Simple & easy modifications for NwAvGuy's O2 headphone amplifier

I published a number of modifications for NwAvGuy's O2 headphone amplifier on the headphone forum of a couple of years ago, here. In this section I'm posting the most simple & easy of the O2 modifications that would apply to the most people.

Your music source feeding the O2 headphone amp is too loud, even with the O2 is the 1x gain position. This problem is common with the O2 used as a desktop headphone amp with desktop music sources that output 2Vrms or more. Portable music players usually output just 1Vrms or less. Even with the O2's gain set at 1x you either get very limited amount of turn range with your O2 volume pot, or none at all.

The modification to solve the problem is extremely simple. The two 274 ohm input resistors on the O2 board (R3 and R7) are replaced with low current noise 4.99K resistors, Mouser 270-4.99K-RC, the same type of resistor used for R14 & R20. The input RF filter capacitors C11 & C12 are replaced by 9pF capacitors, Mouser # 81-RCE5C1H9R0D0A2H3B, to keep the RF filter coner frequency the same 2.6MHz.  The input capacitance of the NJM2068 is assumed to  be around 3pF, so the 9pF gives 12pF total.  Having the corner frequency this high effectlye filters out cell phone EMI, but doesn't cause any significant phase shift in the audio band with high(er) impedance sources, as NwAvGuy explains in his blog,. The result is forming a 1/3 / 2/3 voltage divider (2/3 passthrough, 1/3 attentuation) along with the O2's two 10K gain stage ground return resistors R14 and R20. In other words, your incoming source signal gets knocked down by 1/3 (attenuated), which solves your problem. Your O2 volume pot will have more rotation range.

Remember that even without voltage gain, or with 1/3 attenuation in the case of this modification, the O2 still provides current buffering via the output stage, which your headphones may require.

If you also perform the next modification for increased amplifier input impedance you will have to modify the part values for this modification.  For example, a 1/3 / 2/3 attenuation with a 20K input impedance, rather than the O2's standard 10K, would require a 10K series resistor, not 4.99K.  The RF filter capacitor would have to be decreased to just 3pF (Mouser # 81-RCE5C2A3R0C0A2H3B ), which combined with the NJM2068 assumed 3pF input capacitance gives 6pF total. 

Increased O2 headphone amplifier input impedance.  NwAvGuy wrote in his blog that the 10K input impedance of the O2 was a compromise between being too low, which will load down the audio source, and too high which will cause a small amount of noise pickup if the input audio cable is unplugged at the source end or the source device is turned off.

In actual practice, since the O2 was released, even 10K has proven to be a bit too low and does load some audio sources.  10K is the maximum load on the ODAC, for example, according to its specifications.  20K or 50K are probably better input impedance numbers.  NwAvGuy's concern about noise pickup only happens in situations where you are not listening to music, with the cable unplugged or source turned off.  There isn't much downside in increasing the O2' input impedance.

You might be wondering if the change will increase the Johnson (resistor thermal) noise pickup by the amplifier.  The answer is no, since the input impedance resistor is in parallel with the O2's input, not in series.  The resistor that would matter for Johnson noise is the series 274 ohm resistor (see the input attenuation mod above!).  By the way, increasing the input impedance with this modificaiton has no significant impact on the O2's RF filter corner freuqency, since it simply loads that filter.   This modification will have an impact on the input attenuation level with the modificaiton above, though.  

Another concern may be some increased DC output offset of the gain stage, due to input bias current from the NJM2068 being pulled through the input impedance resistor, which also is the input bias current ground return resistor.  Luckily this effect won't have any signidicant impact on the O2 due to the coupling capacitor after the gain stage.  The output stage is completely isolated from any DC offset from the gain stage (or any DC coming from the source, which is the cap's main function).   The one tiny effect will be the maximum output voltage swing decreased by the amount of increase in DC output offset, subtraced from the power rail voltage.  In other words, the DC output of the gain stage is now a tiny bit closer in voltage to one power rail or the other.  But this is such a small change it is negligible, as calculated next.

