I’m not gonna lie, I was a little disappointed with how my UFX modification project turned out.
I should clarify: for the obsessively minded, disappointment in a few little things tends to translate to disappointment in the whole. It’s a pretty killer mod, overall, and was a ton of work. But still, there were two things that irked me. The digital board of the UFX had crapped out, or rather the DSP engine (probably having to do with the big TI chip in the middle of the board). FX and DURec weren’t working, which I read on the RME forums might have to do with the RAM chip getting old. I replaced the RAM chip (tiny motherf****** pins and all) and there was no change in behavior, something else was going on, I settled with not having FX or USB recording. Replacing the DSP chip would involve a few hail Marys with the hot air gun, plus a reflashing tool with proprietary firmware only available at the RME factory. The wonderful people at Synthax US were kind enough to sell me a new UFX digital board (something they said they don’t normally do)… for $$$… but now it works!
The other disappointment, and the reason for this post, was the headphone amps. If you look at my previous UFX modification post, you’ll see I replaced the stock NJM4580 op amps with OPA1612 out of the PCM4104 DA converter, and high-current OPA1688 for the headphone drivers themselves (there are two op amps per channel, in parallel, for a total of 300mA output current per headphone). I reduced the output impedance from 30 ohms to 1 ohm since the OPA1688 have some built-in short circuit protection, and also because the lower impedance would allow the drivers to have a more linear frequency response across a wider array of headphone impedances, from IEMs to big 600 ohm cans.
But the headphone amps always sounded a little… flat? I’m not sure how to describe it. Compared to the headphone amp in my MixPre, which is truly excellent, there was less dimension, less articulation of details in recordings, and no deep, substantial bass, like there was still a veil across the audio. My modified HP4 also sounded much more engaging, three-dimensional… it’s hard to describe without falling into the common audiophile catchwords. Basically, the headphone amps in the UFX felt a bit compressed, kinda bland.
Thinking about the differences between these headphone amps—the MixPre, the HP4, and the UFX—his shouldn’t be the case. The DA converters are great, PCM4104. They operate at a +2.5V logic level, necessitating the use of DC blocking capacitors in the signal path, but I installed Panasonic FR bypassed with WIMA polypropylene caps, this shouldn’t really sound substantially *worse* than the HP4, which has DC blocking caps at the input. The other big difference in the UFX circuit layout is the lack of substantial, large-value power supply decoupling capacitors near the headphone drivers. The MixPre has a couple of 470uF capacitors on each voltage rail right near the headphone amp, the HP4 as well (which I bumped up from 10uF to 100uF, installing fast-responding organic polymer caps). These serve as local, low-impedance reservoirs of electrical charge when the op amp needs a quick boost of current, especially when trying to recreate big low-frequency transients like a kick or bass drum hit.
General-purpose Nichicon UVZ caps on the +/-5V rails in the MixPre 6; Max Linear XR8052 op amp is the headphone driver. Nichicon Organic Polymer capacitors (red caps) on +/-15V rails.
The UFX only has the standard 0.1uF ceramics from each voltage rail to ground around each op amp, so I decided to add some additional 560uF organic polymer caps I had in stock, soldering the legs directly to either side of the existing ceramic caps. News flash… with the 1 ohm resistors (through-hole types on surface mount pads) there’s basically no room for anything, let alone 8 capacitors in a space slightly smaller than 2″x1″, as you can see in the first photo. I ordered some low-noise, metal-film 1 ohm SMD resistors (which were not cheap). This made some room, as you can see in the second photo, but not much. In the process of doing all this I managed to break a pad from my previous work, which necessitated an expletive-laden repair.
as you can see, there’s not much room… there are op amps in there, I promise *$#!
That did it. F*** yes that did it. Opened it up. Low, articulative sounds, like a bass-drum kick, or pizzicato from a full section of double basses, sound real, impactive, huge, endlessly deep… how many more words can I pull out of my a**? There’s also more realism in the articulation of space, especially when recordings get loud—previously I felt like instruments blurred together at loud dynamics. Everything now feels clearer and more present, more detailed, balanced between treble and bass, even listening at a very soft level. Little details persist right down to silence, reverb tails feel real… No more cranking the volume! RME’s ADI-2 Pro, which uses the OPA1688 as its headphone driver, has an array of big power supply decoupling capacitors at the headphone outputs (Panasonic FM, looks like). My op amps were probably a bit current starved when they really needed it, especially with the low-impedance output and high-current op amps blowing electrons into the void like there’s no tomorrow, so music couldn’t really sound as dynamic as it actually was, hence the feeling of compression. This just highlights the importance of good power supply decoupling to get the best performance out of high-current circuits.

That in mind, I don’t think this is a problem for the line outputs, which, by their nature, are driving much higher-impedance loads (10k ohms, e.g.), and have a higher output impedance (75 ohms, if I’m remembering correctly). I’ve always thought my speakers sounded great.
Truly my best-sounding headphone drivers now, I’m in love with the sound, but this was not an easy mod… I’m obliged to say it: NOT RECOMMENDED!
Hi, Nice work. Did you have this article on another blog at one point? Did you ever get a chance to shoot out the FF802. Looking at modifying my 802 at some point.
LikeLike
I’m not sure?… but I think it’s been shared a few times because traffic will suddenly go through the roof 😛
I’ve never done a comparison to the 802, although the ADDA chips are AK models rather than Cirrus. Signal path might be a little simpler, simply because (if I’m remembering correctly) the AK chips don’t have a DC offset at their outputs, whereas the Cirrus chips have a +2.5V offset that needs to be filtered out with coupling capacitors. That’s what made my UFX mod a veritable capacitor city.
I’m still using my UFX, it sounds amazing. About to mix down another album…
LikeLike