TPA3118 60W Power Amp Module

[Paul.Ruby]

EDIT #2: I've decided this is not worth the effort. The only way to really avoid pop is with proper behavior. Don't plug in a speaker cab until everything is powered. Unplug speaker cab first for power down. A mute button or stand-by on the amp might be a nice-to-have, but still requires proper behavior.

EDIT: Made some tweaks. R7 changed to 56k to mute if power supply is below about 17V. Trying to be aggressive with this to err on the side of muting if voltage is not adequate. Increased R4 to 1M and C1 to 100uf to get a longer power on delay. Made these changes because the pedal board has a lot of stuff that is all acting badly during power up and down. I can hear two pedals ticking relays during power on and they are making cracks and pops on power down. So, this is an attempt to get the power amp muted for as much of this as possible. But, it's still not good enough once combined with other things that can behave badly. Maybe a stand-by switch on the amp is the best option. I've added one to the schematic. When open, the amp is muted. I'm still stewing on this. I find myself wanting a circuit that will be stuck on mute at power on until a "go" button is pushed. Just keep the amp on mute for all power-on until user input is made. This would be a flip-flop with a toggle switch and designed to always power up in the muted state.

Here's a mute circuit I have on a breadboard and works really well. I'll probably wire this by hand and add to the pedals I've already built rather than do a PCB. The brown box is what is already on the TPA3118.

The mute pins on the TPA3118 are connected together, making mute the default behavior instead of unmute. Then, Q1 and Q2 are an AND gate to unmute. Both transistors must be on to unmute. Q1 has a delayed turn-on, held off by C1 when power is first applied. Will take ~100ms with 10uf. Q2 is meant to turn off very fast at power loss. R8 will pulldown the base of Q2 very fast when 24V is lost.

My only remaining concern is that C1 must be discharged to be prepared for the next power on. So far, this doesn't seem to be a problem. Manually pulling power and reapplying fast is not causing any pops. I don't think any further complexity is needed.

 

[Willybomb]

How are you getting around these boards blowing at 24v?

I'm also getting a lot of hum with mine when it's paired with a particular pedal that has a switchable 9->18v chargepump. The pedal is powered from a 1spot, and the amp is powered by a laptop supply @19v. It's starts off relatively quiet and then builds up. I think I gave the good supply to a mate to use with the 100lolwhaats!? build.

The big problem here is that when we start adding anti-this/that stuff, we start getting away from the small form factor possible with these, and then we might as well start playing with transformers and mains power.

 

[Paul.Ruby]

I added an 1N5408 in series with the supply to drop about 1.5V. And double checked that the supply I'm using is not behaving too badly. Really need to make sure you have a power supply that does not overshoot on power up. From one of my earlier posts, here's the stuff I had to do about noise and hum. Supply filtering to sensitive circuits and ground loop avoidance are key.

Summary of my learning building these things...

• TPA3118 is great. Flat frequency response (which is too dark on it's own, IMO). And rejects the noise of a cheapo power supply very well.
• Need a preamp to boost guitar/pedal/instrument level up to line level to drive the TPA3118 well.
• Fender-like EQ choice for the preamp works really well. Flat response from 20hz up to ~1kHz and then passive high-pass to boost the top end above ~1khz. (See C5/R6 and C6 in the schematic. Tweak to taste.)
• Cheapo power supply can work. TPA3118 can handle it but the preamp cannot. Need a good RC filter or better regulator to get clean power to a preamp. With ~20V available, an RC filter is fine for a simple preamp like this drawing under a mA.
• Don't allow more than one path from circuit ground back to mains earth from your pedal board supplies, especially if using a cheapo supply for the TPA3118. Do make sure you have ONE solid earth ground from your pedal board for safety, of course.
• Cioks DC7 is awesome and can drive the TPA3118 using it's auxiliary 24V output. (I was already doing this with a quilter amp, which is also really nice).

Very hard to debug hum and noise issues via chat. The one-spot is connecting all the pedal grounds together via the daisy-chain and also have a ground path through all the pedals via the instrument cables. That's a ground loop and one offending pedal anywhere in that chain could inject noise. Seems you know you have an offending pedal. Try running just that one pedal from a completely independent 9V power supply or a battery to see if removing it from the overall ground loop solves the issue.

You also need exactly one earth ground from your board's circuit ground for safety and need to be careful about sneak paths for ground, such as a direct line from the board to your PA or recording equipment. I use a radial stage bug for direct send with ground lift. The one spot isn't grounded. No idea if your chosen 19V supply has an earth ground. Since your 1-spot is not earth grounded, might make sense to use a supply for your pedal amp that is earth grounded. My Cioks DC7 provides my single safety earth ground to my pedal board.

