Irritating high-pitched noise with 555 PWM driving LEDs (Update: 12 Aug 2015)

[thejoggingmat]

Hi, I was building an adjustable PWM driver with 555 (circuit schematic below):


The LED Array is a salvaged one from a visualiser. It has some SMD LEDs with 300ohm resistors.
The batteries are 18650s. I measured the (800ma) current usage of the LED Array @ 12V beforehand.

Whilst the curcuit seems to be working as intended (dimming the LEDs), there is an extremely annoying high-pitched whine coming from the LEDs, which vanishes when I adjust the brightness all the way up. (The noise increased as I decreased the brightness of my circuit.) The problem still remained when I swopped with 2 other different LED arrays (this time, they are sets of 5mm through-hole LEDs from cheapo LED video lights). I tried adding inductors (Suggested by looking through Google) of different values, capacitors to filter the DC out (again, Google), and they don't work. I have checked the capacitors & resistors in my circuit (and they are all within spec), and swopped out the components for my backup ones. Still the same.

This has been bugging me for a few days now.
How can I remove that noise?


EDIT: Updated schematic. Adding a 1K cap indeed made that squeal mostly softer (still audible), but the frequency of the PWM signal is very unstable (not to mention reduced by a significant amount).
Measuring at the lower end of the brightness level without the 1000uF cap was a stable 30kHz, now it's fluctuating between 7-13kHz. The brightness of the LEDs are also reduced (though I admit this is a good side effect). Q1 now is also extremely hot (~70 Celsius without heatsink) and putting back the 1K resistor yielded the same result. To make sure it wasn't due to the 'wun hung lo' caps I was using, I swopped it for a Nippon cap. Same result.

Also, 150R resistor at R2 was too low a current to drive the transistor. The lowest value I could use was 500R.

The saga continues...


UPDATE2: I've spent some time experimenting more on the circuit and brainstorming a bit.

On a whim, I tried using my circuit to drive a brushless computer fan without any extra protection. Realised that was stupid after burning through 2 555s. It didn't even want to drive the fan at all.

Tried constant current drive, however the circuit drew 2A and I burned through my only 100k pot for some odd reason. Now I'm using a few resistors as voltage dividers in place of my pot to test.

To clarify, here is a quick sketch of the main LED Array I intend to use (as I have a few lying around), after reverse line tracing (not sure if this is the correct term):

There are 4 sets of this, all SMDs. Resistors are all 300R 0603s (metric unit). I noticed that on one of the arrays whose manufacture date is newer, they added 1 cap on each set of 3 LEDs and a large filter cap (all SMDs). The other 2 or 3 doesn't have them. My DMM reads 8µF, though it isn't historically reliable in capacitance measuring. Oddly, the newer array actually sings less than the others without the caps.


Needless to say, I would prefer mitigating as much singing as possible from all the LED Arrays I have.

Right now, I will try migrating everything to veroboard to try out once I purchase what I need from my local electronic parts mall. It will simulate the cramped conditions I need to shove everything into the size of a 3x18650 battery holder.


UPDATE3: I finally got around to shoving everything onto a veroboard. This board below is my 2nd attempt (I removed the caps for troubleshooting, see below):

(** The jumpers lead to the potentiometer)

Unfortunately, migrating to a veroboard caused more problems: https://www.eevblog.com/forum/beginners/circuit-behaves-completely-differently-on-veroboard-from-breadboard/

[Gyro]

Hard to say exactly where the noise is coming from, there are no inductors to whine so it could be from pulse currents the wiring or maybe even the batteries themselves ( ). I'm pretty sure that putting a capacitor (maybe 100 - 1000uF) across the supply as close as possible to Q1 and the LED2 will quieten it down by shunting the pulse current through as small a loop as possible,

[amyk]

Are you sure the noise is coming from the LEDs?

I bet it's coming from the capacitors. Are they ceramics? Those are known to be sound-producing. (That linked paper suggests a single cap can produce almost 60dB!)

(This also gives me a wacky idea: in an embedded system, getting rid of a piezo speaker that isn't required to be that loud, and using the array of ceramic decoupling caps for the MCU as a speaker by running code that causes the appropriate frequencies on the power rails...)

