly when you turn the power off and on quickly.   If the circuit doesn't reset the timing cap still thinks it's in a unmute state and you will get a bang with quick power off and power on.   For *very* fast off and on the power supply caps hold-up and you might not get a bang but there can be an amount of time a bit longer than that where the power supply caps are discharged but the mute circuit is not.

[PRR]

> when power was applied, the near instantaneous VREF

What is Vref??

If it is say 10k+10k across the rails, bypassed with 1,000uFd to ground, it will come-up VERY slow. If outputs are say 1uFd with a 10k bleeder, Vref rises 500 times slower than the output networks will pass.

[composition4]

Thanks Rob, most of the circuits I've found (including the one you linked) don't seem to offer any benefit over what I landed on, and performance seemed worse for the majority of them.

Paul, The main supply is decoupled by 10 ohms into 10,000uF (yes excessive but I really didn't want that voltage drop using a higher value resistor), which is feeding an op-amp buffered 47k/47k divider. So the buffered divider should come up as slowly as the main supply, however I found that it doesn't, maybe because the buffer op-amp is still establishing its power supplies at the same time. And adding a large cap (10uf up to 1000uf) to the vref divider on the opamp input just made one large thump into a series of large thumps... I assume one when the main supplies powered up, and a few more as VREF was coming up to 4.5v.

[Rob Strand]

Thanks Rob, most of the circuits I've found (including the one you linked) don't seem to offer any benefit over what I landed on, and performance seemed worse for the majority of them.

True.  I would recommend adding the reset diode though.

I guess one issue adding to complexity is you needed to add the negative voltage generator.    One way around that is to use P-channel JFETs which can use a positive unmute signal but that  makes availability lower.

An alternative to JFET is to use NPN transistors.  They are used as muting devices in many consumer products.   In some cases the transistors are used in reverse mode with emitter used as the mute terminal.  You want to make sure the BJTs stay off when the signal is high.

Not a bad write-up here, also see comments.  This one uses a special transistor.

https://www.electroschematics.com/muting-transistor-attenuator-circuits-2sc2878/

More common is along the lines of the Sony playstation example on the same page.

Ah cool, our old friend Mr Elliot has something also,

https://sound-au.com/project147.htm
------------------------------------------------
Another consumer product example at the bottom of this page (left schematic).
http://www.renardson-audio.com/SLPG390-mod.html

[composition4]

I'm not sure exactly where you mean to add the timing diode?

I tried NPNs originally but I couldn't get any non-esoteric transistors to work well without adding distortion or not muting correctly.  I know about the 2SC2878, however I want to keep it to in-production and readily available (and cheap) parts.  I didn't even like using JFETs to be honest, given discrete JFETs seem to be disappearing rapidly. There is also the fact that JFETs are depletion mode - "normally closed" - so the instant to power comes on, it's muting already.

I also just changed the oscillator RC combo to 1M/.22nF - it wasn't a problem before, but I figure I could get away with it and it cuts the current of the oscillator by a factor of 10, so even less chance for current injection onto the audio path if I'm sloppy with routing when it comes time to create a PCB.

[diffeq]

A solution used in a commercial thing is using MOSFETs - they don't need negative rail to be switched on.
See Q608-Q609 here: http://hackabrute.yusynth.net/MINIBRUTE/digital-board/schematics/MiniBrute-10-OUTPUT.pdf
Multiple gates can be connected to the gate RC net.
Edit: 5VNUM net might need to be from fast comparator checking power rail raise for that to work. I guess they did it with MCU or something.

[composition4]

What happens when the signal at the MOSFET drains goes ~.7v below ground? Wouldn't the body diode conduct and clip the signal?

And if so, back to back diodes would necessitate 2 MOSFETs per channel, which equates to 10 MOSFETs for my purpose. Not out the the question if driving them uses less components I guess... however I feel that arrangement would end up using more components all up. Would be nice to avoid having to generate a negative voltage though.

[diffeq]

What happens when the signal at the MOSFET drains goes ~.7v below ground? Wouldn't the body diode conduct and clip the signal?

And if so, back to back diodes would necessitate 2 MOSFETs per channel, which equates to 10 MOSFETs for my purpose. Not out the the question if driving them uses less components I guess... however I feel that arrangement would end up using more components all up. Would be nice to avoid having to generate a negative voltage though.

Good point. Can they be placed before coupling caps, where there's positive DC offset?

[composition4]

I don't think so... once the MOSFET gets the control signal to release the shunt to ground, the VREF will flow through and charge the empty output cap, and the thump will occur, it'll just be delayed until the MOSFET releases.

[Rob Strand]

 

I'm not sure exactly where you mean to add the timing diode?

The main idea is it resets the circuit when you get a power-off, so the circuit doesn't think it is still in the mute state if the power comes on soon after.

I tried NPNs originally but I couldn't get any non-esoteric transistors to work well without adding distortion or not muting correctly.  I know about the 2SC2878, however I want to keep it to in-production and readily available (and cheap) parts.  I didn't even like using JFETs to be honest, given discrete JFETs seem to be disappearing rapidly. There is also the fact that JFETs are depletion mode - "normally closed" - so the instant to power comes on, it's muting already.

Sounds like you have done your homework.   I'm not enthusiastic about using special parts unless it's for a special job.

For transistors you need to keep the source impedance low to reduce distortion, like the 1K ohms you see in most circuits.   There's also the signal level.   For high signal levels, ie. in the 5V to 10V peak zone, the transistor will start to breakdown.  So if you want rail-to rail output you will see issues (the Elliot Sound page mentions this).   On a CD player the output voltage is limited by the maximum DAC output, that's why it works.

