LM1875 Whooshing, Noise issue

[6Leinad7]

First off I am a student in Electrical Engineering, however I don't actually know much about audio amplifiers specifically, so don't be afraid to get technical. I'm still learning quite a bit. I made this schematic and board about a year ago, before my course in opamps and before I knew about "star grounding" and some other things I've seen people recommend for noise. I'd like to keep this board as I spent a long time (and quite a bit of money) on it, designing and soldering it. The board is designed to keep all the amps in very close proximity, to reduce size. It's meant for a flexible 5.1 system, so there are 6 LM1875T's total with separated inputs.

Here are my schematic and PCB layout:
View attachment 6amp.pdf
View attachment amp_schematic.pdf

Parts:
C1 1uF, 50V C2/C3 10uF, 50V C4 0.1uF, 50V C5 0.22uF, 50V C6 2.2F, 50V
C7 100uF, 50V
R1/R2/R3 22k, 1/4W R4 1M, 1/4W R5 10k, 1/4W R6 220k, 1/4W R7 1, 2W
I basically made one amp layout stackable and reversible, which is why none of the grounds or power in's are connected to each other. This allowed me to pick which amps handled which channel later, giving me freedom with the enclosure that I have yet to make. (also, I forgot to label some of the inputs... don't worry about that.)
I am getting a audible "whooshing" sound on at least four of the six amps (I didn't bother testing all six). Besides this the audio is clear and unmuddled.

It does not increase with volume, and does not change with the wiggle test.
It is the most loud right when you give it power, then settles down to a lower volume once you start playing music. If you let it sit for a while, just on without any input, it does get louder to a point.

It decreases if I decrease the gain by changing R6. (obviously, I'm amplifying less noise then)

My theory is either noisy ground lines due to poor power supply line in placement, or bad/poor quality caps, but I can't seem to figure out a solution that allows me to keep my current design and board. I suspect my bipolar 0.22uF C5 or the input 1uF C1 but I really don't know much when it comes to noise issues.

Any tips or advice for this?

 

[farley]

Maybe concentrate on just 1 of the circuits and get it working first. No need to modify each copy with every experiment.

- Can you provide some photos of what you've built so far?
- What are you using for a power supply?
- What are you using for a load on the amplifier?
- What are you using for the signal input?
- The .1uF decoupling cap (C4) should really be connected right at the chip between the chips power input pin and ground.
- Is C6 really a 2.2F, 50V electrolytic, in that small package?

 

[adason]

Its likely oscillating...if you hear weird noises in audio band, these may be indication of oscillations outside an audio band

 

[6Leinad7]

Maybe concentrate on just 1 of the circuits and get it working first. No need to modify each copy with every experiment.

- Can you provide some photos of what you've built so far?
- What are you using for a power supply?
- What are you using for a load on the amplifier?
- What are you using for the signal input?
- The .1uF decoupling cap (C4) should really be connected right at the chip between the chips power input pin and ground.
- Is C6 really a 2.2F, 50V electrolytic, in that small package?

Here are some photos. I apologize for them all being in portrait, I took them in landscape and uploaded them in landscape...





After getting up close with the camera you can see there is quite a bit of dust around all the components... I didn't notice this before, and I know that metal pieces surrounding wires can create inductance so that could be part of the issue.

I took a pretty cool power supply out of an old Yamaha receiver that was broken, but I accidentally fried that when making my enclosure over the summer. I was really close to just making my own supply, but I received a 24V, 4A supply as a gift that seems to be working just fine for at least three of these amps at the same time. It's a DROK WX-DC2412 something-or-other. It has substantially less noise (none, as far as I can tell) then the other supplies I've been using to test things.

Right now I have two separate speaker enclosures connected in parallel. One of them is 8 ohms (smaller midrange/tweeter), the other is a full-range woofer and tweeter combo at 6 ohms. I haven't had trouble driving them in the past with my previous supply.

I don't have anything special for the input besides a 10k slide pot, driven from my phone or my computer. The wires are separate all the way to the board.

C4 IS connected pretty darn close to the chip as you can see from the pictures. Does it need to be closer than that, like literally on the leads?
C6 is a 2200uF electrolytic. I must have forgotten a mF or something haha, definitely not that huge of a capacitor.

I am not hearing what I assume to be typical oscillatory behavior, but I could be wrong. There sometimes is some faint whining depending on the condition the amp was just in (ie, briefly turned off the music) but it isn't constant and appears somewhat sporadically.