My power supply has a smooth ramp and settles at 24.4V. Then it's dropped to ~23V internally through the 1N5408. This is the cheapo supply I'm using to make sure things work, not the Cioks DC7. This cheapo supply has no earth ground.

 

[Paul.Ruby]

(EDIT: Scroll down to post 160 for the simplest power-on delay) Simplified mute. This only has power-on delay, which can be tuned to what ever delay via C1. The fast mute during power-down was useless because it only helps when the 24V power plug is pulled directly out of the amp, which should be never. Power-on delay is debatable as the only value add. Still no substitute for proper behavior. I might layout a new volume-only-preamp with this since it's so simple.

Retrofit modules are built. Used 220uf for C1 to get ~5 seconds of power on delay. Built them as free-flying components and heat shrunk each junction and then the whole thing. Here's an installed module (the red heat shrink blob).

 

[Paul.Ruby]

Knowing that some folks have had the amps blow up makes me wonder if over-voltage protection should be included in a general design... Or at least include the pads for such things if folks want...

 

[Paul.Ruby]

In case anyone is interested in the volume-only preamp, here's my latest (final?) PCB. All 3 revisions are in the ExpressPCB file. Rev 3 adds the power-on delay and the LED resistor to the board. Does not add over-voltage protection. That's a high-power thing that should be elsewhere if desired. Also lined up most of the header pins in a row to use whatever headers are available. I quickly used up all my two's with prior layouts. And reduced the power connection to KF2510 since this draws about 1mA, even with the LED.

Schematic to go with this is below as well... Anywhere a 100nf is suggested, 47n or larger will work. Every one of them is intended to pass full bandwidth. I had an abundance of 82n, so used them for these two boards... I still used a 1uf for C4 in these but even 47n is plenty for that, too.

The 220uf, C12, in the power-on delay can be chosen for what ever delay you like. 220uf is ~5seconds to stay off while all the other stuff on the pedal board settles in. This cap will only ever hold 1V, so can be the lowest voltage you can find to keep it small. The 10uf, C11, should be 35V. Can be larger than 10uf if you like, too. I had some 47uf, 50V so used those here.

The NPN transistor type is not important. Any silicon BJT is fine. I have a small pile of counterfeit J201 from ebay that happen to be decent NPN BJTs instead. That was a fun discovery... EDIT: The same transistor can also feed the cathode of a stand-by LED if desired. Can be a 2nd LED or wire the cathode of the single LED to the stand-by connection instead of direct to ground if you prefer the LED to mean "Engaged" instead of just power.

The bright switch or stand-by switch can each be jumpered (closed) if you don't want those on your build. Can leave out the power-on delay components if you don't want that.

I think I can still get this into a 1590B case. Easily into a 125B...

 

[Paul.Ruby]

Decided to add a soft-touch momentary footswitch toggling between mute and unmute. This gave me the chance to also make a toggle flip flop that always powers up muted. Schematic shows how to connect it to the existing design. C13 and the larger value of R19 than R18 provides the power-on skew to make the latch always turn-on muted. The toggle circuit might be finicky, but has worked so far with all the NPN types I have on hand. I suggest breadboarding with the exact transistors you will use to test that it toggles properly. The sketchy part of this is that it uses Q4 as if the collector and emitter can be interchanged, depending on which way the latch must be flipped. There is no such thing as a symmetrical BJT (that I know of). Q2 and Q3 can be just about anything. I've tested the following as Q4: 2N5088; MPSA20; MPSA18; 2N3903; 2N2222; A42; S9013; S9014; S9018; S8050. This circuit could easily have a relay added to make a soft-touch true bypass foot switch as well. I've seen microcontroller based boards that do this. No microcontroller needed and it powers up in a known state.

The power-on delay circuit provides ground to the toggle circuit in the schematic, so this is providing both features. "Belt and suspenders" this way. The power-on delay could be removed. Connect the line at the bottom of the toggle circuit to ground if omitting the delay.

The attached zip file has the ExpressPCB layouts of all three versions of the Volume-Only preamp and the toggle as it's own little board. Updated the PCB layout to bring both Q and Q-Bar to the edge. Schematic is also updated to show that I have the mute lines of the TPA3118 connected to Q-Bar instead of Q. After assembly, Q-bar powers up high. Has something to do with the slow power-on through the delay circuit. Anyway, the flip flop alone will always power up with Q high.


Made one by hand but also have a PCB layout attached.

