DRAMA? (revised 3/16/11) Ok, so nobody is going to want the movie rights, but my experience with the NuForce uDAC-2 had some unexpected drama. When I started making measurements of the uDAC-2, it was doing so poorly on some tests, I stopped. Had I received a defective one? I contacted NuForce and sent some of my preliminary measurement results. That turned into 16 emails back and forth with NuForce finally concluding:
“Your uDAC-2 is within spec, and both of our measurements are correct, just that we are looking at different part of the plot. If you are going to publish your report, I appreciate that you also publish our response, which you will find that we did not dispute your finding, but it makes a compelling case why we TUNED uDAC-2 this way.”
LISTENING TEST (added 3/16): Because NuForce has strongly argued they designed the uDAC-2 to sound good, despite the poor measurements, I set up two different listening trials to let anyone download the files and listen for themselves. The results can be found here: Nuforce Listening Test Results
JUST TO BE CLEAR: I have no other experience with NuForce, and I’m in no way affiliated with any of their competitors. I just bought one of their DACs for my personal use. There’s no hidden agenda here and the subjective comments are only my opinions. The measurements are readily reproducible (and many were verified by NuForce themselves).
DESIGN BY EAR? (added 3/9/11) Apparently, based on comments regarding my measurements from Jason Lim the CEO of Nuforce, the company prefers listening to measuring. For example, in response to this review, Mr. Lim posted a public comment (that he references in the comments section of this article) that includes: “better sounding product still win hands down over better measured product” and “If someone choose to listen with their measurement equipment, there is nothing we can do.” Here are my observations:
- THIS HAS HAPPENED BEFORE: NuForce’s “listen first” design philosophy hasn’t always worked out so well. Thanks to Mainelli’s comment at the end of this article, I learned (3/9/11) John Atkinson at Stereophile used an Audio Precision analyzer to measure the NuForce CDP-8 last year. He found the NuForce to have among the highest amounts of jitter he had ever measured, as well as unusually poor noise modulation and predicted the sound of the CDP-8 likely suffers as a result. Interestingly, I also measured very high jitter with the uDAC-2.
- NuFORCE’s RESPONSE: Jason Lim from NuForce responded to the Stereophile Measurements by saying NuForce believed the performance was “addressed by the design”, they made “decisions based on listening”, and “the measurement was a surprise to us” because it seems they never measured the jitter, or likely the noise modulation, themselves.
- SURPRISES: Apparently my measurements were also a “surprise” to NuForce—just like the ones made by Stereophile. I am concerned when companies that claim to offer superior products don’t even seem to bother (or perhaps have the capability) to measure their basic performance. You can take the finest DAC chips and other components in the world, throw them on a great looking PC board, wrap it in a fancy high-end enclosure, and you will likely get very poor performance. Jitter and noise performance, in particular, are very sensitive to the PC board layout, grounding issues, EMI, etc. So if you don’t properly measure your product, you really don’t know what you’re going to get. And, it’s my opinion, that’s what happened with both the uDAC-2 and the CDP-8 tested by Stereophile.
- TOTAL PERFORMANCE: The ultimate performance is not merely “addressed by design” as Mr Lim is quoted in the article. That would be like Boeing delivering airplanes without ever testing them because they were confident in the design and the plane simply “felt good” to the pilot. As the multi-year delays with the Boeing 787 Dreamliner show, there are many problems that often don’t show up until you conduct the right tests.
- ANOTHER “DESIGN BY EAR” EXAMPLE: I bought a power amplifier from a high-end manufacture similar to NuForce in some ways. After I started listening to it, the tweeters failed in the speakers it was driving. I measured the amplifier and discovered an overly protective current limiting circuit (triggered by my 4 ohm speakers) that created nasty high frequency bursts of powerful and inaudible noise. It was these bursts of ultrasonic noise that destroyed the tweeters. And in contacting the manufacture, much like with NuForce, they seemed completely unaware their amp had a problem. All they could talk about was how it sounds great and pointed to subjective reviews (without measurements) that also said it sounded great. Well, unless perhaps you’re a dog, you can’t hear a 25 Khz burst that destroys tweeters—that requires proper testing and measurements. Their amp, like NuForce’s DAC, also didn’t meet its published specifications and apparently had never been properly measured.
- WHY “DESIGN BY EAR” OFTEN DOESN’T WORK: In my experience, most companies who “design by ear” to varying degrees also don’t believe in blind listening tests. Whenever their equipment is subjected to a blind test, and doesn’t do well, they are very quick to provide reasons why such testing isn’t valid. Consequently, they usually don’t conduct their own blind listening tests. Instead, they use “sighted” listening. And it’s been widely shown sighted listening is significantly flawed, biased, and misleading. If you’re skeptical, please see some of the links in the right hand column of this blog such as Dishonesty of Listening Tests. So, in summary, it’s fairly safe to say most of these manufactures who “design by ear” are designing in a way that’s signficantly flawed and biased. All the proof is already out there. It’s not the best way to design equipment—especially if you also skip making proper measurements as NuForce apparently has.