 

[dimkasta]

My experience with those is that they are pretty stable. The only thing I know that can screw them, and is a bit tricky, is that they have very specific instructions for their grounding scheme.
Specifically the small current and large current returns must be in seperate stars and joined on one point in between. It is not rare to see this proposed as good practice, but with those chips it seems to be mandatory.
Check the datasheet for more descriptive info. Rjm also has a very nice graphic explaining this on his gainclone article (lm3875 but the principle is the same)

By the way don't be afraid to redesign the pcb. There are cheap fabs these days and they provide good quality, especially since we usually do not need very fine pitch.
I would design it with just a single amp on it, and just get more copies to cover each channel.

 

[erikb]

If you use 2200uF caps directly after the switching power supply, it might go into overcurrent protection mode (switching in audible frequencies is possible, you should be able to see it on a scope). If you use six channels, the SMPS sees 13mF, that is alot.
You can try decoupling the boards with something like a beefy 10 ohm resistor and see if the problem goes away, that should be very cheap to test, but is not a permanent solution for a class AB amp (power supply modulation). Decoupling with an inductor is better (no modulation), but more expensive.

 

[djoffe]

I didn't see values in the schematic. Perhaps I overlooked them, but please fill them in.

Also, I ddin't see an output inductor with damping resistor. That could certainly wreak havoc on stability when a speaker is connected.

 

[AndrewT]

Sporadic noise could be interference from other equipment, or intermitant oscillation.
Did you fit input filters to attenuate interference coming in via the cables?
Does the noise still exist when the input is shorted with a dummy zero ohms plug?
Did you fit a Zobel R+C at the output to help stabilise the amplifier when there is no high frequency load?
Did you fit supply rail decoupling? Both Medium Frequency and High Frequency versions?

 

[farley]

By the way don't be afraid to redesign the pcb. There are cheap fabs these days and they provide good quality, especially since we usually do not need very fine pitch.
I would design it with just a single amp on it, and just get more copies to cover each channel.

This is a good option. I don't know where you're located but here in the U.S. getting qty 10 PCBs from China (Elecrow for example) in less then 2 weeks would cost around $40 (USD).

Post your layout here as it progresses. Since you're studying EE the lessons learned should be helpful in your career.

Start with component placement. Post it here for advice and get the parts in the right places before you start routing connections. You can nudge them around later in the routing stage.

 

[AndrewT]

This is a good option. I don't know where you're located but here in the U.S. getting qty 10 PCBs from China (Elecrow for example) in less then 2 weeks would cost around $40 (USD).

Post your layout here as it progresses. Since you're studying EE the lessons learned should be helpful in your career.

Start with component placement. Post it here for advice and get the parts in the right places before you start routing connections. You can nudge them around later in the routing stage.

if you decide to start again, then I strongly suggest you design and build a SINGLE unit. test it as a single.
When you are sure you understand how and why the final version works, then build 5 copies of the successful outcome.

 

[dimkasta]

This is a good option. I don't know where you're located but here in the U.S. getting qty 10 PCBs from China (Elecrow for example) in less then 2 weeks would cost around $40 (USD).

2 weeks delivered from China for 40$ is already very cheap IMHO. You are paying for both super fast manufacturing and super fast shipment.

Normal prototyping times and registered postage should be ~30% of that though, for ~4 weeks max delivery.

Check out Oshpark too. They have a holiday special for the super swift service until 1/1. But it gets expensive quickly for even slightly bigger boards.
Nice quality though.

 

[6Leinad7]

If you use 2200uF caps directly after the switching power supply, it might go into overcurrent protection mode (switching in audible frequencies is possible, you should be able to see it on a scope). If you use six channels, the SMPS sees 13mF, that is alot.
You can try decoupling the boards with something like a beefy 10 ohm resistor and see if the problem goes away, that should be very cheap to test, but is not a permanent solution for a class AB amp (power supply modulation). Decoupling with an inductor is better (no modulation), but more expensive.

I'm not sure what you mean here, the 2200uF caps are directly at the output of the amp, not directly after the supply?

I didn't see values in the schematic. Perhaps I overlooked them, but please fill them in.

Also, I didn't see an output inductor with damping resistor. That could certainly wreak havoc on stability when a speaker is connected.