[Gyro]

I bet it's coming from the capacitors

C2 is just the normal control pin decoupling capacitor, I suppose C1 is a possibility, but it's not seeing much current or fast edges. The thing that stands out for me it the total lack of supply decoupling on a circuit that's pulling significant pulse current.

Edit: The 555 probably isn't very pleased with the situation either - add a 100nF as close to its supply pins as possible.

[mij59]

Try lowering the pwm frequency, change C1 to 100nF, decoupling the power rails is always a good.

[Zero999]

How warm does Q1 get? R2 should be reduced, otherwise Q1 won't saturate properly, try 150R.

The trouble with reducing the frequency is flicker could become a problem

If audible noise is a problem, C1 could be reduced to 470pF to increase the switching frequency beyond the audible range but Q1 might struggle to switch at that speed. You could try putting a capacitor in parallel with R2 (say 470pF) to provide a negative pulse to discharge Q1's base. Failing that, try a MOSFET.

[thejoggingmat]

Are you sure the noise is coming from the LEDs?

I bet it's coming from the capacitors. Are they ceramics? Those are known to be sound-producing. (That linked paper suggests a single cap can produce almost 60dB!)

The caps are indeed ceramic.
Unfotrunately, I just went back to my breadboard and the noise is indeed coming from the LEDs side. I separated the LEDs from the main circuit with 1m wire.

(This also gives me a wacky idea: in an embedded system, getting rid of a piezo speaker that isn't required to be that loud, and using the array of ceramic decoupling caps for the MCU as a speaker by running code that causes the appropriate frequencies on the power rails...)

Hmm... Interesting...

[thejoggingmat]

Hard to say exactly where the noise is coming from, there are no inductors to whine so it could be from pulse currents the wiring or maybe even the batteries themselves ( ). I'm pretty sure that putting a capacitor (maybe 100 - 1000uF) across the supply as close as possible to Q1 and the LED2 will quieten it down by shunting the pulse current through as small a loop as possible,

It's night here now; I will try tomorrow and report back. Previously tried with 10 & 47uF electrolytics and that did nothing. I have to rummage in my parts bin first.   

[thejoggingmat]

How warm does Q1 get? R2 should be reduced, otherwise Q1 won't saturate properly, try 150R.

The trouble with reducing the frequency is flicker could become a problem

If audible noise is a problem, C1 could be reduced to 470pF to increase the switching frequency beyond the audible range but Q1 might struggle to switch at that speed. You could try putting a capacitor in parallel with R2 (say 470pF) to provide a negative pulse to discharge Q1's base. Failing that, try a MOSFET.

Here's the odd thing. Q1 is surprisingly cool (measured ~30 Celsius with thermocouple), whilst the 555 timer is running at ~55 Celsius after about 5min).

Can't increase the switch speed as doing so caused Q1 to act all wacky (It intermittently flickers between full brightness -> off without the PWM dimming) for no apparent reason.

Unfortunately, I'm out of MOSFETs (reason why I'm using a TIP instead) so that's not an option for a while.

I'll try with a 150ohm resistor and report back.

[thejoggingmat]

I bet it's coming from the capacitors

C2 is just the normal control pin decoupling capacitor, I suppose C1 is a possibility, but it's not seeing much current or fast edges. The thing that stands out for me it the total lack of supply decoupling on a circuit that's pulling significant pulse current.

Edit: The 555 probably isn't very pleased with the situation either - add a 100nF as close to its supply pins as possible.

The 555 obviously didn't seem pleased; it runs much much hotter than Q1 (~50-60 Celsius for the 555 vs ~30 for Q1).

[thejoggingmat]

Try lowering the pwm frequency, change C1 to 100nF, decoupling the power rails is always a good.

Sadly I can't lower the PWM frequency anymore; I'm using the lights for my event/wedding videography (my part time job) and the supplemental resistors/caps used were the lowest switching frequency that wasn't picked up by my cameras (my DMM shows ~1.3k - 16kHz).

I'll try decoupling the rails though.

[alsetalokin4017]

I can't help but notice the inconsistent use of "Dots" vs. "No Dots" to indicate connections vs. nonconnected crossings in the Schematic. For example, Pins 4 and 8 are definitely connected but there is no dot there. Are the other "no dot" points where wires join to be connected, or not? Perhaps the OP has miswired something, due to this inconsistent usage.