WHICH USB DAC? I needed a small, USB powered DAC for headphone use with a secondary PC. Here are some I considered and my thoughts on each:
- Behringer UCA202 – My first idea was the Behringer UCA202 as I already had one laying around and it’s nice and small. So I wired it to the dScope to check it out. It measured surprisingly well in almost every way—impressive enough I decided to turn it into a UCA202 Review. But, unfortunately, the headphone output impedance was too high for my preferred headphones. Strike one.
- E-Mu 0202 – I also had an E-Mu 202 already. But a quick check showed the headphone output impedance is 22 ohms. That’s a better number, but still not low enough to work well with my preferred headphones. The E-Mu is also kind of big to try and hang off the back of a slim all-in-one panel PC. And it needs its own proprietary driver that has caused problems on some PCs. Strike Two.
- M-Audio Transit – The Transit is well regarded, small, and I also already had one. But, sadly, while M-Audio claims the 3.5mm Line Out jack also works for headphones, it has the same 50 ohm impedance as the Behringer UCA202. It also, like the E-Mu, needs its own potentially troublesome driver. Strike three for what I already owned.
- FiiO E7 – The $99 E7 gets some good reviews and is relatively popular. It uses a digital volume control to supposedly maintain good tracking between the channels. It can also act as an analog headphone amp, has built-in EQ and an optional desktop “dock”. But, from what I read, it forces the Windows volume control to maximum. While this delivers a bit accurate stream, if I want to mount the DAC on the back of the PC so using only the up/down volume buttons on the E7 wasn’t a viable option.
- Firestone Audio Fireye2 – The Firestone Fireye2 is a little $99 USB DAC from the Netherlands. It has a nice full metal enclosure and has received some good reviews with the exception of some slight hiss and channel balance problems at low volume settings. But I didn’t see anywhere if it worked with the Windows volume control.
- Leckerton UHA-4 (updated 3/2/2011) – The Leckerton UHA-4 is well over my budget at $169, but it’s small and also portable (it has a Li-Ion rechargeable battery). It has some appealing features including adjustable gain (see: Headphone Amp/DAC Gain), crossfeed, and Leckerton claims excellent channel balance tracking (to 0.1 dB) and impressive specs. The bad news is, according to Leckerton, it has a 10 ohm output impedance. That’s still unacceptably high for use with many balanced armature headphones.
- NuForce uDAC2-hp – The $99 “hp” version of the DAC reviewed here supposedly uses the same problematic volume control as the original (1st generation) NuForce uDAC. The problems with the older volume control are well documented—i.e. uDAC2-hp Channel Balance.
WHY THE uDAC-2? So Goldilocks tried a few bowls of porridge, and considered several more, and as luck would have (or it turns out not), I finally chose the NuForce uDAC-2. A big reason was NuForce proudly claimed on their website:
“As a step above the uDAC, the uDAC-2 boasts a highly linear TOCOS volume control for improved channel tracking at low listening levels in addition to 24bit/96kHz USB DAC and an improved headphone amp.”
The statement above led me to believe NuForce had fixed the often reported channel balance problem of the original uDAC (and the newer “hp”). The uDAC-2 specs on their website also looked good (but they have since revised them to be more realistic after our e-mail exchange—more on that later). And NuForce seemed like a good company. Unlike some boutique DAC manufactures, they have an office in the USA, a decent website, and people have said good things about their customer service. So I ordered one.
IT ARRIVES: The uDAC-2 arrived in a professional looking package and seemed nicely made with a solid metal enclosure and decent build quality. It’s really small—the footprint is smaller than a Sansa Fuze. It has a white LED power indicator, a volume control with on/off switch, and a 3.5mm headphone jack on the front. On the back is the USB jack, a coaxial digital output, and the L/R line out RCA jacks. It claims to support 24 bit formats and up to 96 Khz sampling rates. The volume control has a nice “silky” feel and it at least seems like a high quality product.
WINDOWS PLUG-N-PLAY: On both XP and Windows 7 PC the uDAC installed smoothly without needing any drivers. It shows up in Win7 as an “SPDIF Interface NuForce uDAC 2”. Apparently NuForce is using a CODEC chip, not a pure D/A chip, as Windows also think the uDAC-2 can record but it lacks any inputs to do so.
HISS & NOISE: First I checked for hiss with my sensitive Ultimate Ears SuperFi 5 Pro’s. There’s some slight hiss with the volume all the way down, and it increases a bit as you turn it up, but even at maximum volume it’s not very objectionable. It’s similar to the Sansa Clip+ and is quiet enough for me.
UH OH! In trying to listen to the line level outputs I discovered they’re disconnected when headphones are plugged in. But, and this isn’t good, the switches built into the jack that shut off the line outputs, caused one line level output to stop working sometimes even with the headphones unplugged. I had to insert and remove the headphone plug several times to bring the dead channel back to life. And with headphones plugged in, the line outputs are left floating which isn’t ideal either. This might cause hum or noise in some computer audio setups through the speakers when headphones are plugged in. If you’ll never use the line outputs, however, this isn’t an issue.
