capacitor: non polarized to polarized? (need answer fast)

I’m working on an audio amp circuit from an LM380. I haven’t had much luck with these chips, but FINALLY I figured out a problem: I hadn’t noticed on the schematic that 2x 0.1uf capacitors are POLARIZED and I have ceramic non-polarized caps! I think this is why my sound is distorted and scratchy.

The problem is I am going on vacation soon and probably won’t have time to order polarized electrolytics from mouser and install for a friend beforehand. Is there a way I can make the 0.1 cap polarized with a diode or something, or otherwise gimmick a polarized 0.1uf cap from a ceramic non polarized?

edit: I am using this schematic:

I found a 1uf capacitor but I’m scared of blowing it up, do you guys think it would be safe to try?

Where are the caps in the circuit? In general a polarized (typically electrolytic or tantalum) capacitor is not specified for any other reason than they are cheaper or smaller (or both) than the non-polarized equivalent. They typically have poorer characteristics in most other areas. Being polarized is a restriction on its use, not the converse.

Do you have a link for the schematic you are using?
Googling for a typical schematic for the LM380 I see that 0.1uF caps are used in the Zobel network on the output. I would be very loth to use an electrolytic here, as the self inductance of such a cap may lead to instabilities. Typically a foil capacitor with known low self inductance, or a ceramic cap is much less likely to give trouble. A NPO ceramic is better than the ubiquitous Hi-k ceramics, but in this circuit it probably makes no difference.

A scratchy and distorted sound can be symptomatic of an unstable amp, and these sorts of amps are notoriously picky about layout and decoupling. Is the amp make up on a known good printed circuit board design? I don’t suppose you have access to any sort of test gear?

Yeah I edited the OP with the schematic I am using. I have used LM380’s successfully once or twice before but EVERY time I had to fiddle with the grounding to get this weird scratchiness to go away, it is to make an arcade board go from line level to amped to power 8ohm speakers.

I got them working with these ceramics before (they are not that big actually). Test equipment I have access to is a multimeter, basically.

Sorry I am not an audiophile so I don’t know what “decoupling” means, and I don’t know for sure if it is a good layout, but I have had this problem before multiple times and I have to try a TON of minor changes with grounding etc before I magically get it non-scratchy for reasons I can never figure out… the 0.1uf caps I have now are ceramic, but you say that won’t make a difference so I’m stumped.

Oh, and I am using a 220uf capacitor instead of 250 because that’s what I have, should I try adding a smaller cap to it?

The grounding is the key. This is really part of the layout issue. You need to be quite clear about what is going on with the current paths, and in particular get the return paths right. It isn’t just a matter of joining nodes in the shcematic together - wire has inductance and resistance, and wires nearby on another have capacitance and mutual inductive coupling. What this means is that if you get the layout wrong you introduce new invisible components into the circuit, and can reach a point where the circuit goes unstable. Ultrasonic (or radio frequency) oscillation is the usual outcome. This is often manifest as distorted scratchy sound. It would be my first bet. The Zobel network is the capacitor and 2.7 ohm resistor in series that connects the output pin to ground. It is there to help keep things happy when driving possibly uncooperative (capacitive) loads. The power supply decoupling capacitor is the 01. uF cap from the 12 volt pin to ground. That is there to help keep the amp stable. It assumes you are driving the amp with a good solid 12 volt source that itself probably has some thunking big capacitors in it, or a battery. A small value cap here should really be a ceramic. It could be worth paralleling it with a medium value electrolytic cap - say 10 to 100 uF, but so long as your power supply is reasonable this won’t make much difference.

Overall I would lay the blame on layout issues, and most certainly not on the use of a non-polarised cap.

Capacitors are typically +/- 20% A 220uF and 250uF cap are functionally identical. That capacitor is used to block the DC offset on the output of the amp (there because the amp is powered off a single rail) and its size only really affects the low bass performance.

Francis Vaughan thanks so much for helping me late at night. My power supply is a good arcade power supply that is not driving anything but the amp right now (in this role it’s mainly passthrough for testing) so I know I have enough amperage. I added a 33uf cap in parallel at the same voltage as my 220, and you are right it didn’t help.

