Basic Troubleshooting for Headphone Amplifiers

So you’ve built one of the headphone amps described on this site and it doesn’t work...what to do? This article gives several common symptoms and some ways to find out what the problem is.

You will need a basic meter for most of these tests. If you don’t have a meter, troubleshooting is just a matter of guesswork, so click that link for advice on selecting a meter.

Preliminaries

If you haven’t cleaned the flux off the board, do so. Watch Tangent Tutorial #5 to see how to do this properly.

Before you start chasing your symptom, you should take some basic measurements:

1. Measure DC voltage at the amp’s power input pads. If your amp is powered from a wall supply, is the voltage what you expect? If it’s battery-powered, the voltage will differ depending on how fresh the battery is; if it is much lower than the battery’s rated voltage, change the battery and see if the symptom changes.

2. Measure DC voltage from V+ to V- of all the chips in the amp. (See the chips' datasheets to find out which pins these are.) Is it the same as what you measured previously, less any expected voltage drops in the power supply path? For example, in the PPA, D1 and the rail isolation circuits both drop the supply voltage by a small amount.

3. If your amp has an active virtual ground circuit of some sort — as all of the amp circuits on this site do except for the CMoy pocket amp — measure DC voltage from ground to the V+ and V- pins of each chip. If your amp has multiple “grounds,” use the one connected to the audio input jack(s). The two voltages should have the same magnitude, within about a tenth of a volt. If it’s farther off than that, the virtual ground circuit may be damaged or not hooked up correctly.

   (None of this applies to the CMoy pocket amp, because large virtual ground imbalances are normal in this design.)

4. Measure DC voltage from output ground to the amp’s left and right outputs. If there is more than about 20 mV, you need to avoid plugging headphones in for all tests until you correct this symptom.

5. With no music playing through the amp, measure the amp’s current draw. (Click the link for a how-to.) This gives you the amp’s quiescent current (Iq) draw. Is this value what you expect it to be? If not, the amp may be oscillating, or a part may be shorting out; if so, something’s probably getting hot. If you don’t know what current value to expect, it’s easy to calculate it.

If all of the above checks out, try a different source and different headphones. Does the symptom change?

Advice for Specific Symptoms

A problem in one channel but not the other

If the amp allows it, swap chips among the channels one at a time. (E.g. the buffers in a PIMETA or PPAv1, or the op-amps in a PPA.) If the symptom moves to the other channel, one of the chips you exchanged is probably damaged.

If you can’t swap chips because there is only one of a kind in your amp, try a different chip in its place. If the symptom just changes rather than going away, keep reading; the chip may not be at fault.

If you don’t have an extra op-amp on hand, Radio Shack sells the TL082 op-amp, which will work in any of the amps mentioned on this site that take a dual DIP-8 op-amp. (PIMETA, META42, CMoy pocket amp.) If it’s in a CMoy amp, you will need to feed the amp at least 18 V for this chip to pass a decent audio signal. In a buffered amp, the buffer dominates clipping behavior down to about 6 V where the 082’s clipping behavior takes over. If you don’t have a Radio Shack nearby or yours doesn’t have this chip in stock, similar generic chips are the TL072, the NTE858 and the LF412.

Quiet hiss in the headphones

The first step is to determine whether the hiss is being generated in the amp, or is coming from the source. If you unplug the source and the noise goes away, you’ve found your culprit: the source is just noisy. If the amplifier has a volume control at its input and changing the volume level results in a corresponding change in the noise level, that’s another strong indicator that the noise is coming from the source. The only thing you can do about this (short of getting a better source) is to reduce the amplifier’s gain, if you can. That way, you don’t magnify the source’s noise any more than you have to.

Reducing the gain can also help if the noise is being generated within the amplifier, because many of the noise sources within a headphone amplifier are directly affected by the amp’s gain. The optimal gain is where the volume control is at about the 3 o’clock position when the music is as loud as you normally listen. This gives you a little bit of extra control to turn the music up for quiet passages or the occasional rock-out session. If you’re using less than half of your volume control’s range, the gain is too high: not only does that raise the noise floor unnecessarily, it makes the volume control "touchy".

If you’ve selected your amplifier’s resistor values, another thing to check is whether they are too high. All resistors add noise to a circuit, with higher values being more noisy. This effect is called Johnson noise, after its discoverer. All you can do about it is lower the resistor values, if possible. You can play around with different values and see their effect on noise with my op-amp noise calculator.