SUBJECTIVE LISTENING: Before measuring the uDAC-2, I listened to some favorite music with a variety of headphones and the line outs. It has enough power to drive all the headphones I tried to levels I’m happy with. I went back and forth a few times between the uDAC-2 and my Benchmark DAC1 Pre’s headphone output and didn’t notice any huge differences. But the Benchmark did seem to sound better. This could easily be subjective bias, however, as being much more expensive it’s hard not to expect the Benchmark to sound better. I did notice some channel balance problems. Perhaps the measurements would be more conclusive? Please also see the results of the public listening test.
THE MEASUREMENTS: For the non-technical readers out there, the short version is I was really disappointed in the uDAC-2’s performance in several key areas. As mentioned at the start of this review, NuForce has confirmed my uDAC-2 is apparently OK and “within spec” for the first two problems. There are a total of four main problems and a few more minor ones I cover only in the tech section later:
PROBLEM 1: The uDAC-2 (like the original uDAC and uDAC2-hp) Has Serious Channel Balance Problems. Despite NuForce claiming to have upgraded to “a highly linear TOCOS volume control for improved channel tracking” the uDAC-2 still has what most would consider serious balance problems. Even set to nearly half volume (around 12 O’clock) it has a 1.4 dB error--anything greater than 1 dB is generally considered audible. At –30 dB it rises to a 3.3 dB error which is plainly audible. At –40 dB it’s a huge 10.3 db error! You can easily hear the stereo image shift to one side as you lower the volume control. This isn’t a subtle problem. Anyone can hear it and, the more efficient your headphones are, the more it will be an issue as you’ll use the lower end of the volume range more. And it’s also present in the line outputs as they’re controlled by the same volume control. It’s worth noting that channel tracking issues tend to be rather random. Every uDAC-2 will be at least a bit different.
The NuForce Response: Here’s NuForce’s reply to the channel balance problem after measuring their uDAC-2 sample (not mine):
“At 9 o'clock, there is hardly any signal, and analog pots do not behave well…
Volume at 9 o'clock (-41.6 db below reference): Balance Error: 8.8 dB
Volume at 10 o'clock (- 9.6 db below reference): Balance Error: 0.6 dB
Volume at 11 o'clock (reference level ~ 300 mV output): Balance Error: 1.0 dB
We consider this within spec. I will emphasize again that listening test over and over again from so many people convinced us to adopt this current solution. I understand that recording engineer might need a different product so I will post such details and plot online so that there is no mistaken about the intended customers.”
My View: I don’t agree that only “recording engineers” need decent channel balance—not if it’s a plainly audible problem. Someone with sensitive headphones studying in a quiet library, reading, etc. easily might end up with a big balance problem. And it’s also misleading to blame it all on “analog pots don’t behave well”. The circuit design has a lot to do with this issue, not just the potentiometer (volume control) used. I go into more detail in the technical section. NuForce says “so many people convinced us to adopt this current solution” but it seems more the opposite. If you look at all the uDAC reviews, the number one complaint is the channel balance problem. What “so many people” seem to want from NuForce is better channel balance!
A Useful Comparison: As proof this problem is easy to fix just look at the Behringer UCA202. It’s a $29 USB DAC and has a seriously cheap looking little thumbwheel for a volume control. The Behringer tracks within 1 dB over the entire range until –40 dB where it’s still within 1.5 dB. If Behringer can make a cheap volume control work well in their $29 USB DAC why is the $129 NuForce so much worse? My opinion, based on the NuForce response quoted above, is they think this level of error is perfectly OK. See the tech section below and my Behringer UCA202 Review for more details.
PROBLEM 2: The uDAC-2 Starts to “Clip” (badly distort) with a 0 dBFS signal(revised 3/3/11) - Using the line output, the NuForce has over FIFTY TIMES more distortion with a 0 dBFS input signal than a signal just 1 dB lower. With the headphone output it’s more than ten times higher. This matters because you only get a bit accurate full resolution stream with the PC’s volume control all the way up—see this Computer Audio Setup Guide for more details. And a lot of pop music is heavily compressed to make it sound louder with peaks frequently hitting 0 dBFS. Unfortunately, by NuForce’s own admission, the uDAC-2 “clips” 0 dBFS peaks. And it clips in this way at any setting of the volume control from the line or headphone outputs.
The NuForce Response: Here’s the NuForce reply after their own measurements on a different uDAC-2:
“I confirmed that we did this so we can have more dynamic and useful volume for practical use, such as listening to music. Our experience was that with MP3 or other format ripped music files, the output level could be very low, and there is not enough dynamic range. At the end of the day, customer buys the uDAC2 to listen with their ear, not their scope.
For testing, I said that by having -1.3dB, it will go below clipping. In practice, there is no music recorded at 0dB. You are taking measurement at MAXIMUM digital volume. We maxed out exactly near 0dB, so the THD is around 1% as intended (hitting the voltage rail limit so that we have the loudest and un-distorted pracitical music playback. If we don't have 1% THD at 0dB, that means we are not fully using our voltage limit and we used 'less range' for music.
And customers can just slide the volume on their PC if they are playing such a track. There is no perfect optimization for such as entry level product (DAC1 cost almost 10X the price of uDAC-2 and it is linear powered with 30V for dynamics). But I think uDAC-2 is a better engineered product given the limitation (of USB powered device and cost).”