I posted the values below the schematic. When I was typing up the post I pasted all the values in as a table. It looked just fine but I guess it all got truncated into a couple lines.... I should put them in the schematic, I'll do that when I have the chance later today.

How large of an inductor and resistor should I use, and how should I arrange them? I've only seen that used as a filter on the output or input, but with audio I don't think I want to high-pass the output unless it's a very low high-pass?

Sporadic noise could be interference from other equipment, or intermitant oscillation.
Did you fit input filters to attenuate interference coming in via the cables?
Does the noise still exist when the input is shorted with a dummy zero ohms plug?
Did you fit a Zobel R+C at the output to help stabilise the amplifier when there is no high frequency load?
Did you fit supply rail decoupling? Both Medium Frequency and High Frequency versions?

What kind of input filters should I be using, just LP and HP filtering out anything outside the 20-20k range?
I've never heard of a Zobel filter before, are you talking about this? Car Audio - ZOBEL Filters for CROSSOVER Networks
If so, they seem pretty useful. I'll implement it as soon as I can find some parts that work.
I'll do some research on supply decoupling and try to implement that as well.

My experience with those is that they are pretty stable. The only thing I know that can screw them, and is a bit tricky, is that they have very specific instructions for their grounding scheme.
Specifically the small current and large current returns must be in seperate stars and joined on one point in between. It is not rare to see this proposed as good practice, but with those chips it seems to be mandatory.
Check the datasheet for more descriptive info. Rjm also has a very nice graphic explaining this on his gainclone article (lm3875 but the principle is the same)

I've looked at the datasheet and seen the grounding practices they recommend, and I wish I looked at that before I printed my boards! I'm not quite sure which lines hold the small and large current returns though. Are those the signal input and the power input channels? I'll check out Rjm's article.

Thanks for the swift and helpful replies. It was suggested multiple times that I build another set of test PCB's with just one amp on them which I definitely should do. I'd like to try any options that allow me to keep the current PCB in the interest of time and money (unsoldering, resoldering, waiting for PCB's, etc) because I only have another week or two of winter break before my classes start again. The more I can do now to troubleshoot the better, but I'll start designing a new PCB while I still have time.

 

[dimkasta]

I've looked at the datasheet and seen the grounding practices they recommend, and I wish I looked at that before I printed my boards! I'm not quite sure which lines hold the small and large current returns though. Are those the signal input and the power input channels? I'll check out Rjm's article.

Yes that is right.

The article is here.
RJM Audio - LM3875 Non-inverting Gainclone

You can see on the diagram how he names his returns signal ground and power ground.

 

[farley]

I'd like to try any options that allow me to keep the current PCB in the interest of time and money (unsoldering, resoldering, waiting for PCB's, etc) because I only have another week or two of winter break before my classes start again.

In that case some things you can try:

1. Disconnect everything from all the amplifier circuits. The photos don't show where all the wires are going or what they're connected to. Starting at a known point will make things simpler.

2. Add a .1uF (100nF) ceramic capacitor directly across pins 3 and 5 of the LM1875 (it can be soldered to the back of the PCB, directly to the pads for pins 3 and 5) in one chosen amplifier circuit.

3. Connect a speaker (8 Ohms if you have it, otherwise use what you have) to the output of that same amplifier circuit. Don't parallel or connect the speaker to/with anything else.

4. Connect your input signal coming from your volume potentiometer (assuming the potentiometer is wired correctly) into the audio inputs of the chosen amplifier circuit. Set the volume control to minimum. If you're unsure of the potentiometer wiring, let's address it before going further.

5. Connect the power supply to that same, chosen amplifier circuit and nothing else.

6. What results do you get when you apply power?

 

[6Leinad7]

In that case some things you can try:

1. Disconnect everything from all the amplifier circuits. The photos don't show where all the wires are going or what they're connected to. Starting at a known point will make things simpler.

2. Add a .1uF (100nF) ceramic capacitor directly across pins 3 and 5 of the LM1875 (it can be soldered to the back of the PCB, directly to the pads for pins 3 and 5) in one chosen amplifier circuit.

3. Connect a speaker (8 Ohms if you have it, otherwise use what you have) to the output of that same amplifier circuit. Don't parallel or connect the speaker to/with anything else.

4. Connect your input signal coming from your volume potentiometer (assuming the potentiometer is wired correctly) into the audio inputs of the chosen amplifier circuit. Set the volume control to minimum. If you're unsure of the potentiometer wiring, let's address it before going further.