If lowering the gain and resistor values is impossible or ineffective, try adding some resistance in series with the amplifier’s output. If the amp design allows it, try putting the resistor inside the feedback loop first. (This is R5 in the CMoy pocket amp, and R8 in the PIMETA, for instance.) Putting the resistor inside the feedback loop is often effective, and has the fewest bad side effects. The incentive is to use as low a value as possible, to avoid side effects. Start with 10 Ω, and work your way up to 100 Ω at most. If that doesn’t work, try adding the resistance outside the feedback loop: just put the resistor inline between the amp’s output and the headphone jack. This gives a stronger effect, but with more bad side effects.

If you’ve lowered the gain and added up to 100 Ω outside the feedback loop and the noise is still there, give up on this tack. The problem is probably something more serious, like a broken component, interference, or oscillation. Read on for discussion of these topics.

Noise in the headphones

This section is for diagnosing more serious noises than in the previous section: buzzing, loud hissing, humming, etc.

If the chips in the amp are getting hot, the amp may be oscillating. If you have access to an oscilloscope, you can simply look for oscillation. Another way to detect oscillation is to measure the amp’s current draw, and see if it is too high. If you believe the amp is oscillating, try adding bypass caps. You may have to solder-tack bypass capacitors into the circuit, if positions for them aren’t provided already. It’s best to use a film or ceramic cap for this, about 0.1 µF, with one cap soldered as close to each power pin as possible. Run the other leg to ground; this lead’s length is not critical.

If the chips aren’t getting hot and the amp’s current draw is reasonable, you’ve most likely got a grounding problem, a power supply problem, or interference.

The easiest of these three to rule out is wall power supply noise: simply switch to a battery power supply, if you can. If that makes the noise go away, the information in the companion article Op-Amp Power Supply Quality Considerations is likely to be on-point. If the noise remains, you can pretty much rule out power supply problems.

If the sound changes or goes away when you touch something metal on the amp, that suggests a grounding problem. The most useful advice on grounding can usually be found in the amplifier’s documentation. Some circuits have very specific grounding requirements that must be observed to get a working circuit. If the documentation doesn’t help you, search the various forums and the web for advice specific to that amplifier. Failing all of that, try going through reference material, such as Rane’s System Sound Interconnection application note.

The noise could also be some kind of interference. First try moving the amplifier to a different part of the room, or to a different room altogether. If that changes the noise, it’s picking up some noise within the room. That could be any number of things: RF hash from fluorescent lights or computers, AC line borne noise, etc. The best cure for that is to use a grounded metal case for your amp.

If the amp is wall powered and moving it doesn’t cure the noise, the AC power circuitry may be too close to the amplifier circuitry. Put some distance between the wall power supply and the amp. If you can’t put much distance between them, try adding a plate of some ferrous metal between the two circuits: iron, steel, nickel, cobalt, and specialized alloys like Mu-Metal. Only ferrous metals are effective in screening against magnetic interference, such as that from power supply transformers. Other metals such as copper and aluminum are only good for screening against RF interference, which usually isn’t audible in a headphone amplifier.

If the sound only happens when you touch the volume knob or the volume control’s shaft, you need to connect the pot’s chassis to the amp’s signal ground. (In all of the amps on this site, that means virtual ground or input ground.) If you’re using a metal case, be sure the case isn’t already connected to something else, like V- via the DC input jack. In that situation, you’d have to insulate the pot from the case or else you’ll connecct V- to ground through the case, which would be bad.

Loud click when you turn the amp on

Measure DC offset at the output of the amp. (That is, DC volts from output ground to output left and right.) Make sure your meter is on the millivolts scale, if it has the option. High DC offset will manifest as a loud click. 20 mV is tolerable, and lower is better. With high-end ICs in high-end amp circuits, offset should be down in the single-digit millivolts. (If you’re testing a PPA with bass boost, this is with the boost turned off; boost multiplies DC offset by the amp’s gain.)

If your amp has a ground channel (e.g. PIMETA or PPA) also measure DC offset from input ground to output ground. This number should be exceedingly low. I’d be bothered by 1 mV here, and would be sure of problems with 5 mV.

If DC offset isn’t the problem, it’s probably just a circuit turn-on transient. If you have an oscilloscope, measure the transient. If it’s less than a volt or two, it’s not worth worrying about. If it’s larger than that, something is probably wrong with the amp, since none of the circuits on this site should have large turn-on transients. The most likely source of problems is in the power supply circuitry. Try a different power source, if it’s a wall-powered amp. If that doesn’t fix it, look at the power rail caps, and work outward from there.

Chips are getting warm

If you can’t leave your finger on the problem chip for several seconds, skip to the next item. This test is about deciding whether a warm IC is too warm. Truly hot ICs are a different class of problem.