My View: The explanation above is based on the idea NuForce accepted some clipping to get a larger dynamic range. But this argument doesn’t make sense to me. If you reduce the level by even 1 dB the uDAC-2 no longer clips. So we’re talking about giving up only a trivial 1 dB of “range”! The uDAC-2 has about 94 dB of dynamic range. If they had designed it not to clip, it would have roughly 93 dB of dynamic range. That’s about a 1% improvement. Is 1000% (ten times) more distortion on peaks worth a 1% improvement in dynamic range?
Real World Music: (added 3/3/11) NuForce said above “In practice, there is no music recorded at 0dB”. This is simply false! If you pop a best selling Lady Gaga CD into the drive on your computer and click play, with the PC’s volume set to max so you get the full 16 bit digital resolution (an accurate bit stream), below is what gets sent to the uDAC-2. All those blue lines reaching “1.0” is where the signal is hitting, or getting very close to, 0 dBFS and hence would be much more seriously distorted by the uDAC-2 (click for larger). If you’re skeptical they’re really mastering CD’s this way, just download the free Audacity software and look at some for yourself:
Does NuForce Understand This Stuff? (revised 3/9) It’s my understanding the comments above are from the Vice President of R&D at NuForce who also claims to be an engineer. The stuff I’m measuring is audio engineering 101. So it’s hard for me, a fellow engineer, to understand why a company would knowingly trade 1% more dynamic range for 1000% more distortion. I personally think the comments from NuForce are more of a contrived excuse than a real explanation. I suspect this is a design error, or just sloppy engineering, and was not intentional. Just as with the Stereophile Review of their CDP-8 ,they likely never bothered to properly measure the uDAC-2 as, by their own admission, they prefer to “design by listening”.
I’m not aware of any rational reason for a DAC to clip 0 dBFS signals. And it’s not something I’ve ever seen in any other product. And their solution, lowering the PC’s volume, no longer delivers a bit accurate stream to the DAC—i.e. your ripped CD becomes “processed” by the PC’s operating system and is delivered to the DAC with less than the full 16 bits of digital resolution. I show a graphical representation of this in the tech section. The main reason most people buy stand-alone USB DAC’s is to get better performance, not worse, than what’s built into their PC.
Another Comparison: The Behringer UCA202, has virtually no distortion playing a 0 dBFS signal (< 0.008% THD+N). This is nearly 100 times better than the uDAC-2. And the noise of the UCA202 is only 2 dB worse (-92 db vs –94 db). So, as with the channel balance, Behringer manages to do a much better job with a much cheaper product.
PROBLEM 3: High Output Impedance. I didn’t bring this up in my emails to NuForce, but apparently that “improved headphone amp” NuForce claims on their web site (see quote above) doesn’t include having a sufficiently low output impedance. The uDAC-2 measures about 6 ohms and this is well above the 2 ohms I consider acceptable for use with typical Balanced Armature headphones. Even the lowly $29 Sansa Clip+ and $20 FiiO E5 portable amp have much lower output impedance. I explain why this matters in my Amp Impedance article, but all you have to do is look at the frequency response below in blue:
In the graph, the popular Ultimate Ears SuperFi headphones cause over 4 dB of frequency response deviation with the uDAC-2. While the flat yellow line across the middle is essentially what you’d get if it had a much lower output impedance. Most would consider the blue response to be readily audible—especially as, in this case, it’s worst at midrange frequencies where the ear is most sensitive. With the uDAC-2 the frequency response will change in unpredictable ways--the above is just one example--with different headphones.
PROBLEM 4: Too Much Maximum Gain. I also didn’t bring this up to NuForce, but one can argue the uDAC-2 has too much maximum gain or should have adjustable gain. The measurements show with most any load, it clips above about 1 O’Clock on the volume control. Even the line outputs clip. This means the entire upper range of the volume is control is wasted when listening to most music sources. And given its high output, few headphone users will ever go above 11 or 12 O’Clock. This makes the channel balance problem even worse as it forces most people to use the lower part of the volume range. It may also make noise problems worse as I explain here. A partial solution is to reduce the volume in Windows, but, just like with the clipping issue, this means reducing 16 bit CD quality music to less than 16 bits of resolution. The only reason I know of to have a product designed this way is for the rare use where someone has seriously inefficient headphones and is also listening to an improperly ripped track recorded at a much lower than normal level. Why design a product to work for 1% of the users at the expense of the other 99%? See the technical section below for more details and graphical examples.
THE OTHER MEASUREMENTS: In most other ways the uDAC-2 measured similar to, or in some cases much worse than, the vastly cheaper Behringer. It did much worse, for example, with high frequency distortion and jitter. See the technical section for more details.
WHAT DOES THIS SAY ABOUT NuFORCE? This is a really good question. Before I tested the uDAC-2, and received NuForce’s responses quoted above, I respected the company. They have some happy customers, they’re responsive to problems, have a good refund policy etc. So, in some ways, they seem like a respectable business. They also seemed to take my concerns seriously and were very prompt in their response.