So I’m assuming the decoupling pin and the zobel network cannot share a ground, then? :slight_smile:

Another thing I noticed wrong: The diagrams I find from hobbyists do not have anything connected to pins 3-5, and 10-12 but on the PDF it says that these should be connected to a heatsink?

Over here it is now late at night on a Friday. Just home from a pleasant evening.

OK, some guidance. The simple circuit you have doesn’t really convey a lot of what is going on. It is sort of a mix of layout and schematic. The datsheet for the LM380 is available here. (PDF) There are a few points to be made. With a 12 volt supply and an 8 Ohm speaker the maximum device dissipation is a shade over one Watt. This is low enough that you can actually get away with no heatsinking of the package - although good practice and probably improved longevity of the part might suggest that some heatsinking is a good idea. The datasheet suggests that soldering the package to a printed circuit board provides good enough heatsinking if it has enough copper foil. Pins 3,4,5,10,11,12 are all a common ground pin - and soldering all these to a PCB where there is a nice slab of copper foil connected to them will provide significant extra heatsinking, enabling higher power dissipation… With your application I would not worry greatly. (I have seen applications where copper foil has been soldered directly to pins 3,4,5, 10,11,12 - but that was to make an amp using a much higher power supply voltage.)

Where you need to be careful is in the grounding and layout. The datasheet is clear that the power supply runs must be not be more than a couple of inches before you need the 0.1uF decoupling capacitor on the power pin. There is a clue here that the capacitor should connect right next to the pin on the IC, with minimal lead length. If you were constructing this dead bug style (ie the IC on its back and directly soldering to the pins) you would solder right to pin 14. The other end of the capacitor should go straight to ground.

Ground is the critical issue. The layout you have does not convey the crucial point. There needs to be one point in the layout that is the designated ground point. Ground is not a diffuse idea that you can link together all over the place. The place of this single point is pin 7, or a point right next to it. Everything in the schematic that is shown running to ground should connect directly to this point, and not go via some other circuit element that also is shown connected to ground. So pin 3 is connected directly to pin 7. The decoupling capacitor goes straight to the ground point, the shield of the input wire, the bottom leg of the volume potentiometer, the -ve connection to the speaker, and most importantly, the 0 volts of the power supply. Clearly this is going to get crowded, which is why it doesn’t actually all connect to pin 7, but all these runs should connect to a point nearby. Such layout is commonly called a star ground. It eliminates a host of problems, and I suspect that it will eliminate any stability problems your amp has, and remove the distortion issue. There is a good diagram at the bottom of this page, (although they show pin 6 coupled to ground, I would not do that. They have also missed the needed connection of pin of of 3,4,5,10,11,12 to pin 7/Ground, but the ideas are there.)

Its been a while since I have done this stuff, but I will throw this out there. I seem to recall some power supplies, especially switching power supplies, can get noisy if the load on them is too low. Your arcade power supply may have been designed to operate efficiently at a much higher power than you are currently using.

I think Francis Vaughan covered the capacitor and heatsink issues pretty well. It is critical that the 0.1 capacitor in the upper right be connected between pins 1 and 14 (Vs and GND) with as short of leads as possible. If there is any length of wire between these connections then the capacitor loses its effectiveness and you can end up with signal distortions, oscillations and instability, and other similar problems in your amplifier. This should typically be something like a ceramic disk capacitor or some other type that works well over a broad range of frequencies. You wouldn’t want to use an electrolytic type capacitor for example because its poor high frequency response would make it about worthless.

Is the board providing the sound from your arcade system isolated from the power supply? The LM380 has a differential input, so it’s measuring the difference between the audio signal and “ground”, and if your amplifier ground and the arcade board’s ground aren’t the same ground then this could cause you problems.

Also, since the circuit is set up to measure the difference between the audio output and amplifier ground, if the audio output has a DC offset in it (i.e. it’s driven by a single transistor output or something similar) you could have problems. A simple fix is to add a capacitor between the audio input and the 10k pot. And actually, depending on the output impedance of your arcade board, that 10k pot in series really may not be a good way of doing things. You could also be getting clipping just from the voltage being too high.

