Power On/Off Thumps and Plops

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

buckethead

Well-known member
Joined
Nov 15, 2011
Messages
125
Location
Austria
Hi guys,

after searching the internet for infos about power on/off thumps/plops without success, I wanted to ask what you guys think about it.

I'm working on a compact monitor controller (2 stereo in/2 stereo out) with a Muses 72323 volume chip. For buffering/debalancing/balancing/mono-summing I'm using opamps - everything works great so far. The problem is powering the circuit on and off. With speakers attached and powered on, powering on the controller results in a huuuuuuuge thump/plop, seriously enormous. Powering off the controller is not as bad, but still quite noisy. It's not that I'm surprised about it, I'm more in search of why this actually happens or rather if there's any way to minimize power up/down thump noises in amplifier circuits (besides something like delayed relay muting stuff etc.).

Any input would be highly appreciated :)

Thanks and best regards,
Mathias
 
Check the outputs for DC, and also with a scope to see what transient voltage there is.

Some pop and thump is often inevitable unless you get really creative. Best practice is generally turn the monitors on last and off first.
 
Check the outputs for DC, and also with a scope to see what transient voltage there is.

Some pop and thump is often inevitable unless you get really creative. Best practice is generally turn the monitors on last and off first.
There's just a tiny amount of DC offset on the outputs, nothing that makes me worry. The transient after turn on is huge, I'll check it again on the scope tomorrow for an exact number.

What kind of creative ways are coming to mind? I'm thinking about some "soft-start" circuitry for the power rails, but it feels like overengineering and probably won't change the thump too much.

Proper power-on sequence is mandatory, but I have these "ooooops-days" that force me into thinking about foolproofness. :)

Add a muting relay for the outputs with a time delay circuit driving the coil? I've seen equipment with a "slow on/fast off" muting relay circuit.

Bri
I'm working on a new PCB version at this very moment, fighting with limited PCB real estate and thinking "Do I seriously need two additional relays or is there any other way?". But I can't think of a better solution...
 
Over the decades I thought about this a lot.

Customers don't like loud turn-on/off clicks, thumps, farts..... They are generally considered a sign of inferior design while there is no actual correlation other than that premium products design in benign turn on/off characteristics as a feature because customers expect it from premium designs. Actually customers expect it from all designs.
===
For today's TMI I recall years ago when dbx designed their first active PA system processor. While not new to product design it was their first foray into sound reinforcement and when their line level box is connected to thousands of watts and PA speakers, turn on clicks and pops matter. I don't recall them actually harming any speakers with the spurious turn on/off noises, their customers were disappointed and they corrected it by their next generation design.
===
There are multiple ways to deal with this... the premium cost is no object approach is to use a bypass relay that connects input to output when power supply is removed, and connects to the active output after a suitable settling time delay. The control circuity wants to turn the relay off immediately when power is removed and on slowly after power is applied.

The more cost effective approach (and I used lots of these for Peavey) is to use JFET shunts*** on the output with similar on slow/off fast characteristic. For the JFET to offer effective attenuation you want some series output resistance. In extreme click/thump mitigation you can cascade multiple series resistor and JFET shunts for more attenuation. The downside to this is that you get increased output impedance. Generally not an issue for line level SKUs. The shunt JFETs can be damaged by static and often diode clamps on output lines protect against some of that. Customers care far more about clicks and thumps than output impedance.

JR

**** in the wayback machine cheap tape recorders and the like used bipolar transistors for output mutes. These mute transistors were special design devices for low saturation voltage, and high reverse base-emitter zener voltage. I used bipolar transistor mutes in my old 4TR tape recorder for Peavey(AMR). I inherited that design that was pretty much cut and paste from a Dolby NR chip set app note.
 
Wow, very interesting read! I've been successfully using JFETs for momentary muting while channel switching in high gain guitar amps - although I remember having problems with measureable noise/distortion (probably bad mixed signal GND layout) + those JFETs .

For now I'm going the cost-no-object-route with relays
 
For my 4x250 home brew power amp I built back in the 70s P1010072.JPG
I built my amp into an old Western Electric chassis (from the 30s). The WE PS chassis had a 3 position on-on-off power switch that allowed old school designs to turn on the tube heaters to warm up, before applying full power. I used the first on position to charge the amp power supply through a resistor to limit inrush current. I didn't energize the 120vac loudspeaker relays until the 2nd on position. The speakers disconnected immediately upon power off. Click and pop free...

JR
 
Take a look at Ch. 16 in Doug Self's Small signal audio design book if you haven't yet. It discusses relay, JFET and CMOS muting circuits and distortion fairly in depth. (You can find a downloadable version if you google "small signal audio design pdf", but you didn't hear that from me).