But it’s hard for me to accept a company shipping a $129 product with the sort of flaws the uDAC-2 has. It’s even harder for me to understand with a second generation product like the uDAC-2. NuForce already knew about the channel balance problem with their original uDAC and claims they fixed (or at least seriously improved) the problem. But did they? Given how much better the first generation (now many years old) Behringer USB DAC measures even using much older technology, and at a much lower price, it’s especially hard for me to forgive the uDAC-2’s poor measurements.
DID NuFORCE EVEN KNOW? (revised 3/9) After they received my measurements, NuForce didn’t seem to even know what the typical performance of the uDAC-2 should be. This was especially a surprise considering the VP of R&D, also an engineer, was apparently one of the two employees corresponding with me. Shouldn’t he already be familiar with the uDAC-2’s performance? It’s not like NuForce is a huge company or has lots of products. But, instead of knowing if my results were reasonable, or not, NuForce had to go borrow another company’s Audio Precision analyzer to measure their own product. And only then could they tell me if what I measured was normal. That just doesn’t make sense to me unless they really had no idea of what the real world performance was? This is also consistent with what happened with the Stereophile Test of the CDP-8 last year. NuForce admitted to being “surprised” as they apparently had not made similar measurements to Stereophile’s standard set of tests.
PRODUCT MANAGEMENT 101: Based on the above, I have to wonder, did anyone at NuForce properly measure the production version of the uDAC-2 before initially shipping it? Did they check to make sure they really had solved the channel balance problem? Did they verify the basic performance specs? Did they set up quality control standards to make sure the production units are within spec? Do they hold the factory to documented specifications? If the answer to any of these questions is “yes”, I would expect them to have already known what is “within spec” and answer my questions without needing to go out and take measurements. But, in this case, even the VP of R&D apparently didn’t know or have access to that sort of common information. Does that seem weird to anyone else? And if they didn’t know, where did the specs on their website come from?
MARKETING HYPE: (revised 3/9) Besides the claim of a new “highly linear” volume control with “improved tracking”, other NuForce specs influenced my decision to buy one. For example, up until a few days before publishing this review, NuForce listed on their website:
- Dynamic Range 110 dB
- THD+N = – 95 db
If you convert –95 dB to a percentage, it’s 0.0018 % THD+N. NuForce, since our email exchange, has removed this spec from their website and now just lists “0.05%” instead—which is a much more mediocre but realistic –66 dB. They also lowered the dynamic range from 110 dB to only 90 dB. These are huge changes in specs. The previous specs were impressive audiophile-grade numbers. The new specs are similar to the sound hardware built into many PCs. I have screenshots of the old web page for anyone interested in their changes. Don’t you think a company should base their published specs on some actual measurements? Apparently NuForce did not. Or perhaps they intentionally wanted the uDAC-2 specs to appear better than they really are?
WHAT ABOUT OTHER NuFORCE PRODUCTS? Have they used similar “tuning” and compromises in their other designs? Do their more expensive products have similar measured or performance issues or not meet their published specifications? Do they all have 6 ohm output impedances? I really don’t know but I do wonder.
BOUTIQUE AUDIOPHILE COMPANIES: To be fair, NuForce isn’t alone. There are plenty of small “boutique” audiophile companies that found a market niche selling poorly engineered, or what many would consider marginal performing, products. Nearly all of them, when someone challenges the performance, respond with excuses including some version of NuForce’s “At the end of the day, customer buys the uDAC2 to listen with their ear, not their scope”—i.e. that specs and performance don’t matter and it’s better to "design by ear”. And there’s often at least the implication, and sometimes the direct claim, that better measured performance would somehow sound worse. This is not true. Look at the Benchmark DAC1. It has among among the best measurements of any audio DAC anywhere. And it’s also been rave reviewed by some of the most critical, high-end, audiophiles writing for the uber-esoteric audiophile media. And if you don’t like that example, just look at how the Behringer USB DAC performs. It’s entirely possible to design gear that measures well, sounds great, is reasonably priced, and doesn’t require the manufacture to make any excuses. Behringer is just one example. There are many others.
WHAT ABOUT THE POSITIVE uDAC-2 REVIEWS (added 3/2/11)? The channel balance issue will literally be random from sample to sample of the uDAC-2. If the potentiometers NuForce is using are not individually tested, then you could end up with one that tracks fairly well or one that’s really poor. So some of the good reviews you read might be from those who got lucky. And if someone is already lowering the volume on the PC side they are less likely to use the lowest range of the uDAC-2’s volume control where the tracking is worst. The same is true if they have really power hungry low sensitivity headphones.
The 0 dBFS distortion issue only shows up on certain music and with the PC volume set for a bit accurate stream. It’s also possible it may vary from sample-to-sample due to component tolerances. The output impedance interaction is very headphone specific. With some headphones it may be minimal or tweak the frequency response in ways someone actually prefers. While with other headphones it may cause more severe variations of the sort that don’t usually sound good (like the big midrange boost and HF cut with my SuperFi 5’s). The excessive gain issue, again, can be partly overcome by turning down the volume on the PC side and depends on the headphones used. So it’s easy to see how some people wouldn’t care as much about the problems I’ve found. But, to me, that doesn’t excuse the glaring flaws when they’re so easily corrected.