All ICs generate heat when running, so the thing to do is decide if you’re just feeling normal heat. The best way to decide this question is to measure the amp’s current draw without any load. Next, calculate the expected current draw, and compare the two. If the measured and calculated values are close, the heat you’re feeling is normal. If the value you measured is much higher than the number you calculated, yet no component is so hot that you can’t hold your finger on it, it may have a low-level oscillation. The only reliable way to test for that is to use an oscilloscope.

Socketed chips are getting truly hot

If you can’t leave your finger on a chip for more than a few seconds, it is probably either shorted internally, or it is oscillating.

First, don’t let the amp run too long with a hot chip. The chip may survive a short duration of overheating, but left on too long it will certainly die.

The easiest test is to replace the problem chip. If the new chip also gets hot, try putting it back into another circuit. If the previously hot chip works fine in another circuit, the chip itself is not the problem. Some other thing is causing the circuit to misbehave in the original circuit. Common causes are oscillation, or a short on an output pin.

If the amp is a PPA, that opens some interesting options. Here, the three-channel design means you have two identical sub-circuits, and a third nearly identical to those other two. Try swapping chips among channels. If the hot chip is still hot in its new position, the chip itself is the problem. Otherwise, the chip you moved into the problem location is probably now getting hot, meaning that the problem is elsewhere in that channel. The ground channel is a bit of a special case, since correct sound depends on both the signal channel and the ground channel. So, in your debugging, try to get the ground channel fixed first. Also, the ground channel is the only one running at unity gain: the requirements for stability are more stringent in this channel. Try swapping in a very tame chip here, such as an OPA227, or even a TL071.

Soldered-down parts are getting hot

Since you can’t swap soldered-down parts easily, the best thing to do is simply replace them if they’re getting hot. They’re either shorted out internally, or there’s something downstream from them drawing a lot of current. If the new part also gets hot, look downstream.

If multiple parts in a section of a circuit are all getting hot at once, don’t replace the parts. It’s likely that the problem is elsewhere.

Trace a Signal through the Amp

If all that fails, this test can often nail down the problem source.

You will need an oscilloscope or a meter. The scope shows more details, but it isn’t essential.

You will also need some kind of signal generator. There are dedicated units available, but the audio quality ones are expensive. For audio frequency, you can get better quality for less money with any digital audio source: PC, CD player, MP3 player, etc. The simplest method is with software that plays test tones out to the sound card. The free version of TrueRTA will do this on Windows, and on Mac, there’s Tone Generator X. If you don’t want to use your computer directly for some reason, the other way to go is to get a program that will write those same test tones out to a file. Then you can then burn them to an audio CD or download them to an MP3 player, preferrably in a lossless format. There are fewer choices for freeware in this regard. I only know of SigJenny for Windows, and nothing on Mac. If you’re willing to pay for it, there are several alternatives on both platforms.

The most useful tone for this test is a 1 kHz sine wave. If the program gives you a choice, you want the tone to be full-scale, or “0 dB.”

Attach the tone generator to the amp’s input jacks, and measure AC voltage at these points in the amp relative to virtual ground in both the left and right channels:

1. Solder lug on the inside of the input jack — At this point, you should be seeing the full voltage of the source. For a portable source, this will probably be somewhere in the 0.3 to 0.5 V range; if your source is connected via the headphone jack, be sure to turn the volume all the way up. If it’s a full-size source, the voltage will probably be more like 1 to 2 V. If the voltage is very much lower than these values, or 0, the input jack or the source is damaged.

2. Pad where the input wire attaches to the amplifier board — At this point, you should be reading a voltage very close to what you measured at the previous point, if not identical. If there is more than a tiny voltage drop, the wire is probably damaged.

3. Wiper of the volume control — The voltage should vary from what you measured previously down to nearly 0 as you turn the volume knob. Because a volume control’s taper isn’t linear, the voltage drop rate should increase as you turn the volume down.

4. +IN pin on the op-amp — This should be nearly the same voltage as at the volume control’s wiper, possibly less a small voltage drop if the amp has an input resistor. (E.g. R1 in the PPA, PIMETA or MINT.) If there’s a big drop, either there is a bad solder joint somewhere between the wiper and the +IN pin, or the input resistor’s value is incorrect.

5. -IN pin on the op-amp — This should be nearly identical to the voltage at the +IN pin. If it isn’t, probably one of the components in the feedback loop is broken or not hooked up correctly.

6. OUT pin on the op-amp — This should be the voltage you measured at the +IN pin multiplied by the gain of the amp. You should adjust the amp’s volume until the +IN voltage is about 0.1 V before doing this measurement. If the measured voltage is too low for the gain you’ve selected, lots of things could be at fault. If the voltage is zero, likely there is a bad solder joint or one of the chips in the amp is damaged. If the voltage is just lower than you’d expect from knowing the gain, perhaps the gain resistor values are incorrect.