 

[Paul.Ruby]

This is just a heads-up that the 100W boards that look like this are not 100W. They are not even TPA3116 chips. These use the 16-pin cheapo chips and these boards can only deliver about 43W into 4 ohm. See the 2nd picture below with the heat sink pulled... These work... And sound OK. But they not even as good as the smaller TPA3118 boards. This is the exact amazon link that sold me the fake version.

Amazon.com

 

[Paul.Ruby]

EDIT: Verified PCB layout attached with a pair of mounting holes added.

Decided to make a a stand-alone circuit with an improved toggle flip-flop and power-on delay. Added Q5 to isolate the flip-flop from the circuits that are being controlled since that was messing with the flip flop operation. This allowed adding a stand-by LED. D1 is any diode. 1N914 will do. It keeps the 100k pull-down resistor on the TPA3118 from turning on the LED. ExpressPCB layout attached. The height of the board can fit along the side of a 125B enclosure. This will power-up in stand-by and won't allow coming out of stand-by for about 5 seconds.


Here's my first PCB under verification:

And planned mounting location...

 

[zgrav]

thanks for all of the updates as you are optimizing the build!

 

[tcpoint]

Here's another approach to the power on delay. It was designed for a stereo headphone amp but solves the problem with a delayed relay. E12 circuit.

 

[Paul.Ruby]

Here's another approach to the power on delay. It was designed for a stereo headphone amp but solves the problem with a delayed relay. E12 circuit.

I'm considering some higher-power builds. This might come in handy. I may use a relay to disconnect the speaker output for amps without a mute. And, I will likely want the power-on delay to be unpowered if there is no speaker plugged in so that power-on delay starts when the speaker plug is inserted. I have 200W and 500W modules coming... I built a 400W bass amp with high-end power-supply and amp modules. That one needs none of this because the amp itself handles all this gracefully. Those were expensive... I'm going to try some cheapos... I'm certain there will be nothing graceful about them.

 

[Paul.Ruby]

FYI, here's what my "full featured" build looks like. Volume-only preamp, bright switch, power-on delay, power-LED, stand-by-LED, toggle flip-flop, and soft-touch momentary foot switch. The toggle flip flop is the small board next to the foot switch. I'm keeping that 1" tall so it will fit on the side wall.

 

[Paul.Ruby]

EDIT: Verified PCB layout attached with a pair of mounting holes added.

Decided to make a a stand-alone circuit with an improved toggle flip-flop and power-on delay. Added Q5 to isolate the flip-flop from the circuits that are being controlled since that was messing with the flip flop operation. This allowed adding a stand-by LED. D1 is any diode. 1N914 will do. It keeps the 100k pull-down resistor on the TPA3118 from turning on the LED. ExpressPCB layout attached. The height of the board can fit along the side of a 125B enclosure. This will power-up in stand-by and won't allow coming out of stand-by for about 5 seconds.
View attachment 23704

Here's my first PCB under verification:
View attachment 23751
And planned mounting location...
View attachment 23752

FYI, updated post 149 above with a verified ExpressPCB layout for the stand-alone power-on delay and toggle flip flop for use with a soft-touch momentary foot switch that always powers up in a muted state. It's about 1.5" wide and 1" tall to be mounted with two stick-on nylon posts.

 

[Paul.Ruby]

Just tried using a Tone TwEQ version 1 from guitarPCB in comparison to the Cold Turkey from PedalPCB. These are very similar in design, sound and even the same size. Cold Turkey is easier to work with being only value changes. Tone TwEQ needs a cut on the board and different pots. The only notable difference of the Tone TwEQ is the three EQ pots in a row, if that is something someone really wants.

I have several more boards coming from PedalPCB to try in the near future...

Here's the schematic to use Tone TwEQ with a TPA3118. R2 and R3 get the gain correct. R11 and R12 changed to reduce current draw. C9 and D1 are not needed and one of the holes for D1 is used for the added 470R, which forms the needed RC power supply filter with C8. R13 gets the LED current correct, if you're going to use the LED. Also consider adding a bright cap of ~680pf to the volume pot.

 

[Paul.Ruby]

This is just a heads-up that the 100W boards that look like this are not 100W. They are not even TPA3116 chips. These use the 16-pin cheapo chips and these boards can only deliver about 43W into 4 ohm. See the 2nd picture below with the heat sink pulled... These work... And sound OK. But they not even as good as the smaller TPA3118 boards. This is the exact amazon link that sold me the fake version.