THE PSYCHOLOGICAL ASPECT: How psychological and emotional aspects influence what we hear is well documented. For one good reference, see Dishonesty of Listening Tests. Humans, even if they’re not aware of it, are very influenced by expectations, the comments of others, the money they spent, the physical look and feel of a product (and here the uDAC-2 does well), the brand name, and other factors. For example, the Behringer A500 amplifier was widely criticized as being a “cheap clone” and sounding awful. It was indeed very cheap. But in an interesting blind test, serious audiophiles couldn’t tell it from a much more expensive audiophile amp. Virtually all of the uDAC-2 reviews and forum posts I’ve seen have been conventional “sighted” listening and hence are subject to this same well documented bias. You can draw your own conclusions but I’d encourage everyone to follow the two links if you haven’t already. They might help you save lots of money :)
BETTER ALTERNATIVES TO THE uDAC-2 (revised 5/2): As of May 2nd, I’ve now tested the FiiO E7 and can safely say it easily outperforms the uDAC-2 in nearly every area and represents a much better choice and value. I’d still like to get my hands on the Firestone Fireye2 as, based on this review and other information I’ve seen, it may offer better performance than the uDAC-2, at a lower price, with fewer problems. If you only need line outputs (no headphone jack), the similarly priced HRT Music Streamer II is worth a look.
BOTTOM LINE: To conclude the less technical part of this review, I believe the uDAC-2 has some inexcusable flaws. How much some of these problems are actually audible is open to debate. But, given nearly all the problems could have easily been fixed, I think it’s hard to defend the problems in a supposedly “audiophile” DAC selling for $129 from a company that specializes in DACs. NuForce argues their design is “better engineered”, but when a $29 Behringer DAC outperforms it in nearly every area, I have a hard time accepting that. Personally, I don’t see any reason to buy a uDAC-2 given its flaws and I’m going to return mine for a refund. But, ultimately, everyone has to decide what’s best for them.
TECH INTRO: With the drama out of the way, I’ll try to focus on the detailed results. A few are perfectly fine, a few are questionable, and some are the worst I’ve ever seen. For more details on how I measure, and the equipment I use, see my Testing Methods.
THD AT 0 dBFS: I run this same test on every digital device I test. It’s the standard “reference” used in most digital audio testing to compare other levels and measurements to. Here’s the uDAC-2 via the headphone output with a 15 ohm resistive load:
The high distortion here is from the analog side of the DAC stage clipping (hitting the supply “rails”) with a 0 dBFS signal. The spectral “signature” of clipping is having higher odd order harmonic distortion than even order harmonics. And if you look at the circled spectrum above, you’ll see exactly that. Odd order harmonics are widely considered to be far more unpleasant, and hence more audible. In this case they’re about –48 dB below the signal. Remember cassette tape hiss? That was generally much quieter--around 65 dB below the signal. So we’re talking about “unpleasant audio garbage” that’s a whopping 17 dB louder than typical tape hiss. Some might argue this level of distortion could be audible under some conditions. It’s important to note this happens at any setting of the NuForce’s volume control with any load. As explained earlier, I believe it’s a design flaw. Here’s the same test on the line outputs at 1 volt RMS into 100K:
If we drop the level by even a barely audible 1 dB, here’s what happens to the distortion of the Line Outputs:
Same thing on the headphone output (but the uDAC-2 headphone amp has about five times higher distortion in general):
The huge drop in distortion is because the DAC circuit is no longer running into the power supply rails (clipping). For reference, here the $29 Behringer reproducing the exact same 0 dBFS test signal that gives the NuForce fits on the exact same Line Out dScope test:
Note with the Behringer the more pleasant sounding 2nd harmonic is way down at –82 dB and the unpleasant 3rd harmonic is at –100 dB which just about everyone will tell you is inaudible. The rest of the harmonics are essentially lost in the noise floor. Compare this to the “forest” of distortion products of the NuForce at 0 dBFS above (The odd ultrasonic “bump” in the Behringer spectrum is above the audible range and explained in the UCA202 Review.)
NOISE LEVELS: Here’s the headphone noise under the same conditions as above but with a barely visible –115 dBFS signal to give the DAC something to do (many DACs mute and shut off with no signal giving an unrealistic noise measurement and making the specs look better—but no cheating allowed here):
Here’s the same measurement via the Line Outputs. This is good noise performance—especially for a USB powered device. It’s about 2 dB better then the Behringer UCA202. This agrees with my subjective review of the noise earlier":
MAXIMUM OUTPUT: Into a 15 ohm load, here’s the impressive maximum headphone output into 15 ohms just under 1 % THD+N:
Here’s the same thing into 150 ohms to better approximate higher impedance headphones:
This is very respectable output. Into 15 ohms it’s 47 mW and into 150 ohms it’s 32 mW. The uDAC-2 should drive most any headphone plenty loud.
