Just made an LM1875 amplifier. Having problems with humming

[Plecto]

Hi. I've tried making an LM1875 amp in the past. The last time I had the rectifying diodes and decoupling caps on the same board as the amps and I even put on of the diodes across an input trace (kind of asking for humming). This time around I had the rectifier and decoupling caps on a separate board with a space of around 10cm between them. This helped a lot, but I later figured that a gain of 21 was not enough. Playing at full volume with my android as a source was no where near any form of distortion. I then tried a gain of 28 and finally 40. I still haven't made the amp clip though, but I'm afraid to play too loud because of my small heatsinks (I've ordered a much bigger one). The humming increased as the gain increased and I feel that it is unacceptable.

Both the LM1875's and rectifier boards are homemade PCB's. The spacings and signal traces are 28 mils, output and power traces are 40 mils. I have a ground plane that covers most of the board, but I haven't payed any other attention to grounding than that. What can I do to fix this humming and what causes it? I can see that having a power AC wire close to a signal wire can cause humming, but these wires are no where near each other in my design. Is it really the puny ripple voltage from the decoupling caps that causes it?

Here are some pictures (they are too large to post directly I believe):
https://i50.tinypic.com/27y8jee.jpg
https://i46.tinypic.com/2rfwvwy.jpg
https://i49.tinypic.com/20hkqkm.jpg
https://i45.tinypic.com/2rf3uqw.jpg

 

[BradtheRad]

I believe the hum would be less if your input signal were inside a shielded co-ax cable.

Just those few inches, unshielded, can expose it to mains hum emanating from the power supply board.

 

[betwixt]

BradtheRad's advice is good. You should use shielded cables between the PCB and the volume control and again between the volume control and the input sockets. Even then, route them away from the transformer. I would also use thicker wires between the PSU and the amplifier, those look like 7/0.2 wires which will drop more voltage at peak current than is good for high quality. If that 8-pin IC is a preamplifier, also check it has decoupling on it's supply pins.

Brian.

 

[Plecto]

Okay, I'll try that. Do I then just connect the shield to ground wherever I can? What if the signal input and ground are far a part, should I then just connect the shield to ground in one end and not the other or will that be pointless?

I'll get some thicker cables in there The Dip-8 is a comparator controlling the output relays.

 

[betwixt]

Connect both ends of the shield or you will have no signal ground return path. At the amplifier, connect the shield as close as possible to the ground of the appropriate IC and if possible keep the grounds of each channel apart from each other. In other words, don't link the L & R grounds at the input sockets or volume controls, run separate grounds all the way back to each amplifier. The idea is that the input wire and it's return path through the shield carry ONLY that channels signal. They are electrically joined on the PCB but the intention is to keep the signals apart as much as is possible until they get there. What you must do is ensure continuity of the shield as it enters and leaves the volume control, so you have incoming and outgong shields soldered to the bottom of each volume control, the input wire to the top and the wire to the amplifier to the wiper.

Brian.

 

[Plecto]

Okay, I shorted the wires from the secondary, replaced the wires from the rectifier to the 1875 board with thicker ones and replaced all signal wires with shielded wires. I connected them like this:

https://i50.tinypic.com/2z54x3s.png

This made a huge difference. The humming in one of the channels can only barely be heard by putting my ear really close to the speaker. The other channel still has humming, but it's a great improvement. I don't understand why one channel has more humming than the other though, they are connected the same way and all the distances are the same.

I would like to understand what's going on here though. Why should the ground wires going to the board be close to the non-inverting input pin of the amps? The current flows either from or to the center tapping of the transformer, shouldn't it ideally be connected directly to the transformer? Also, the shielding is to prevent electromagnetic waves from changing the voltage of the input signal. Why won't increasing the distance between the transformer and the signal wires help? I've tried moving the wires and LM1875 board around, but that doesn't change anything.

 

[betwixt]

Your wiring is correct. The transformer is toroidal so it would have relatively small external magnetic field anyway. Keeping all signal carrying wires away from potential sources of interference is always a good strategy.

The connections are ideally arranged so no current other than the desired signal is carried through the wiring. Whenever you draw current, no matter how small it is, you get a voltage drop, it may be very, very tiny but don't forget you are feeding it into an amplifier which makes it more problematic. So for example, your transformer wiring is passing alternating current to the rectifiers and they in turn to the reservoir capacitors and so the wiring between them has a voltage dropped along it. If that voltage reached the input of the amplifier it would be heard as hum, either at mains frequency, twice mains frequency or a mix of both. By wiring to the amplifier input and ground, and not creating a looped connection anywhere, you are preventing all possible current paths that could conduct the hum. Even tracks on the PCB can pick up tiny voltages, you can test this by moving it around and seeing if the hum changes.