5. Connect the power supply to that same, chosen amplifier circuit and nothing else.

6. What results do you get when you apply power?

Following these instructions exactly (disconnected everything, 8ohms speaker, minimum audio pot level, directly connected to the PS), the whooshing sound is almost completely eliminated. It is still there, just hardly audible to the point that I have to be right next to the speaker to hear it. Now, I am wondering why this fixed it. I already have a 0.1uF cap across the V- and Vcc pins, however it is directly at the input on the board rather than across the chip's pins. Did I actually need a 0.2uF cap (they are added in parallel now), or did it need to be right at the pins? My understanding is that this cap filters noise coming from the supply, so I guess I needed redundancy?

As I was typing the above paragraph, the whooshing noise came back momentarily. It slowly ramped up to not quite the volume it used to be, then back down to nothing. That sounds like a capacitor very slowly charging and discharging to me, but I really have no idea at this point!

After typing THAT paragraph, I played some music through the one channel and then paused it. The noise came back again momentarily, higher pitched than it was before then dissipated again to nothing. Could be a coincidence, I can't seem to replicate it again.

After leaving it plugged in without any signal the noise seems to have come back, still mostly quieter. Turning it off and back on fixes it again.

 

[farley]

In it's current position, as shown on the 6AMP.PDF drawing, C4 (.1uF) is likely not contributing much of anything to the circuit and you could probably remove it. The added .1uF cap at pins 3 and 5 are acting as the decoupling/bypass capacitor. At one time I had a link to a really good article on placing decoupling capacitors but I can't find it now. However, Dave Jones on EEVBlog has a video on the subject:
https://www.youtube.com/watch?v=BcJ6UdDx1vg

- Does the magnitude of the noise follow the setting of the volume pot? In other words can you make the noise louder or softer by adjusting the pot?
- Is the noise present if you short the audio input of the amplifier (connect +IN to GND)? - Remove your music source before doing this.
- Do you have an alternate power supply to try? It doesn't have to be 24VDC but I wouldn't go any higher than that. I can't find in the data sheet any mention of a minimum operating voltage. I would think that anything between 12VDC and 24VDC that could supply a couple of Amps should work for testing.

 

[6Leinad7]

I've been testing it for an hour or so and the noise is still there, it is just very low (volume and frequency). Now that I know what it sounds like I can hear it pretty easily in a very quiet room.

I watched that video and promptly tried connecting three different valued caps across pins 3 and 5 to no effect.

The noise does not change if I adjust the pot.

It is still present if I short the audio input.

Using a different power supply:
I have a very simple, barebones 0,-15,+15,+5V supply I built a year or two ago that I use to test simple projects. As it is literally one of the simplest supplies you can build, it is unregulated (besides the voltage regulators obviously). Using this supply to test, I get absolutely zero noise. Re-connecting my other supply yields the same noise as before.

I know that switching/regulated supplies can potentially have much more noise then unregulated, but I also thought that unregulated supplies are kind of horrible for audio applications. I can't use my test supply because it doesn't put out near enough current for six amps on the 15V rail, however I could make another easily.

Any suggestions on this? Should I just bite the bullet and make my own supply or try to filter out the noise somehow?

 

[farley]

Based on your response it would appear that the switching PS is causing the noise. If it were me I would not spend any more time with that Drok supply unless the goal is to understand how to properly design a switching power supply for this particular application.

Most audio amplifiers use unregulated power supplies. A transformer followed by a rectifier and filter capacitors is all one typically needs. A power supply such as that, suitable for your 6 amplifiers, running simultaneously, is a topic for another discussion.

 

[6Leinad7]

Alright, I'll look at making my own supply. Thank you for your input and help!

 

[AndrewT]

Fit the input filters and the Zobel.
They are on your sch.
The High Pass filter is C1 & R3 try 1uF and 47k
The Low Pass is missing. Add 1k to the input line. Add C of about 1nF in parallel to R3
The amplifier output Zobel is R7 & C5. Try 4r7 and 150nF
This is different from a speaker Zobel, which modifies the input impedance of the speaker.

Do not connect a speaker until after you complete your testing.
Leave the output open with just the Zobel attached.
Short the input with a zero ohms dummy plug.
Measure the output Hum+Noise using the 199.9mVac scale of your DMM.
measure the output offset using the 199.9mVdc scale of your DMM.