The input side of this circuit is a bit better, IMHO. The capacitor will filter off any DC bias present in the input, and the 50k pot will act as a better volume control than your 10k series pot because it ends up making a voltage divider instead of just adding a linear resistance to the input circuit. It will also present a more stable and consistent load to your arcade board.

Yes, this is true. Switching power supplies typically need somewhere around 5 to 10 percent of their rated max load in order to regulate themselves properly. Some switching power supplies have a built in load just for this purpose. Many don’t. They can not only get noisy, but can become unstable, and in some cases can even damage themselves.

Linear supplies are typically much less efficient than switchers, but can regulate themselves down to little or no load without any problems.

Wow, you would think that the schematic would include that if it was so important to have sound that wasn’t crackling and popping all over the place, that is more of a “rough guide” then a “schematic”. :mad:

I’ll try a dummy load later, good idea. :slight_smile:

Gonna try this too, thanks!

I’m making a box to convert a home console to arcade standard and I have video working, and of course the home console’s line level sound needs to be amped to run the arcade speakers, so I’ve always been confused about which grounds to use on the lm380 circuit, either the arcade connector ground connected to the power supply or the ground from the AV port on the game console, I think right now it’s connected to the arcade connector, maybe this is where my distortion is coming from! I wish I was at home and not at work :stuck_out_tongue:

My confusion is because there are three “grounds” in the setup, the arcade power supply ground, the pin on the JAMMA connector marked “Audio Ground” (which is usually looped to “Ground” anyway I believe) and the ground from the AV port on the game console.

I forgot to mention. I’m not using a pot, because what I am using the circuit for is just a box to go game console>arcade connector, and in the arcade cabinet it should have it’s own pot for volume, I don’t think this matters though.

To be fair, bypass capacitors are such a standard part of circuit design and most schematics just assume you know what they are these days. It’s not even that uncommon for a schematic to have all the 0.1 uF bypass capacitors all listed on a separate page nowhere near the ICs they are being used on.

Schematics also generally don’t say anything specific about ground layouts, even though these can be critical in the proper operation of some circuits. Sometimes you will see very specific notes about grounds or component layout on a schematic, but these are rare.

Yeah, grounds can get confusing. One thing to keep in mind is that the LM380 has a differential input, which means that the signal you are amplifying does not need to be referenced to the same ground as the amplifier. Just run the arcade board’s audio ground into the LM380’s INPUT- terminal and leave all of the other grounds on the amplifier circuit referenced to the amplifier ground.

Even though the arcade cabinet has its own volume control, you still don’t want to overdrive your final output stage. You don’t have to use a volume control type of pot here. You can use a little trimpot on the circuit board. Set the volume on the arcade as loud as it can do and adjust the trimpot so that it doesn’t distort on the output, then just leave the trimpot on that setting.

If you are mass producing something where the cost of every little component matters you don’t want to do it this way, but for hobbyists it’s a very quick and easy way to connect things without doing a lot of calculations to find out exactly how to match the signal level up to the amplifier.

What I meant was, for people to post the schematic on an arcade hobbyist website where I might know how to solder but not really care too much about the arcana of audio amplification (my exact scenario, can’t you tell? :stuck_out_tongue: ), it should list what every pin needs to be connected to and not make assumptions, or else overtly list that it’s a rough guide… in my mind “schematic” means follow this diagram and everything should work fine with no modifications?

I’m gonna hopefully figure this out tomorrow using the many good suggestions, and then issue some thank yous. :slight_smile:

Something that threw me the first time I looked at the simple schematic and only made sense when I read the fine print of the datasheet. Whilst the LM380 looks externally like a power op-amp, it has an internal feedback loop that defines a gain of 50, and is designed so that you can use it without connecting the -ve input pin at all. This is why pin 6 isn’t shown connected. They make a few notes about options of connecting it in various circumstances, but seem to default to leaving it unconnected. Do either that or connect it to the star ground

As to grounding, if you use a star ground, everything goes to the star ground point. Makes it very easy to reason about and usually avoids all the possible weird issues you may see. The suggestion about a trimpot is very good.