Is the circuit built on a solderless breadboard? We have seen other breadboarded timer circuits that malfunction because the breadboard contacts are flaky. I really don't think the 555 should be heating up so much in this kind of service. But maybe, since it is acting as a current source for turning on the power transistor. I would strongly suggest using a mosfet instead of the TIP transistor. You can probably pull a good mosfet out of a defunct CFL lightbulb power supply, everybody has those lying around don't they? (Although some do use bipolar transistors.)

The convention of using non-dotted crossings to indicate no connection has always driven me mad. There should be a "jump" symbol or just a little gap in the line where non-connected crossings occur in schematics, and definite Dots where they do connect.

[android]

Shouldn't the batteries be in series?

[thejoggingmat]

Shouldn't the batteries be in series?

Oh s**t, my bad. They are supposed to be in series.

[thejoggingmat]

I can't help but notice the inconsistent use of "Dots" vs. "No Dots" to indicate connections vs. nonconnected crossings in the Schematic. For example, Pins 4 and 8 are definitely connected but there is no dot there. Are the other "no dot" points where wires join to be connected, or not? Perhaps the OP has miswired something, due to this inconsistent usage.

Is the circuit built on a solderless breadboard? We have seen other breadboarded timer circuits that malfunction because the breadboard contacts are flaky. I really don't think the 555 should be heating up so much in this kind of service. But maybe, since it is acting as a current source for turning on the power transistor. I would strongly suggest using a mosfet instead of the TIP transistor. You can probably pull a good mosfet out of a defunct CFL lightbulb power supply, everybody has those lying around don't they? (Although some do use bipolar transistors.)

The convention of using non-dotted crossings to indicate no connection has always driven me mad. There should be a "jump" symbol or just a little gap in the line where non-connected crossings occur in schematics, and definite Dots where they do connect.

I had no idea how to use a 'jump' in EAGLE. Tried using the 'arc' but EAGLE keeps automatically adjusting to meet the wire which I wish to intersect in the first place. I ended up showing those junctions that I wish to join as dots.

I'm currently using a solderless breadboard to test, though I swopped my jumper wires for plain solid core because the jumpers were heating up.

You can probably pull a good mosfet out of a defunct CFL lightbulb power supply, everybody has those lying around don't they? (Although some do use bipolar transistors.)

I have to consider myself an odd one out here. Everybody I knew scrapped them long ago. I've ordered some jellybean MOSFETs from evilBay last week, and they're going to take another month++ to reach me.

[thejoggingmat]

I've updated the schematic, and tried some changes that you guys suggested.

Now the circuit seems more weird than before (see edit).

[mij59]

Hi, I was building an adjustable PWM driver with 555 (circuit schematic below):


The LED Array is a salvaged one from a visualiser. It has some SMD LEDs with 300ohm resistors.
The batteries are 18650s. I measured the (800ma) current usage of the LED Array @ 12V beforehand.

Whilst the curcuit seems to be working as intended (dimming the LEDs), there is an extremely annoying high-pitched whine coming from the LEDs, which vanishes when I adjust the brightness all the way up. (The noise increased as I decreased the brightness of my circuit.) The problem still remained when I swopped with 2 other different LED arrays (this time, they are sets of 5mm through-hole LEDs from cheapo LED video lights). I tried adding inductors (Suggested by looking through Google) of different values, capacitors to filter the DC out (again, Google), and they don't work. I have checked the capacitors & resistors in my circuit (and they are all within spec), and swopped out the components for my backup ones. Still the same.

This has been bugging me for a few days now.
How can I remove that noise?


EDIT: Updated schematic. Adding a 1K cap indeed made that squeal mostly softer (still audible), but the frequency of the PWM signal is very unstable (not to mention reduced by a significant amount).
Measuring at the lower end of the brightness level without the 1000uF cap was a stable 30kHz, now it's fluctuating between 7-13kHz. The brightness of the LEDs are also reduced (though I admit this is a good side effect). Q1 now is also extremely hot (~70 Celsius without heatsink) and putting back the 1K resistor yielded the same result. To make sure it wasn't due to the 'wun hung lo' caps I was using, I swopped it for a Nippon cap. Same result.

Also, 150R resistor at R2 was too low a current to drive the transistor. The lowest value I could use was 500R.