Another option I don't think was mentioned so far are optically coupled relays. Google PhotoMOS and you'll find some good starting points. Advantages are that they are small compared to mechanical relays, and have much lower reliability / lifetime issues. Distortion is also very low, in most cases unmeasurable. The capacitance between the LED and the detector is very small (< 1 pF usually), and doesn't have as much voltage dependency as you'd get from a JFET or regular MOSFET switch, so non-linearity isn't imparted much.

Here's a photoMOS output disconnect circuit example:

1628433186048.png
Normally for a muting circuit, there would also be a second pull-down switch. In this case, I'm just trying to mostly prevent start-up pops from making it to the output. R91 (100 kΩ) should be much smaller than the PhotoMOS switch off resistance, so the divider formed will mostly keep the output at ground during startup. I'm driving the output enable line from a MCU, but it's not too difficult to use RC delay circuits. C82 and R89 slow the turn-on of the LED once output_EN is asserted, which prevents the output enabling causing a transient of its own which would happen if there's any signal being driven when enabled.
 
Last edited:
, I'm more in search of why this actually happens or rather if there's any way to minimize power up/down thump noises in amplifier circuits (besides something like delayed relay muting stuff etc.).
The cause is that the different parts of the circuit must suddenly go from zero to something volts. In an opamp, typically, the Vas stage (the one that procures the voltage swing and a large part of the gain) is basically a transistor that is turned-on. Depending on how the base current and the supply voltage behave respectively, there is a large hump.
One way of fixing this is using a perfectly symmetrical circuit. Some power amps take benefit of this approach, but it's expensive. I don't know of any monolithic opamp based on this technique, although some unity buffers did.
 
Not cheap ($2 digikey) that looks like it would work. The data sheet only shows on resistance vs temperature (15-35 ohm). That looks like a TG (transfer gate) but with optical drive.

JR
 
That looks like a TG (transfer gate) but with optical drive.
Yes, the datsaheet makes no secret of it. The major advantages I see is the near-perfect isolation between the command circuit and the signal path and the lack of need for powering them, like a CMOS switch would need. Not really significant in that application, though.
 
It looks like it depends on which one you look at. TLP172AM has an Ron of only 2 ohms. Because these are opto-isolated (and they have to be for various reasons), the LED current is the same or more than a relay. So solid state relays are just delicate versions of mechanical relays with high(er) Ron, lower voltage rating and lower power. For a line out they'll work fine. But there's just no major advantage over a mechanical relay other than size. And they're not super tiny either (7mm x 4mm vs 14mm x 9mm).
 
They're not one-size-fits all, obviously, but can have several advantages over electromechanical relays in certain applications:
- Significantly smaller size (3-4x at least)
- Significantly greater lifetime expectancy (EM relays are common failure points in a lot of the gear I've serviced from personal experience at least)
- Usually cheaper
- Controllable switching speed. Can be made to quickly fade in and out over a couple of ms, similar to JFET switching.
- Lower drive power (10 mA from a 2.5V or 3.3V rail is enough power, which is usually 3-4x less than than driving a non-latching EM coil)
- ON resistance stays the same over the product lifetime (No switch contact oxidation)

Distortion is comparable to an electromechanical relay, so comparing them to a JFET or regular CMOS switch isn't really fair.

Lots of Ron values are available, and there are driver-only parts available as well if you want to roll your own photo switches with power MOSFETs that have mΩ's of resistance.
 
Last edited:
They're not one-size-fits all, obviously, but can have several advantages over electromechanical relays in certain applications:
- Significantly smaller size (3-4x at least)
I use TQ2-5V which are 14mm x 9mm and TLP172AM is 7mm x 4mm. But in practice you need space around the SMD pads so not even 2x.

Also with a mechanical relay you get two poles which more often than not is actually used. So the size is not really better at all.

- Significantly greater lifetime expectancy (EM relays are common failure points in a lot of the gear I've serviced from personal experience at least)
I wouldn't be so sure about that. There is a note in the TLP172AM datasheet that reads:

"Using continuously under heavy loads (e.g. the application of high temperature/current/voltage and the
significant change in temperature, etc.) may cause this product to decrease in the reliability significantly even
if the operating conditions (i.e. operating temperature/current/voltage, etc.) are within the absolute maximum
ratings."

- Usually cheaper
Usually. But TQ2-5V are less than $3 is reasonable quantities.

- Controllable switching speed. Can be made to quickly fade in and out over a couple of ms, similar to JFET switching.
I use L9823 to limit slew rate. That's fine for limiting transients in the supply that might get picked up elsewhere. It won't help with popping from DC but I have always just designed circuits to not have DC across switching contacts / pots so that there is no popping in the first place.

- Lower drive power (10 mA from a 2.5V or 3.3V rail is enough power, which is usually 3-4x less than than driving a non-latching EM coil)
TQ2-12V is only 11.7mA. The TLP172AM datasheet does not show Ron vs If so it's not clear what sort of current you really need to get 2 ohms.