This is true for isolated inputs, but not differential inputs. For differential inputs, you usually need some kind of path between one of the input pins and the amplifier’s ground in order to allow some input bias current to go into the front stage. Even a high impedance path will usually work.

As an example, take a run-of-the-mill but modern thermocouple readout. It has differential inputs. If the thermocouple is completely isolated from the readout’s ground, it won’t work. Connecting a 100 KΩ resistor (or whatever) between the negative input and ground will fix the problem. (If your thermocouple workout works fine with an isolated thermocouple, it means there’s a internal resistor referencing the negative input to ground.)

Having said that, the above *might *not be applicable to a LM380; not sure. It would be applicable if it behaved like an op-amp configured as a differential amplifier.

Based on that circuit, the top 0.1uF cap is acting as power supply decoupler (commonly used to filter out fast transients in digital circuits). In this application, it can be eliminated without affecting the operation of the circuit. The lower 0.1uF cap is part of a Zobel network to present a flatter impedance to the output of the chip because of the inductive rise in impedance of a typical speaker. Again, in an application with a low quality power opamp like this, its effect is minimal and you could easily eliminate that cap and resistor and not notice any difference.

Here’s a much more typical circuit for that chip that will give you a couple of watts output, along with volume control and a DC decoupled input.

Not really. That power supply cap is going to be critical to maintain stability of the amplifier. The application note is quite specific about this too.

The schematic linked is almost identical to the application note, but adds the optional bypass capacitor on pin one that is used to mitigate noise injection from the power supply into the common leg of the differential pair. (Given the way the feedback loop is used to autobias the output, the chip will not have a good PSRR, and even with the bypass cap it is only a medicore 38dB. So adding this cap is not a bad idea.) The inductance of a couple of inches of wire between the power supply capacitors and the amp would appear to be enough to cause stability issues with this amplifier. Given the OP is clearly having stability issues anyway, deleting the PS decoupling cap is not really advisable.

Adding a decoupling capacitor on the input is something missed before, and given the application something I would strongly suggest be included. Indeed DC on the input could be enough to be causing a lot of problems. That also raises the question of overall system grounding. Although we have talked about ground with respect to the amp chip and its power supply, this can’t be looked at in isolation from the rest of the system, especially if, as I suspect, there are a number of separate mains powered power supplies involved. In such system it is possible to discover that there are very poorly defined return paths, and scope for all manner of noise injection and unexpected signal paths.

You’ve obviously never played around with an LM380. They’re pretty bulletproof, and although they have a tendency to severely asymmetrically clip the output, you could literally connect power, ground, input, and output with no other components and get usable output (less than 0.5W though).

It doesn’t really need a power supply decoupler cap, even with long power supply leads, unless you’re doing something like parking it inside an RF transmitter or wrapping the output wire around the power supply wires for inductively coupled feedback goodness (and when I say goodness I mean badness).

The circuit whose link I posted works well and is straightforward. It’s nearly identical to what I’ve used successfully. C1, C2, and C3 can be standard film caps (or electrolytics in a pinch), but I’d avoid ceramic or cermet types, especially for C1 (more distortion) and C3 (power dissipation).

Actually I have, albeit a very long time ago (back when it was a modern component.) However I have never tried violating the design note’s suggested use, and so can’t really comment about how well it will manage outside its recommended use. The datasheet isn’t very forthcoming about the more detailed specifications, so is it is a bit hard to judge how sensitive it might be to poor practice. As EECompGeek suggests earlier, I think an issue is that if you have some inherent idea of what you are doing, you naturally avoid a lot of pitfalls that are not documented but taken as read, and it appears that design is more robust than it is if you do blunder onto those pitfalls.

In the end the OP has a LM380 that is not behaving robustly, and is in need of help to make it so. Further, the symptoms are suggestive of instability. Although I worry about the lack of an input decoupling cap, it would be a rare consumer device that does not have one on the output, so this is likely not an issue, but worth checking.