GAIN AND VOLUME CONTROL: In the last two graphs, note the NuForce has reached max output at between 1 O’Clock and 2 O’Clock on the volume control. This means listening to typical material with any headphones, or the line outputs, the entire upper range of the volume control is wasted (assuming you send the full 16 bit accurate signal to the DAC from the PC). And 30 – 50 mW is more power than most can safely use with typical headphones. So, as a result, few users will likely get past half volume and the entire upper half of the volume control’s range will be unused unless they turn down the volume in the PC’s mixer. But this will reduce their CD quality audio stream to less than 16 bit resolution. Here’s the NuForce with a –10 dBFS signal (so it won’t clip at full volume) at exactly half volume (12 O’Clock) which is about as loud as most will ever use (~20 mW into 16 ohms at 0 dBFS):
Leaving everything the same, with the volume all the way up to max at 5 O’Clock:
So there’s at least 10.6 dB of “wasted gain” for most users. So to use the full range of the NuForce volume control, you have to turn the volume in Windows down by 10.6 dB. Here’s the PC’s output to the DAC at full volume with peaks hitting 0 dBFS:
And here’s what it looks like if you lower the PC’s volume slider by 10.6 dB to correct for the NuForce having too much gain:
Now only about 25% of the digital (numeric) dynamic range is being utilized. The other 75% of possible bit values are completely unused. Some may argue over how audible this is, but it’s clearly a compromise. If the uDAC-2 had less overall gain, or even better adjustable gain like Leckerton, Firestone, and others offer, it wouldn’t be as compromised.
HEADPHONE OUTPUT IMPEDANCE: Here’s the No Load output referenced to the same 400 mV at 15 ohms level used earlier. The voltage rises to 561 mV with no load. If you do the math, this is a 6 ohm output impedance:
WHY IMPEDANCE MATTERS: Here’s the same graph from earlier, the frequency response with a 15 ohm resistive load vs the Ultimate Ear SuperFi 5 Pro’s on the uDAC-2:
THD vs FREQUENCY: I could have run this test at 0 dBFS where the NuForce would have had a miserable result. But, to be fair, I ran it at –3 dBFS where the uDAC-2 has about 0.05% THD+N at 1 Khz. Here’s the sweep (in blue) plotted on the same graph as the Behringer UCA202 (shown in red):
You can see above 500 hz the uDAC-2’s distortion heads skyward. The Behringer does far better here. This implies poor design (or odd design trade-offs) in either the DAC implementation, low-pass (“reconstruction”) filters and/or the headphone amplifier.
IMD DISTORTION: Here’s the SMPTE plot at –2 dBFS which is just below the clipping value of the combined SMPTE IMD test signals (60 hz and 7 Khz in a 4:1 ratio):
Here the DAC is likely starting to clip at –2 dBFS, just as it does with a single tone near 0 dBFS. See the “mountain” around the base of the 7 Khz signal? Here’s a closeup:
All those spikes spaced multiples of 60 hz apart are IMD—that’s the 60 hz signal interacting with the 7 Khz signal in undesirable ways. Here’s the lowly Behringer with the exact same –2 dBFS signal and test. Check out the almost completely clean noise floor with only a few spikes that barely make it to –100 dB:
WHY THIS MATTERS: The clipping point of an amplifier isn’t determined by the level of any single frequency. It’s the combined level of all frequencies. With the NuForce, even though both tones individually are well under the clipping point, the combined result pushes the compromised DAC design into far more distortion than the Behringer even though it technically shouldn’t be clipping. The IMD test better approximates the complex interaction in real music.
FREQUENCY RESPONSE: With a purely resistive load of 15 ohms, the only interesting thing here is the channel balance problem (blue is left, yellow is right). Otherwise the uDAC-2 has sufficiently flat frequency response:
However, as previously shown, the uDAC-2 doesn’t do nearly as well with a real world load like the Ultimate Ear SuperFi 5 Pro’s headphones:
DAC Linearity: As mentioned in the Sansa Clip+ Review, this can be a good indicator of the quality of the DAC chip and here the NuForce does well rendering the –90 dBFS signal at –90.2 dB. At this low level, all the distortion is hidden in the noise.
JITTER SPECTRUM & PITCH ACCURACY: Another key difference between DACs, and especially USB DACs, is their jitter performance. There’s a lot of debate about this topic, but I can personally attest to having heard plainly audible jitter in a listening test. So don’t let anyone tell you it’s always inaudible. You can read more about it, and how it’s typically measured, in my Jitter Post. The USB interface and clock design can really make a significant difference. Here’s the Nuforce running my standard jitter test:
And, for comparison, here’s the Behringer:
There are some interesting differences here. The Behringer’s jitter side bands are at about –120 dB while the ones on the NuForce are some 15 dB higher at –105 dB. This is a substantial difference. Also notice the much wider “spread” at the base of the signal. This is generally an indicator of very low frequency jitter (which some argue is the more audible variety) and the uDAC-2 has quite a bit of spread which isn’t a good sign. In fact it may have the worst jitter performance I’ve ever seen since I started doing this test. I can’t say with any certainty if the uDAC-2’s jitter is readily audible, but I can say it’s a lot worse than even the much cheaper Behringer. At least the pitch is OK-but it almost always is.