7. OUT pin of the buffer, if any — This should be very nearly equal to the voltage you measured at the OUT pin of the op-amp. In the unlikely event that these are different, either the op-amp or the buffer in that channel is probably damaged.

8. Output jack solder lugs — This should be equal to the voltage at the OUT pins of the amp’s output stage. If not, the headphone jack wiring is suspect.

9. Output jack proper — Plug a mini-to-mini extension cable into the headphone jack and measure the voltage at the other end of the cable. Use the cable’s ground lugs, so you’re testing everything. If the signal isn’t getting to the other end of the cable without attenuation, the output jack may be broken.

How to Measure an Amp’s Current Draw

A standard amperage meter (ammeter) measures the current flowing through the meter. To measure an amp’s current draw, you must break the amp’s power supply circuit and use the meter to “heal” the break. The amp’s power flows through the meter, and you get a reading.

All meters I’ve used require you to move the red probe to a dedicated amps measuring jack. Many meters have two such jacks, with different current ratings. Use the higher rated of the two first; switch to the lower one only when you’re certain that the current is within its range. If you exceed the current rating of a meter’s ammeter jacks, you will either blow an expensive fuse or damage the meter. Trust me, I’ve done both, and it isn’t fun.

With a battery supply, measuring current draw is easy: just disconnect one battery terminal from its connector, and put the meter across the gap.

With a wall supply, things are a little trickier. One way is to desolder one of the wires going between the power supply and the amp, and put the meter across the break. You can also lash an external power supply to the amp with alligator jumpers and the meter.

How to Calculate an Amp’s Expected Quiescent Current (Iq)

Once you have measured the amp’s current draw, how do you know if it’s a reasonable value? With typical amps, it’s fairly easy to calculate the quiescent current that the amp should have when operating normally. (Quiescent current is most commonly symbolized Iq, so that’s what we’ll use from here on. Occasionally, you will see other symbols instead, such as ISY, for “SupplY current.”)

The amp’s total Iq value is the sum of the Iq values for all of its components. You will get an adequate estimate for most circuits simply by adding up the major current draws:

1. The biggest single current draw in most modern circuits is the ICs. The datasheet for each IC will give you the expected Iq value. Beware, this value often varies depending on the supply voltage; you may have to look at several pages in the datasheet to figure this out. Less commonly, you will come across chips like the BUF634, where Iq depends on the value of external components. Another common trap is in a dual op-amp datasheet, where sometimes the current is given per amplifier, and sometimes for the entire device. There is always at least a small range for Iq, if only because of manufacturing differences. Just try to figure out the most likely expected Iq value for each chip. We’re not trying to calculate the correct value down to microamps here.

2. If your amp has large discrete sections, the amp’s documentation should give the expected Iq. Often this is variable. For instance, in the PPAv2 design, the output buffers have trim pots that let you set the buffers’ bias current. Add up all of these currents.

3. Add up any remaining small currents of 1 mA or more, such as the power LED, and the current drain of any class A biasing devices.

Add all these values together. This is a close approximation of your amp’s total expected Iq.

If the measured value isn’t close, the amp could be oscillating, or there could be a short circuit. Damaged components often “fail short.” Oscillating and shorted components usually get hot. Follow that link for advice on what to do next.

Let’s work an example.

Say you have a PIMETA amp with an AD8620 in the OPALR position, an AD8610 in the OPAG position, single buffers on the outputs, and 220 Ω in all the R11 positions. The AD8610/20 datasheet says these chips will draw 2.5 to 3.0 mA of quiescent current at lower voltages, and 3.0 to 3.5 mA at higher voltages. 24 V is near the maximum supply voltage for this chip, so we’ll take the highest value, multiply by the three channels, and round off the fractional milliamp to get an even 10 mA. The BUF634 datasheet is a little trickier: you have to go to the graph on page 5 to figure out that with a 220 Ω bandwidth resistor, Iq will be about 8 mA. There are three buffers, so that’s 24 mA. There’s also the TLE2426; it draws well under 1 mA, so we could ignore it like we did the fractions above, but let’s add 1 mA to the total instead to roll up all these fractions into a single approximation. Finally, let’s add in another 2 mA for our power LED. The grand total is 37 mA.

We wouldn’t suspect oscillation if the measured value were a little higher, or even lower for that matter. Remember, this is an approximation, meant to test if what we’re measuring is even close to correct. If bad things were happening, you would probably see much higher values, over 40 mA in this example.

This space intentionally left blank. :)