A big "thank you" goes out to Currawong and the other mods for allowing this O2 headphone amp related DIY post!
This DIY project is a PC board I've whipped up that replaces the two NJM4556A output chips on the O2 headamp with an OPA140 + LME49990 pair on each channel. The board also has a few (optional to populate) O2 modifications I've posted elsewhere over the last couple of years. The board fits in the top slot of the standard B2-080 headphone amp case. The parts are all surface mount and fit between the board and the top of the case. I've kept the minimum SMD part size to 1206 though (larger) to make it as DIY friendly as possible.
The OPA140 is a DC precision op amp, meaning it has a very small input offset voltage. One of my goals with the project was to reduce the O2's output DC offset voltage (DC going into the headphones) from the typical 3mV. In a photo at the link below you can see that the result of this board I measured at 10uV on one channel and 36uV on the other. That is MICRO volts, vs the 3mV = 3000uV normally with the O2, a 98%+ reduction in output DC offset.
The OPA140 is a jfet input op amp with a miniscule input bias current that won't cause a large voltage drop across the O2's 40.2k ground return resistor on the inputs of the NJM4556A's. That IR resistor voltage drop caused by the NJM4556A input bias current reflects to the chip's output (being a unity voltage gain buffer stage) and accounts for about 1/2 of the O2's typical 3mV output DC offset voltage. So the OPA140 being jfet input is another factor in why the output DC offset voltage of this board is so low. The 40.2k resistors in the O2 still can't be reduced though since they would load the 10K pot.
For lower resistor values in the center of the O2 circuit, take at look at my version of an O2 Desktop Amplifier (ODA) DIY project boards that I wil post in another thread sometime. That ODA headamp uses LME49990 gain chips and takes full advantage of their datasheet rated ability to drive a 600 ohm load to reduce the pot in the middle to 1K (plus feedback loop resistors equals about 750 ohm). The NJM2068 gain chip in the O2 headamp is only good for driving loads down to 2K loads, as per the datasheet. Lowering the pot value reduces Johnson noise, and in turn allows the ground return resistor on the output stage to be lowered from 40.2K to 4.99K. That amp has 6 NJM4556A chips in parallel on each channel (SIP package rather than DIP to dissipate more heat) to increase output current and increse the S/N ratio (signal adds linearly in parallel but noise adds by RMS). The resulting background noise level is just dead silent, even at the full gain setting.
The OPA140 chip in the O2 booster PCB is also very battery friendly at just 2mA idle (quiescent) current. The LME49600 audio buffer is being run in "low bandwidth" mode (which is still 110mHz, way more than is required for audio), which pulls just 7mA. Together the OPA140 and LME49600 pull just slightly more current than the NJM4556A chip they replace, having a very tiny impact on battery life.
The datasheet THD (distortion) graph of the OPA140 is about 8x better than the dScope analyzer numbers the person who designed the O2 published. The distortion levels rise with frequency, so that "worse case" of 8x better occurs at the 20kHz end. At lower freqencies the numbers are even better. I don't have access to a dScope or Audio Precision analyzer unfortunately, so these are purely off the data sheets, NOT measuured. So bottom line, there is also a fairly good possibility that the THD levels using this O2 output booster board are lower than with the stock O2 headamp.
The O2's designer noted in his postings that the THD levels of the two paralleled NM4556A output chips swamp that of the NJM2068 input gain chip, making them the limiting factor on THD and making it pointless to use an even lower THD chip for gain, such as the LME49720 or LME49990 on an SOIC adaptor. This upgrade board solves that problem, at least up to 8x lower THD or so, meaning a lower distortion chip can be used in place of the NJM2068 and the lower THD would make it through the output stage.
The upgrade board also allows for higher voltage levels for higher impedance and low-sensitivity cans. The O2's designer noted that although the NJM chips are rated for +/15Vdc power rails, the higher voltage (the O2 normally runs on +/-12Vdc rails) could cause excess power dissipation in the two NJM4556A output chips under some load conditions. Both chips on the upgrade board are fully specified in their data sheets for +/-15Vdc rails. The LME49600 output chips are well heatsinked via thermal vias to the foil on the back of the PC board (4 layer board), so runing at +/-15Vdc is no longer a problem. The bill of materials (BOM) and build instructions for the project at the Google Drive link below have instructions for that mod. The O2's voltage regulators have to be changed, along with one capacitor and the battery charging resistors.
A +/-18Vdc modification is also possible with a little more work. The LME49600 has to be replaced with the higher voltage LME49610 version. The NJM2068 gain chip on the O2 has to be replaced with the higher voltage LME49860 chip (which also has better THD numbers, it is just the high voltage version of the LME49710). The LME49990 on DIP adaptors can also be used since it is fully specifiec in the data sheet for +/-18Vdc rails.
The board includes a couple of other O2 related modfications. One puts two resistors across the power rails after the O2 mosfets to reduce turn off thumps. Another puts LEDs in series with those to show when the O2's mosfets have turned on (after the O2 power management circuit) and if each of the two power rails is running correctly. Another circuit is a latch for the O2's power managment circuit to prevent oscillations when the batteries get low. And finally there is a spot under the board to mount a 1/4" Neutrik jack upside down, although if that one is used the taller B3-080 case must be used.
The mod on the board with the 2 LEDs is a useful one. The one existing LED on the O2 is before the mosfets and power managment circuit. When it is on it says nothing about whether the mosfets are on and/or whether both power rails are up. I've helped a bunch of folks trouble shoot their O2's over the past couple of years and that question always comes up: are the mosfets "on" to supply power to the amp circuits. It is never obvious and DMM readings are required. These two new LEDs only come on when the O2 mosfets are on, one LED per power rail, so right there both troubleshooting questions get answered automatically. If the new LEDs are both off your mosfets are off. If one is on and the other off, the one that is off is a missing power supply rail.
The schematic, layout, BOM, photos and build instructions are all at this Google Drive link:
If some additional testing goes well I'll post the Gerber PC board files. Those Gerber files can be sent out to any PC board fab house that does 4 layer boards. I've been using Seeed Studio. I've posted instructions for submitting the files to a fab house in the link. 20 4-layer boards with blue solder mask were $140, including the $37 or so worth of DHL express shipping. That works out to about $7 a board. Their minimum quantity is 5 boards, although the cost per board goes up with smaller lots.
Have (DIY) fun!
EDIT 12/3/2013. Added mention of the OPA140 chips being jfet input and the effect. Added mention of the ODA and the ability to drive a 1K pot in the center for lower Johnson noise. Expanded on the two LED optional modification. Cleaned up some typos.
Edited by agdr - 12/3/13 at 6:17pm