SQUARE WAVE TEST: Not much exciting here, the NuForce does fine:
CHANNEL SEPARATION: Here the NuForce is average at about 47 dB and flat with frequency (which isn’t a problem):
Here’s the cheaper Behringer, once again performing much better, for comparison (although the 150 ohm load gives it an advantage):
CHANNEL BALANCE: If you scroll way back, you’ll see the channel balance error was 0.55 dB at the reference level of 400 mV. Turning down the volume down just 3.5 db from there (a typical moderately loud listening level), you get an audible 1.4 dB of balance error—not good:
What happens at –20 dB? It’s a still (barely) audible 1.2 dB:
Next is –30 dB which a person might use say studying in a library with sensitive headphones. Here it’s a very audible 3.3 dB error:
And, at –40 dB which is about as low as anyone is likely to listen at, the error is a huge 10.3 dB:
The above is the worst channel balance I’ve ever seen in any product by a wide margin. And it’s worth noting, at a similar control position (about 9 O’Clock), even NuForce measured a very poor 8.8 dB on their uDAC-2 sample and considered it “within spec”. This is an avoidable problem even with a really cheap volume pot. Behringer’s “thumbwheel” volume control probably costs them about $0.15 but it performs much better. This is likely due to more careful engineering and a better circuit design. At –40 dB the Behringer has only about 1.5 dB of error and it’s below 1 dB (inaudible) at all settings above that. If Behringer can do so much better in a much cheaper product, what’s NuForce’s excuse?
MEASUREMENT SUMMARY: I’ve hopefully written enough about the numbers. I personally think these measurements, and NuForce’s response, call into question the integrity of the product and their overall design philosophy. But you can form your own conclusions.
SUBJECTIVE vs OBJECTIVE CORRELATION: (added 3/3/11) When I originally, before doing any measurements, compared the uDAC-2 to the Benchmark DAC1 I thought I heard differences. And I believe those differences may correlate with one or more of the following measurements:
- The most likely thing I heard is the 4+ dB frequency response deviation with my SuperFi headphones. The midrange is broadly boosted and the treble is sharply cut in a narrow notch. If you’ve ever played with a notch or other steep slope equalizer/filter you know you get a very odd audible effect when you sharply cut a narrow range of frequencies. It can sound like the music is playing through a plastic tube. I think I heard a hint of that combined with a slightly glaring “in your face” midrange boost.
- I also heard some image shift--even if it’s only a few dB at the volume setting I was using. This is especially obvious when you’re switching back and forth relatively quickly. The SuperFi’s are very efficient so the volume control was set fairly low where the tracking is worse. This is no surprise given the very poor channel balance measurements.
- I also may have heard some increased distortion on peaks making loud passages sound more “gritty'” or “dirty”. This is because the uDAC-2 clips the peaks with distortion products as high as –48 dB which is easily audible by itself. The only question is if it’s fully masked by the music.
- I may have also heard some audible effects from the uDAC-2’s unusually high levels of jitter (among the worst I’ve ever seen).
- See also the results of the public listening test.
NOTES TO NuFORCE: (revised 3/9) In the interest of constructive criticism, and having a better product available, if I were designing a “uDAC-3”, I’d make the following changes:
- Fix the DAC circuit so it doesn’t clip at 0 dBFS. This shouldn’t be hard to do as every other DAC and digital player I’ve measured over many years can handle a 0 dBFS input signal without a problem. It’s probably a gain error or something similar?
- Use a better volume circuit so errors in the potentiometer are not made much worse by the respective ratios of the voltage divider, input impedances, etc. And sourcing a better pot wouldn’t hurt either.
- Lower the output impedance to around 1 ohm or less. This might be as simple as a resistor change. If the amp isn’t stable (or able to withstand a short circuit) without 6 ohms of series resistance, correct the feedback loops and/or use a better IC/design. You could even just copy the amp circuit in the $20 FiiO E5 which has a near zero output impedance.
- Reduce the overall gain by around 5 or 6 dB to provide a better match for most uses and headphones. This will help volume tracking and lower the noise floor in some circumstances. It will also make volume adjustments less “touchy”. Even better would be a choice of gain with a jumper or switch like some of your competitors have. Or perhaps offer two versions: A low gain and high gain model with different internal component values. Lowering the gain will do more to improve the noise performance than clipping 0 dBFS signals ever can.
- Find out what’s causing the excessive high frequency distortion and reduce it to more typical levels.
- Don’t route the line output through cheap switch contacts in the headphone jack that are apparently prone to problems (and/or contamination from the PCB wash process). If you still want to switch the line outputs, don’t leave them floating when headphones are plugged in and find a better way to do it.
- Establish a full set of performance specifications based on properly measuring the final prototypes and make sure the contract manufacture is required to perform testing and meet the same specifications. This will prevent being “surprised” when Stereophile, or people like me test your products.
- Publish the above specifications on your website with reasonable confidence the product the customers receive will actually meet them instead of specifications that seem to have no basis in reality.
IF YOU ALREADY OWN A uDAC-2: I’d suggest accepting the lesser of evils and dropping the volume on your PC a bit to help with some of the problems. You’re still stuck with the high headphone output impedance causing frequency response problems, high levels of jitter, and excessive high frequency distortion, but at least it won’t clip and you can use more of the uDAC-2’s volume control’s range to lessen the channel balance problem.