The saga continues...

The 1000 uF capacitor should be on the power supply rail NOT parallel on the transistor !

[mij59]

To avoid led flicker you could use a buck boost converter based on the XL6009 like this http://www.ebay.nl/itm/Hot-Selling-DC-DC-Boost-Buck-Step-Down-Up-Converter-XL6009-Solar-Voltage-Module-/301596262442?hash=item463889c02a

[Gyro]

The 1000 uF capacitor should be on the power supply rail NOT parallel on the transistor !

Absolutely! and as close to the transistor and LED array as possible.

You also still need to add local supply decoupling to the NE555 - 100nF between pin 1 and pin 8.

[thejoggingmat]

The 1000 uF capacitor should be on the power supply rail NOT parallel on the transistor !

Absolutely! and as close to the transistor and LED array as possible.

You also still need to add local supply decoupling to the NE555 - 100nF between pin 1 and pin 8.

Tried putting the cap on the power supply rail, but the sound got much louder. By default, shouldn't filtering the power supply make the power line cleaner (and thus reduce this problem)?

[thejoggingmat]

To avoid led flicker you could use a buck boost converter based on the XL6009 like this http://www.ebay.nl/itm/Hot-Selling-DC-DC-Boost-Buck-Step-Down-Up-Converter-XL6009-Solar-Voltage-Module-/301596262442?hash=item463889c02a

So I should use this module instead of the 555-based circuit?

[mij59]

To avoid led flicker you could use a buck boost converter based on the XL6009 like this http://www.ebay.nl/itm/Hot-Selling-DC-DC-Boost-Buck-Step-Down-Up-Converter-XL6009-Solar-Voltage-Module-/301596262442?hash=item463889c02a

So I should use this module instead of the 555-based circuit?

Yes, considering the price you can't go wrong.

[mij59]

An other module with constant current adjustment, should give a better control of the brightness http://www.ebay.nl/itm/Automatic-Converter-Boost-Buck-LX6009-4-35V-to-1-25-25V-CC-CV-Voltage-Regulator-/400873719224?pt=LH_DefaultDomain_3&hash=item5d55ef85b8

[thejoggingmat]

An other module with constant current adjustment, should give a better control of the brightness http://www.ebay.nl/itm/Automatic-Converter-Boost-Buck-LX6009-4-35V-to-1-25-25V-CC-CV-Voltage-Regulator-/400873719224?pt=LH_DefaultDomain_3&hash=item5d55ef85b8

To avoid led flicker you could use a buck boost converter based on the XL6009 like this http://www.ebay.nl/itm/Hot-Selling-DC-DC-Boost-Buck-Step-Down-Up-Converter-XL6009-Solar-Voltage-Module-/301596262442?hash=item463889c02a

So I should use this module instead of the 555-based circuit?

Yes, considering the price you can't go wrong.

I'll order a few samples in a while to try out, but they're going to take a while to reach me.

In the meantime, are there any long-term implications if I were to use per the schematic I posted?
As 'dirty' as the circuit is, there's no denying that it makes the least audible high pitched noise. (Unless I can whip up a better one)

[mij59]

An other module with constant current adjustment, should give a better control of the brightness http://www.ebay.nl/itm/Automatic-Converter-Boost-Buck-LX6009-4-35V-to-1-25-25V-CC-CV-Voltage-Regulator-/400873719224?pt=LH_DefaultDomain_3&hash=item5d55ef85b8

To avoid led flicker you could use a buck boost converter based on the XL6009 like this http://www.ebay.nl/itm/Hot-Selling-DC-DC-Boost-Buck-Step-Down-Up-Converter-XL6009-Solar-Voltage-Module-/301596262442?hash=item463889c02a

So I should use this module instead of the 555-based circuit?

Yes, considering the price you can't go wrong.

I'll order a few samples in a while to try out, but they're going to take a while to reach me.

In the meantime, are there any long-term implications if I were to use per the schematic I posted?
As 'dirty' as the circuit is, there's no denying that it makes the least audible high pitched noise. (Unless I can whip up a better one)

Running components at high temperature will decrease their life expectancy.
The audible noise could drive you really mad, since its a long wave transmitter the Pentagon may ask you to shutdown the device so the can communicate with their submarines.