Can you post the schematic, I'm a bit worried when you mention "non-inverting input pins", the shield should go to the amplifier ground as close as possible to the IC. Ground isn't normally one of the input pins. I have to ask what kind of screened wires you used, are they audio grade cables or TV antenna types, the latter are not adequately screened.

Brian.

 

[Plecto]

I misunderstood you slightly, you said this:

At the amplifier, connect the shield as close as possible to the ground of the appropriate IC

That's why I mentioned non-inverting input pin. The LM1875 does not have a ground pin when operated with a dual sided supply. What I don't get it why the signal ground has to be close to the appropriate IC.

So for example, your transformer wiring is passing alternating current to the rectifiers and they in turn to the reservoir capacitors and so the wiring between them has a voltage dropped along it. If that voltage reached the input of the amplifier it would be heard as hum

You mean the wires from the psu to the board? The connection between the diodes and the reservoir caps are just traces. This is what I'm wondering though, you say that if this voltage get's close to the input of the amplifier, it will be heard as humming. This voltage drop has to be tiny! I have 9400uF cap's for both the negative and positive side, when the amp is just idling the current has to be very small so the ripple current has to be very tiny as well. I assume it's this ripple voltage combined with the voltage drop across the cables that causes it? So that either reducing the ripple or increasing the wire diameter will help?

Here's the schematic:

https://i48.tinypic.com/212wk6b.png

The psu is a seperate schematic, but it just consists of 4x4700uF caps and four rectifying diodes. Just like this:

**broken link removed**

So to reason for the humming I have now is the pcb itself? What about the points when the shielding leaves the signal wire (the pot and RCA connectors)? Perhaps these areas are too small to pick up on any of the humming?

The shielded wires are from one of those phono to phono cables you buy in the store, they are not TV antenna types.

 

[Plecto]

Another thing. If it's the voltage drop in the wires from my decoupling caps to the LM1875 board that is the problem, then adding voltage regulators should further do a massive improvement, right?

 

[betwixt]

Voltage regulators will do no harm and will to some degree reduce any residual ripple on the supply wwires that might be causing the hum. Be careful though, a regulator has to carry the full current of both channels at peak volume and also needs a few volts 'overhead' at the input pin relative to the regulated output. You have to be sure the voltage at it's input will not dip below the overhead requirement at peak load or the regulator wont work and could make things far worse. A few years ago I was sent to investigate a misbehaving industrial machine and traced the problem to the power supply. The manufacturer was running 12V regulators with 15V at their inputs while off load, a borderline situation, when the load was driven, the 15V dipped and the regulators produced an almost square wave output which exploded (yes, big bang!) several electrolytic capacitors.

At this stage it becomes necessary to experiment to find out where the hum comes from. In reality, it may only be a very tiny voltage, maybe less than 1mV, but being amplified it is still troublesome. Do some tests for me:
1. Are the LM1875s running warm without any audio going through them? If they are, you might have instability at a frequency higher than you can hear which is increasing the current drain from the PSU.
2. Does the hum change with the volume control setting?
3. With the volume at maximum, what happens if you short out the input sockets? (center to outside) Does the hum level change?
4. If practical to do, can you release the amplifier board from it's mountings and see if holding it at different angles makes a difference. That may sound daft but if you have magnetic pick up on the tracks it will change according to their orientation in the magnetic field.

Brian.

 

[Plecto]

Yeah, regulators aren't really something I wan't to implement as I would need a higher supply voltage and the regulators will need heatsinking aswell.

As for the questions:

1. They do get warm, but not unusually warm. They are supposed to dissipate some heat regardless of the input and I have gotten other LM1875s to oscillate before and they generate far more heat than this one do. Both 1875's generate the same amount of heat as well, I really don't think any of them are oscillating.

2. The humming is constant until I reach about 60% on the volume knob. The humming will then go down a tiny bit and then stay constant for the remaining 40%.

3. Shorting the input sockets made no difference what so ever.

4. I tried this before I mounted it. I tilted it, moved it back and forth and even flipped it around, it made no difference what so ever.

 

[betwixt]

From what you describe, I'm guessing there are a few mV of ripple on the supply lines and thats how it gets into the amplifier inputs. It's undesirable but quite normal to have some ripple on the supplies but in your case you seem to have coupled it into the amplifier inputs. Again, some experimenting is needed:

DO NOT SWITCH ON WITH THE GROUND DISCONNECTED. But, try moving the GND wire from the PSU to a different place on the ground track of the PCB. I think you will find the hum changes according to position, even though all the ground points are electrically joined. See if you can find the optimum spot for lowest hum and leave the ground wired there.

