Cmoy With Gain

BACKGROUND: I recently reviewed a $39 Cmoy purchased direct from China on eBay and was somewhat impressed except for one obvious problem: It had a gain of “1X” or 0 dB which rendered it useless for most applications.

NOW WITH GAIN! Fixing the gain problem required replacing 4 resistors. The photo shows the original resistors removed and the 4 new ones laying above the Cmoy’s circuit board (click for larger version). The Tech Section goes into more detail on the resistor values, etc.

HISS & NOISE: Even using my most sensitive headphones, Ultimate Ears SuperFi Pro 5s, and even with 13 dB of gain, the Cmoy is still very quiet when driven from a low impedance source. Subjectively it’s even quieter than the Mini3 at some volume settings. There’s a “pop” when you first turn it on that’s not that bad—especially with more typical headphones. There’s a softer click when you turn it off.

SUBJECTIVE SOUND QUALITY: With a proper amount of gain the 4556 Cmoy sounds at least decent even with my Sennheiser HD650s. It has plenty of power for the 650s and also my 80 ohm Beyer DT770s using my iPod Touch or Sansa Clip+ as a source. Without the Cmoy, and the gain modification, neither source works well for those headphones. So the modified Cmoy is a big improvement. I might conduct some blind listening tests in the future to find out just how good a $39 amp can be.

MEASUREMENT SUMMARY: As expected from using less feedback there’s more distortion but the measurements are still respectable and better than the Mini3 in several areas. In particular, the Cmoy has a much higher output voltage than the Mini3 with roughly five times more power into high impedance loads. That could make a big difference with some power hungry high impedance cans. And the Cmoy’s real ground design is free of the serious interchannel distortion problems that plague the Mini3. It also has less distortion at high frequencies, a lower output impedance and much better channel separation (less crosstalk). While the noise measured 14 dB worse compared to the gainless version it’s still slightly better than the Mini3 (both are fairly quiet subjectively):

Measurement	Cmoy 13 dB Gain	Cmoy 0 dB Gain	AMB Mini3
Frequency Response	+/- 0.1 dB Excellent	+/- 0.1 dB Excellent	+/- 0.1 dB Excellent
THD 1 Khz 150 Ohms	0.002% Excellent	0.001% Excellent	0.002% Excellent
THD 1 Khz 15 Ohms	0.030% Good	0.003% Excellent	0.017% Good
THD 20 hz 15 Ohms	0.028% Good	0.005% Excellent	0.01% Very Good
THD 20 Khz 15 Ohms	0.040% Very Good	0.02% Excellent	0.45% Poor
IMD CCIF	0.010% Good	0.003% Excellent	0.043% Fair
IMD SMPTE	0.012% Good	0.003% Excellent	0.009% Very Good
Noise (ref 400 mV)	-89 dB Very Good	-96 dB Excellent	-94 dB Excellent
Max Output 15 Ohms	100 mW Excellent	67 mW Very Good	104 mW Excellent
Max Output 150 Ohms	180 mW Excellent	180 mW Excellent	38 mW Fair
Output Impedance	0.67 Ohms Very Good	0.1 Ohms Excellent	0.9 Ohms Very Good
Crosstalk 15 Ohms	65 dB Very Good	68 dB Very Good	40 dB Poor
Channel Balance Error	1.1 dB Good	1.1 dB Good	1.14 dB Good
DC Offset	21 mV Good	4.5 mV Excellent	4.1 mV Excellent

FIRST CLASS:

• $39 for a complete portable amp that can drive high impedance headphones is a bargain
• Respectable distortion performance into 50+ ohms
• Real ground is free from virtual ground/rail splitter problems
• DC blocking capacitor on input helps protect headphones
• Simple design is easy to modify (no surface mount components)

ECONOMY:

• Headphones could be damaged if one battery becomes disconnected
• Headphones could be damaged if one battery dies before the other
• Lacks AC power option, batteries must be removed for charging
• Marginal distortion measurements with 16 – 50 ohm loads
• Limited output current into difficult loads
• Some channel balance error (cheap pot)
• DC offset could be marginal with some components
• No RF protection