An *unexplored method* would be to use two output caps in series and connect the transistor *collector* to the cente of the two caps, then in between add a very large valued pull-up resistor from the collector to +V.   When the transistor is off the resistor biases the center point positively and helps prevent the transistor junction from conducting.

Distortion can also be caused by capacitive coupling from the junction capacitances of the transistor.  When the base is left open the CB capacitance will leak base current and the transistor gain will make that worse.   I suspect reverse mode might be better here as the gain is low but I'm only guessing.

I also just changed the oscillator RC combo to 1M/.22nF - it wasn't a problem before, but I figure I could get away with it and it cuts the current of the oscillator by a factor of 10, so even less chance for current injection onto the audio path if I'm sloppy with routing when it comes time to create a PCB.

Good idea just the same.


As for MOSFETs you can put two in *SERIES*: gate to gate, source to source, two drains are the mute switch connections.   

Technically you need to drive the gate relative to the internal source + source point.   There might be some problems driving the just the gate relative to the drain connecting to ground.

As with transistor you might get problems with capacitive couple causing distortion.   You might help stop that with a cap or resistor from the gate to the sources.  I'm only thinking out allowed here.

Yet another idea is to use one those MOSFET optocouplers.   These can switch AC (these types *switch* AC and are different to the optos which can be *controlled* by AC).   You need the type which has the two MOSFET connected back to back (as I described).   These are parts which are available.

[composition4]

The main idea is it resets the circuit when you get a power-off, so the circuit doesn't think it is still in the mute state if the power comes on soon after.

Actually that's exactly what I did when I inverted the pulse driving circuit to get rid of the NPN as Paul suggested... I did it to prevent the pulse going negative with respect to ground, but it works for your reason too! I also replaced the two diodes in the negative voltage generator with Schottkys, just to get closer to -9v (it brings the voltage at the JFET gates to -8.2v instead of -7.5v... The J109s are specced at VGS threshold of -2v to -6v, but doesn't hurt to leave a bit of headroom. See below for what I've got thus far:

So if you want rail-to rail output you will see issues (the Elliot Sound page mentions this).

Yes rail to rail is important in my application, one of the outputs needs as much as I can squeeze out of a 9v supply.

Yet another idea is to use one those MOSFET optocouplers.

I considered these too as they are perfect for my application... unfortunately I'd need 5 per board for my two stereo and one mono outputs, and they aren't a cheap solution when all I'm trying to do is prevent a power-on thump. No problem if you're making only one, but I'm designing so that I can build a trial run of 50-100 so trying to keep costs down so I can sell at a reasonable price.

Interesting other ideas. I'll be certain to try the transistor/MOSFET suggestions - If I can get something that works well with similar parts and minus the oscillator/negative voltage generator required for the N-JFETs, I'll go for that option. 15 parts for the driving circuitry plus 5 JFETs for muting seems okay, but always looking to economise where I can.

Thanks for your help, and everyone else. If you've got any more suggestions or comments please feel free to keep the thread open! I'll update if I find a better solution for my application than what I've got so far.

[Rob Strand]

See below for what I've got thus far:

From what I can see it looks OK.

You are lucky in that JFETs naturally mute with 0V and the -9V rail is used for *un*muting.   If it were the other way around the supply needs to come-up first and it would be too late to mute!   I suppose this is where the JFET could actually beat the BJT mute.   The BJT mute circuit relies on grabbing the output in the early stages of the supply voltage rising.

Yes rail to rail is important in my application, one of the outputs needs as much as I can squeeze out of a 9v supply.

Not an unreasonable requirement, and something like a mute shouldn't impose any limits on performance.

unfortunately I'd need 5 per board for my two stereo and one mono outputs, and they aren't a cheap solution when all I'm trying to do is prevent a power-on thump. No problem if you're making only one, but I'm designing so that I can build a trial run of 50-100 so trying to keep costs down so I can sell at a reasonable price.

Agreed.

Thanks for your help, and everyone else. If you've got any more suggestions or comments please feel free to keep the thread open! I'll update if I find a better solution for my application than what I've got so far.

Thinking outside the box a bit.   There's a few more angles to try.

A common trick to reduce thump on a single supply is to make Vref=Vcc/2 rise more slowly by using a fairly large Vcc/2 cap.

In fact, you can also built a one transistor circuit to slow down the rate of rise of the *whole supply*.  Here's a MOSFET version but you can also use BJT's.
https://farm5.staticflickr.com/4462/36990363054_f4dbe6e8e2_z.jpg

The slowing of Vcc/2 as well will also help.

Your thump problem might have a different cause.   From your output stage pic you really need a resistor to ground on the output side of C77.   Without that resistor, the turn-on thump from the output cap is made worse than it need be as C77's charge current is forced to go through the amp.

This one is a bit more unexplored territory:  For a single supply you might be able use a BJT with a collector mute and avoid distortion by connecting the mute transistor to the junction of the R105 and C77.   The key difference here is in normal operation the voltage at the output of U23:A sits at Vcc/2 and this should help prevent distortion from the mute transistor.  BUT the down-side of this is when the mute transistor turns off you have another mute problem because DC needs to charge-up through C77.  The way to cope with that is to make sure the mute transistor turns on releases quite slowly.  (and you have to have that resistor to ground at the output.)

[EDIT: dumb typo's fixed]

[PRR]

Just saying the obvious: if you gotta make a negative rail for your mute(?), why not go wild and run the WHOLE burrito on +/- rails? One reason for +/- rails on serious audio is less turn-on/turn-off thump, since OUT pin is near zero either on or off. (It does not always stay zero as rails come through 1V-2V.....)