Two very typical headpone amplifier input impedances are 20K and 50K.  To make the change simply replace R14 and R20 on the O2 board with Mouser part number 270-20K-RC.for 20K, or 270-49.9K for 50K.  The average input bias current of the NJM2068 is 150nA from the datasheet.  The O2's standard 10K resistor would produce and input offset voltage of 150nA * 10K = 1.5mV, which then gets added to the chip's inherent average input offset voltage of 0.3mV for a total of 1.8mV.  With the new 20K resistor the math works out to be a total of 3.3mV, and for 50K 7.8mV.

Ground the pot faceplate and shaft to reduce hum and noise pickup. Grounding the pot shaft is very important to prevent noise and hum pickup. Using a nut on the pot shaft would be useless since the front panel is anodized on both sides. The anodization acts like an insulating layer. The nut and pot faceplate would not be able to make good contact with the panel for a ground.

The solution is a modification to run a wire from ground to a washer that can be slipped over the pot's threaded nose and the pot's nut clamped down on the washer. The easiest place to get ground on the same ground star segment is the metal case of the gain switch.

Solder one end of a small (24 or 22 AWG) stranded wire to the shell of the gain switch (or to one of its legs) then the other end to the edge of the washer that comes with the pot. The pot's faceplate is aluminum and won't solder, but the washer will take solder. A solder lug with a 5/16" or 9/32" hole can also be used. Then the pot nut can be tightened down over that washer or solder lug.

Potentially lower distortion gain stage resistors.  NwAvGuy may have over-done it a bit in his efforts to optimize the O2 gain stage for minimum noise.  He chose a 1.5K feedback resistor for the gain stage, but the NJM2068 datasheet shows the chip (lilke most older op amps) is really only designed for a minimum of a 2K load at low distortion levels.  Worse case is with 6.5x gain, which puts a 274 ohm resistor in series with the feedback resistor to ground for a op-amp load of 1774 ohms, less than the 2000.

Compounding the problem is that the 10K pot, which is being driven by the NJM2068 gain chip, winds up in parallel with that 1.774K resistor to ground, resulting in a 1.5K load.  Worse still, if you look at the AC model (within the audio frequency band) where the O2's coupling capacitor has nearly zero ohms, the output stage 40.2K op-amp ground return resistor winds up in parallel with the 10K volume pot, reducing the effective value.   The overall worse-case situation happens with the pot all the way up, where 10K || 40.2K = 8K.  8K || 1.774K (feedback resistor + ground return resistor) = 1452 ohms load on the NJM2068.

To solve this problem use one of these increased-value resistor sets to keep the load on the NJM2068 at 2K or above.  Increasing the resistor values adds a tiny amount of Johnson noise, but the effects on THD reduction are probably more important than the tiny noise addition. The first set is based on a 2.37K feedback resistor and 150pF compensation capacitor, which results in 2.1K load on the NM2068 in the worse-case 6x gain situation.  The second set of resistors, with values that might be easier to find (all resistors in both sets are currently available at Mouser Electronics) is based on a 3.01K feedback resistor and 120pF compensation capacitor with a worst-case (6x gain) load of 2.5K on the NJM2068.

Credit where credit is due! I had been thinking about this O2 problem for years, but sgrossklass on the heaphone forum made an excellent post (#4826) on the issue:

that finally motived me to quantify the whole thing here. :) Also please note that the title of this section says "potentially" because all of this is based on math and the datasheets. I'm not aware of anyone who has run a verifying dScope or AP test yet, but the odds of that coming up with a different result are low. This is pretty straight-forward circuit stuff. :)

Increased battery runtime. The part numbers to use are R25 = 2.1 meg 1/8W, Mouser #603-MFR-12FTF52-2M1 for $0.15, and R9 = 36.5K 1/8W, Mouser #270-36.5K-RC for $0.15.