Brian.

 

[FvM]

I'm missing a systematical approach in narrowing down the hum source. There are different ways hum can enter an audio amplifier:
- unshielded high impednace nodes by capacitive coupling
- voltage drops on ground lines and ground loops, in general unsuitable ground wiring
- supply voltage ripple in combination with insufficient bypassing of sensitive circuit parts
- magnetical coupling into ground/signal loops

These different coupling pathes can be distinguished by 1. thinking and 2. systematical tests.

 

[betwixt]

You are absolutely right FvM. Initially, the wiring was completely unscreened and close to the power transformer and fixing that has made a great improvement. I don't think there are any high impedance points and orientation of the wires and PCB make no difference so I'm ruling out magnetic coupling. My gut feeling is the residual low level hum is due to a ground wiring problem but on the PCB itself. I suspect the PSU ground traces are not optimal. It's unlikely, but not impossible that the relay coils are inducing current into the tracks.

Plecto, can you show a schematic and a photograph of the underside of the PCB and point out where the PSU wires are attached to it please.

Brian.

 

[FvM]

I tend to agree to your conclusions about likely causes of hum. But to succesfully design audio amplifiers, it's necessary that the OP is able to distinguish the different interference mechanisms by himself and fight them directly.

 

[Plecto]

I'm sorry for a late reply. I forgot about this thread in the christmas hurry

Here's a layout of the board: https://i49.tinypic.com/4tskds.png

You can see where the ground from the psu comes in. The ground from the inputs and pot has been moved so that they are closer to the signal wire.

It's such a pain to unscrew the board now so I'm not going to do any more resoldering. The humming is so weak now that you can barely hear anything if you stand 1 meter away.

I bought a new heatsink for it, but I'm unsure if it's big enough. I'm having a hard time figuring out how hot the heatsink and the packages can become. I know this is a little off topic, but how hot is it reasonable to allow it to become? I know the IC's can handle quite a bit of heat, but thinking of the enclosure, wires, other components, safety etc. how hot should it be?

 

[betwixt]

I think the layout is the problem but if you can live with the residual hum I wouldn't go to the trouble of redesigning it.
Quote from the LM1875 data sheet:

"When designing a different layout, it is important to
return the load ground, the output compensation ground, and the low level (feedback and input) grounds to the
circuit board ground point through separate paths."

This includes not returning the grounds from one amplifier through tracks of the other!

As a rule of thumb, if you can touch the heatsink it's OK. Obviously, cooler is better and the ICs have thermal shutdown anyway if they get too hot. As long as you use common sense regarding the proximity of plastics to the hot bits you should be fine.

Brian.

 

[Plecto]

"When designing a different layout, it is important to
return the load ground, the output compensation ground, and the low level (feedback and input) grounds to the
circuit board ground point through separate paths."

I'm not following at all. I thought it was the voltage drop across the power wires going to the LM1875 board and the proximity these wires has to the input pins that was the problem? Why would it matter where the ground from the load and input goes? These wires can't be the source of humming.

 

[betwixt]

It isn't the wires causing the hum, it's the tracks on the PCB.

What you have to understand is that the amplifier is sensitive to any signal difference between it's two inputs (pins 1 & 2). You have to make sure no signal other than the one you want to hear can appear across them, if i does, it gets amplified and fed to the speakers. You haven't posted a schematic so I'm going from your PCB image: you have a resistor from pin 1 to ground and a resistor/capacitor from pin 2 to ground but they don't have separate paths to a common zero point. Although they are electrically joined correctly, any current flowing through the ground track between them and to the loudspeaker will cause a voltage drop. The resistance is very low but suppose it's 1 milli-Ohm and 1 Amp flows through it, you get 1mV dropped and that is amplified by say x20 to give you 0.02V out of the loudspeaker, enough to make an annoying noise!

What you need to do, if you want to cure it, is split the grounds so that the supply ground, loudspeakers, resistor on pin 1 and the feedback capacitor meet together at one point but via different routes. Ideally, the supply decoupling capacitors should do the same. Doing this makes sure all the signals are a returned to a point at the same potential so no unwanted voltage can appear between them.

Brian.

 

[Plecto]

I'm beginning to get the picture. I actually have a gain of 40 now so I can see the millivolts making a difference What if I don't have a ground plane and rather chose a "ground spot" somewhere on the board, but away from the chips. I then run jumper wires from each spot (filtering caps, input, output, zobel network etc) to this single 'ground spot'? It won't be as neat and it will take longer time to make, but the layout will be easier.