BOTTOM LINE: While adding a suitable amount of gain degraded the distortion it’s still respectable—especially considering the price. The distortion is also much less likely to be audible than the Mini3’s virtual ground related distortions. My biggest concern is possible headphone damage if you ignore the warning signs (bad sound) and run mismatched batteries too low (see the original Cmoy article). Or if one battery becomes disconnected (or isn’t connected properly in the first place) the result could be tragic. This design also has some DC offset issues and higher distortion than I would like to see into low impedances. But, overall, this amp is a relative bargain for headphones around 50 ohms and higher. This amp generally matches the bargain FiiO E5 into lower impedance loads but this Cmoy is capable of much higher output voltages for driving higher impedance cans properly. The E5 struggles, for example, with the HD650s but this Cmoy handles them easily.

---

TECH SECTION:

THE ORIGINAL PROBLEM: The original Cmoy had a 470 ohm feedback resistor (R1) with 100K to ground on the negative op amp input (R2). This resulted in very nearly 100% feedback and essentially unity gain.The original design is shown in the schematic  to the right.

MODIFICATION: Resistor R1 was increased from 470 ohms to 1.3K to avoid excessive loading and R2 was reduced from 100K to 360 ohms to provide the proper gain. The result can be calculated as follows:

Gain = 1+ (1300/360) = 4.6X = 13.2 dB

This means a typical 500 mV RMS input would yield a healthy 2.3 V RMS output—plenty to drive even full size Sennheisers like my HD650s.  R2 could be as low as about 200 ohms for even more gain if desired or increased for less gain.

IT’S A FLOOR WAX AND A DESSERT TOPPING: In a classic Cmoy design, such as this one, there’s just one op amp to do all the work. In the gainless Cmoy it only had to serve as a unity gain buffer and drive the headphones. With gain, however, it has to serve as the gain stage and also as the output stage. That’s less than ideal as most of the distortion is generated in the output stage but now there’s substantially less feedback to correct that distortion. It’s sort of like a front wheel drive car where the same set of tires have to provide the steering and put all the power to the pavement. It’s a compromise which is why the highest performance cars are rear wheel, or all wheel, drive. And there are more problems with having one op amp do everything. Such as DC offset.

DC OFFSET: Op amps, especially ones with bipolar inputs like the 4556, often have a significant amount of input bias current. In this design, the bias current flows through R3. The datasheet lists the current as 50 nA typical and 500 nA worst case. So V=I*R gives 50nA*100K Ohms = 5 mV. With a gain of 1 that yields 5 mV DC offset on the output. I measured about 4 mV. With a gain of 4.6, however, 5 mV becomes 23 mV and I measured about 21 mV. But the worst case number of 50 mV becomes 230 mV or nearly 0.25 volts. 23 mV is not much of a problem but 230 mV is too high. This is another downside to having a single op amp do everything. Less gain clearly helps here so, especially in a design like this, you only want as much gain as you really need.

BIAS CURRENT TRADEOFFS: You can lower the value of R3 to reduce the offset voltage. But that compromises the –3 dB low frequency roll off caused by C5 and R3 and, if you lower it too much, will hurt the bass response and create more phase shift. R3 could be lowered to 22K and the –3 dB point would still be plenty low with a 2.2 uF capacitor. That would improve the DC offset by a factor of 5 and is worth doing. So it would be around 2 mv – 5 mV typical which is fine but worst case could be as high as 50 mV which is marginal.  Trying to improve it further, however, involves more compromises. If you swap the 4556 for an op amp with less bias current you will have a hard time finding one with similar levels of output current (the main strength of the 4556). Even the expensive AD8397 used in the Mini3 has similar bias current specs to the 4556. So you would give up output current for lower DC offset—probably not a good trade off. You could make C5 bigger, and then R3 can be even lower, but a much bigger film cap is expensive and won’t fit in the tiny Cmoy enclosure so you would have to use an electrolytic which have more distortion. It also lowers the input impedance and changes how the volume control works. So 22K, or possibly 10K, is probably about as low as you want to go with R3.