The details about this modification are posted here in post #106 of my O2 modification thread at, with the resistor results at the bottom of that post. When NwAvGuy first released his O2 headphone amp design he used R25 = 2.74 meg and R9 = 40.2K. The result was only a 0.62Vdc difference between the power management circuit being "on" and "off". It turned out the battery voltage would rise on its own when the load was cut off. Often it would rise more than the 0.62Vdc, turning the PM circuit on again after a low-battery cutoff. To fix this he changed the values to R25 = 1.5M and R9 = 33K. This change helped by increasing the on/off voltage spread to 1.26Vdc, but at the same time went a bit too far and increased the turn"off" voltage to 7.07Vdc from 6.33Vdc. The result is slightly decreased battery runtime.

This modification changes R25 and R9 to produce on/off values in the midde between NwAvGuy's original and revised numbers. The net result is a 0.87Vdc on/off spread and a new "off" voltage of 6.72Vdc to restore some increased runtime over his final BOM values.

Old batteries are the #1 cause of the "motorboating" sound that can happen when the batteries are dead and the power managment circuit toggles on and off at a rapid rate like this. Old batteries tend to rise up in voltage when the load is cut off more than new ones. The Tenergy "9v" NiMH batteries are only rated for just 100 full charge/discharge cycles. After that the capacity starts to decrease. If you use your O2 a lot on batteries you should be putting in new batteries twice a year.

Replace the U1 gain chip witth the ($47!) MUSES02 chip. The NJM2068 gain chip that NwAvGuy used in the O2 headphone amp is really pretty good, as he showed by tests in his blog. Good enough that relatively few chips would yield any substantial improvements in measured numbers.   One chip to consider is also made by that same company, New Japan Radio, the MUSES02 available at Mouser as part #513-MUSES02 for a whopping $47 vs. just $0.60 for the original chip. There really is nothing in the MUSES01 datasheet numbers that jumps out at being a big significant improvement over the NJM2068, certainly not worth 78x the cost. The slew rate is 5V/uS but again with the NJM4556A output chips in place you still won't exceed 3V/uS. The MUSES02 is written up by NJR as their "flagship" audio chip with "excellent sound". I've read one posted review of the chip (an amp other than the O2) where the person was claiming a noticable sonic improvement. So there you go - for the "how it sounds vs. measurements" audiophiles out there, here is a chip to try and see what you personally think. :)

Moving the power jack to the rear. You can use a panel-mounted power jack (5.5mm outer, 2.1mm inner, fully insulated) to mount the power jack on the rear O2 case panel. There are a couple of details involved though. The jack you use should be fully insulated, in that the threaded "nose" shouldn't be connected to either of the jack power terminals. Those are harder to find. Most 2.1mm DC power jacks have a threaded metal nose that is connected to one terminal.

You don't use the side terminal out of the 3 O2 PC board power jack holes. Only use the two holes on the centerline of the jack to connect your wires. That side terminal is a switch inside the jack that isn't used. Same goes for the external jack - don't use the 3rd side terminal, if it has one. Then finally the two wires going from the O2 PC board to the external jack should be tightly twised to keep AC hum from getting into the audio. The wires should be at least 24Awg stranded and ideally 22 awg.

Higher capacity 300mAh NiMH batteries. NwAvGuy used the 250mAh capacity "9V" NiMH batteries in the O2 headphone amp. A higher capacity alternative exists from Maha Powerex, a 8.4V 300mAh "9V" NiMh battery. The 300mAh battery's datasheet shows a low 0.5 ohm impedance at 1kHz, a good thing. The battery is part number MHR84V and is available on or from the company's website at No modifications are needed to the O2 to use this battery. It just plugs right in. The same charging circuit in the O2 works just fine since this is also a NiMH battery.

NOTE - possible point of confusion! Maha also makes a 9.6V 230mAh battery that looks exactly like this 8.4V 300mAh unit. You cannot use the 9.6V battery though in the O2 without some modifications. Be sure to check the labeling on the battery to make sure yours is the 8.4V 300mAh version. For those folks into a more involved modification, I have posted what is needed to use the 9.6V 230mAh batteries on on my O2 modification thread at>headphone forum.

Entire website copyright (C) 2013 - 2017. All rights reserved.