Amazon.com

View attachment 23702
View attachment 23701

The following version is real. The heat sink is screwed on, so easy to take off and put back on. No grease, so I'll be adding that... Also, although the mute signal is not brought out to a connector, it is easily accessible as it is routed on top over to R7 if control is desired.
EDIT: Amazon link removed. No longer points to a real version. Dang it. Can't find these originals anywhere now. :/



EDIT: Trying out the real board... It has a simple power-on delay circuit attached to the mute pin. C4 is the delay capacitor. It instantly couples the 24V supply rail to the mute pin which is then discharged to ground.

Scope trace of the mute signal at power-on, using 10V as the supply for this testing. Red trace is the mute signal. Green is the output, relative to ground, which has some DC voltage when not on mute, so makes a nice marker for the un-mute timing. This is "good enough" for basic operation but not long enough on pedal boards with poorly behaving pedals. Since the components are easy to access, I will be removing C4 and R6 to connect the toggle-flip flop and 5-second power on delay. I will leave R7 (or jumper it) because the trace from there goes to the edge of the board where C4 is. That is a perfect spot to add a proper connector and pull-up resistor to power on in mute until unmuted externally.

Mute circuit changed to match the TPA3118 board with a super-glued connector. Once C4 is removed, there's plenty of room there for stuff.

 

[Paul.Ruby]

The following version is real. The heat sink is screwed on, so easy to take off and put back on. No grease, so I'll be adding that... Also, although the mute signal is not brought out to a connector, it is easily accessible as it is routed on top over to R7 if control is desired.
EDIT: Amazon link removed. No longer points to real version. This sucks.

View attachment 23830

EDIT: Trying out the real board... It has a simple power-on delay circuit attached to the mute pin. C4 is the delay capacitor. It instantly couples the 24V supply rail to the mute pin which is then discharged to ground.
View attachment 23853
Scope trace of the mute signal at power-on, using 10V as the supply for this testing. Red trace is the mute signal. Green is the output, relative to ground, which has some DC voltage when not on mute, so makes a nice marker for the un-mute timing. This is "good enough" for basic operation but not long enough on pedal boards with poorly behaving pedals. Since the components are easy to access, I will be removing C4 and R6 to connect the toggle-flip flop and 5-second power on delay. I will leave R7 (or jumper it) because the trace from there goes to the edge of the board where C4 is. That is a perfect spot to add a proper connector and pull-up resistor to power on in mute until unmuted externally.
View attachment 23855

FYI, post 159 above updated with info about the mute control of the real TPA3116 board.

 

[Paul.Ruby]

FYI, if all you want for the mute is a power-on delay for the TPA3118 board, this is the simplest possible. (EDIT: Not simplest. Scroll a couple posts down.) Gives about 5 seconds. Must use a 25V or higher voltage capacitor since it will be sitting forever at 24VDC once powered for 5 seconds. This method is not combinable with other forms of mute because you have to yank around the voltage on the capacitor to move back to mute. It will not mute fast nor will it unmute fast. Don't consider just adding a switch to bridge the capacitor, that will cause a spark internal to the switch and wear it out pretty quick and likely will cause an audible crack through the speakers prior to being muted. If you want anything more than power-on delay, use the the transistor circuit for it, which is easily combined with other methods of muting. I've fully settled on the toggle flip-flop and power on delay circuit as it's own board with a soft-touch momentary foot switch (post 149 above). It is working great with el-cheapo 2n2222 transistors in all spots.

 

[Paul.Ruby]

Since the TPA3116 board already has power-on delay, if you only want power on delay but want longer (about 5 seconds), remove R6 (under the heat sink, see post 156 above) and add a 22k resistor from ground to negative terminal of C4. Like this:

 

[Paul.Ruby]

FYI, if all you want for the mute is a power-on delay for the TPA3118 board, this is the simplest possible. Gives about 5 seconds. Must use a 25V or higher voltage capacitor since it will be sitting forever at 24VDC once powered for 5 seconds. This method is not combinable with other forms of mute because you have to yank around the voltage on the capacitor to move back to mute. It will not mute fast nor will it unmute fast. Don't consider just adding a switch to bridge the capacitor, that will cause a spark internal to the switch and wear it out pretty quick and likely will cause an audible crack through the speakers prior to being muted. If you want anything more than power-on delay, use the the transistor circuit for it, which is easily combined with other methods of muting. I've fully settled on the toggle flip-flop and power on delay circuit as it's own board with a soft-touch momentary foot switch (post 149 above). It is working great with el-cheapo 2n2222 transistors in all spots.
View attachment 23914

Duh. Even simpler power on delay for TPA3118... Just solder in a cap.