NO GAIN VS GAIN: The graphs below tell most of the story with respect to what adding gain does to the distortion performance.  The short version is, into low impedance loads, there’s about 10 times more distortion. But the distortion was so low to begin with, this isn’t as bad as it sounds. The details can be found below.

THD+N vs OUTPUT 15 OHMS: The bad news is there’s about 10 times more distortion into 15 ohms at all power levels. The good news is it’s still fairly modest below about 0.03% and remains fairly flat until clipping versus the steeply rising distortion of the Mini3 into the same load. Shown below is the original amp without gain in red and the modified amp with 13 dB gain in blue. The slightly higher clipping voltage is due to the batteries being more fully charged not the gain change. It hit 100 mW at 1% THD:

THD+N vs OUTPUT 33 OHMS: Here’s the difference at 33 ohms. At typical listening levels for 33 ohm headphones (i.e. < 1 V RMS) the distortion is still below 0.01% (i.e. the distortion products are all below –80 dB). This also outperforms the Mini3 (again, the higher clipping voltage is due to fully charged batteries):

THD+N vs OUTPUT 150 OHMS: This is as above but into 150 ohms. This time blue is the original and yellow shows the result with 13 dB of gain. Even at 2 volts RMS the distortion is still very low at about 0.004% but it is higher across the board as expected:

THD vs FREQUENCY 15 OHMS: This compares the the gain version (blue) to the no gain version (yellow) into 15 ohms from 20hz to 20 Khz. Again, the distortion is about 5 – 10 times higher but still impressively consistent across the band and still relatively modest (the droop above 10 Khz is due to the analyzer bandwidth used):

CCIF IMD 15 OHMS: The CCIF IMD shows higher sidebands around the high frequency signals up to almost –70 dB. This is still decent performance, and still better than the Mini3 in some ways, but notably worse than the no gain version:

SMPTE IMD 15 OHMS: This is also higher, as expected, but still a decent result considering the 15 ohm load. The relatively benign 2nd harmonic of the 60 hz tone (which is not considered IMD) is over –80 dB but everything else is comfortably under –80 dB:

GAIN & SPECTRUM: Here’s the gain and spectrum at 1 volt out into 100K at 1 Khz:

NOISE: As expected, the noise performance is much more modest but still very respectable. The 60hz hum component is likely due to loop area on the PCB layout and is limiting the unweighted measurement. This hiss is mostly below –130 dB making it subjectively very quiet:

OUTPUT IMPEDANCE: 400 mV no load (100K) at 1 Khz dropped to 383 mV which gives an output impedance of 0.67 ohms. With no gain, it measured about 0.1 ohms. The higher impedance is due to less feedback and the relatively high inherent (open loop) output impedance of the 4556 op amp. Anything under 2 ohms is fine so this still isn’t a problem.

CHANNEL SEPARATION (CROSSTALK): The crosstalk was a few dB worse but was still excellent at around 80 dB into 150 ohms and better than 60 dB into 15 ohms. The dominant factor, by far, is the output jack and ground traces on the PCB. So changing the gain didn’t make much difference.

CHANNEL BALANCE: This was unchanged as it’s related to relative errors in the volume pot and is independent of gain.

PHASE: The phase response was unchanged and still excellent.

SQUARE WAVE RESPONSE: The square wave performance was still excellent and the slew rate remained faster than 3 V/uS.

TECH SECTION SUMMARY: While the distortion and noise numbers have gone from being fairly amazing to merely decent this is still a respectable performance for $39. But there are some concerns. My particular 4556 has about 20 mV of DC offset in this configuration which is fairly high but still acceptable. But the worst case values on the datasheet could yield much worse. So I would suggest measuring the offset of any similar 4556 Cmoy. If need be, R3 can be lowered to 22K or even 10K as explained under DC Offset at the start of the Tech Section. It’s also worth noting not everyone needs 13 dB (4.5X) of gain. If you can live with say 2X (6 dB) gain with your source and headphones, you would get significantly better performance (it would roughly split the difference between the no gain version and the 13 dB gain measurements). At higher than 13 dB gain the distortion would start to be more of a concern with this single stage design. As explained earlier, you can only expect so much of something that’s a floor wax and a desert topping.