- ON resistance stays the same over the product lifetime (No switch contact oxidation)
Contact resistance of a relay is 10's of milliohms. Contacts are gold plated.

Distortion is comparable to an electromechanical relay, so comparing them to a JFET or regular CMOS switch isn't really fair.
Comparably yes. But 2 ohms vs 0.01 ohms is not comparible. I've never made the leap to solid state relays, but distortion would be something that I would want to check carefully. After all you are passing signal through semiconductor junctions whereas with a mechanical relay you are not. It's physical contacts.

Lots of Ron values are available, and there are driver-only parts available as well if you want to roll your own photo switches with power MOSFETs that have mΩ's of resistance.
Rolling your own could definitely be an option for really high voltage / high power. But it would take up some board space for sure.

I think that solid state relay would work just fine for most things. But the benefits are not that great and are not related to signal quality.
 
I use TQ2-5V which are 14mm x 9mm and TLP172AM is 7mm x 4mm. But in practice you need space around the SMD pads so not even 2x.

14*9/4/7 = 4.5x footprint advantage. Keepouts are present for all parts, so a bit of an unfair point. Height is also way smaller for these type of SSR.

I wouldn't be so sure about that. There is a note in the TLP172AM datasheet that reads:

"Using continuously under heavy loads (e.g. the application of high temperature/current/voltage and the
significant change in temperature, etc.) may cause this product to decrease in the reliability significantly even
if the operating conditions (i.e. operating temperature/current/voltage, etc.) are within the absolute maximum
ratings."
That's just a boilerplate reliability liability statement. You find the same or similar wording in numerous other types of parts & product families.

The same general statement can be applied to pretty much every electronic component (including EM relays). If you're using the part at high temperatures, or cycling temperature a bunch, or the part is dissipating a lot of power, it's time to failure is going to decrease. That's just reliability 101.

Usually. But TQ2-5V are less than $3 is reasonable quantities.
TLP172A is less than $1 for even small quantities.

I use L9823 to limit slew rate.
Not what I'm referring to. I mean you can slow the action of the switch itself with an SSR, so they can be used in mute, bypass, or other switching applications live during a mix. Impossible to do with an EM relay without clicking unless you could somehow guarantee switch close/open exactly at the zero crossing, which can't be done.

TQ2-12V is only 11.7mA. The TLP172AM datasheet does not show Ron vs If so it's not clear what sort of current you really need to get 2 ohms.
11.7 mA * 12 V = 140 mW for the EM relay.
TLP172A on resistance is specified at 1Ω for 5mA drive current at Vf=1.27V, so that's 6mW.

Contact resistance of a relay is 10's of milliohms. Contacts are gold plated.

At the start of life. Try switching it a few thousand times and see if you measure the same resistance, or wait a few years. In some cases, contact resistance can increase 10-100x over the lifetime of the relay.

And on signal quality and distortion.... unless you're still driving 600Ω loads you aren't going to measure any distortion with an opto switch. As always, application matters. but at least for the vast majority of my applications where I use SSRs, there's no measurable difference in signal fidelity vs. an EM relay.
 
Take a look at Ch. 16 in Doug Self's Small signal audio design book if you haven't yet. It discusses relay, JFET and CMOS muting circuits and distortion fairly in depth. (You can find a downloadable version if you google "small signal audio design pdf", but you didn't hear that from me).

Another option I don't think was mentioned so far are optically coupled relays. Google PhotoMOS and you'll find some good starting points. Advantages are that they are small compared to mechanical relays, and have much lower reliability / lifetime issues. Distortion is also very low, in most cases unmeasurable. The capacitance between the LED and the detector is very small (< 1 pF usually), and doesn't have as much voltage dependency as you'd get from a JFET or regular MOSFET switch, so non-linearity isn't imparted much.

Here's a photoMOS output disconnect circuit example:

View attachment 83465
Normally for a muting circuit, there would also be a second pull-down switch. In this case, I'm just trying to mostly prevent start-up pops from making it to the output. R91 (100 kΩ) should be much smaller than the PhotoMOS switch off resistance, so the divider formed will mostly keep the output at ground during startup. I'm driving the output enable line from a MCU, but it's not too difficult to use RC delay circuits. C82 and R89 slow the turn-on of the LED once output_EN is asserted, which prevents the output enabling causing a transient of its own which would happen if there's any signal being driven when enabled.
Why not put a small electrolytic (1 or 2.2 uF) after R96 and lose C82 and R89?
 
Why not put a small electrolytic (1 or 2.2 uF) after R96 and lose C82 and R89?
A cap on the base of Q19A is just going to delay the turn-on of Q19. Once the cap reaches Vbe, Q19 is going to turn-on relatively quickly. So, I don't gain anything vs. just delaying the enable line assertion from my MCU.

The purpose of C82 is to slow the transition of the switch, not to delay the switch.
 
Back
Top