Header

Objective Reviews & Commentary - An Engineer's Perspective

August 10, 2011

Op Amps: Myths & Facts

op amp ad8397 5532INTRO: The op amp is a fundamental building block of audio gear. Op amps are widely used and, not surprisingly, they’re a popular topic among audiophiles. So what are the myths and facts?

MYTH: MOST OP AMPS SOUND DIFFERENT - There’s a general perception that op amps sound different. Many gear manufactures tout op amp brands and part numbers in their marketing literature. The $3 OPA2134 is supposed to sound much better than the $1 NE5532 and the $10 AD8610 is supposed to sound better still. But do they? The chip companies don’t help the perception implying some of their op amps offer better sound. But used properly, in a typical audio application, I’ll challenge anyone to a listening test and bet they won’t be able to tell the above three op amps apart. The only catch is it would be a blind test and the listener won’t know which op amp is which.

HISTORY: The Operational Amplifier (op amp) was invented in the 40’s. Bell Labs filed a patent in 1941 and many consider the first practical op amp to be the vacuum tube K2-W invented in 1952 by George Philbrick. Texas Instruments invented the integrated circuit in 1958 which paved the way for Bob Widlar at Fairchild inventing the uA702 solid state monolithic op amp in 1963. But it wasn’t until the uA741, released in 1968, that op amps became relatively inexpensive and started on the road to ubiquity. And they didn’t find their way into much consumer audio gear until the late 70’s and early 80’s.

MYTH: DISCRETE IS BETTER - For audio use the op amp’s main competition is a fully discrete amplifier made out of transistors, resistors, etc. Which is better? It turns out, for nearly all applications for which IC op amps are suitable, they easily outperform discrete designs in the following areas:

  • Better Performance – It’s very difficult to match the overall performance of even the  inexpensive 5532 op amp with a discrete circuit. The discrete circuit is at a disadvantage in many areas including component matching, bias stability, and the need to use off-the-shelf components (every “component” in an IC op-amp can be custom tailored and optimized to its task).
  • Simplicity – To even come close to the performance of an IC op amp many more components are required. You need differential pairs, multiple stages, current mirrors, constant current sources, bias circuits, protection circuits, etc. You end up with dozens or even hundreds of components to try and match a single dual op amp IC in a little 8 pin package.
  • CMRR/PSRR – Common Mode Rejection Ratio is how well an amplifier can reject unwanted noise.  Because their internal components are so well matched it’s easy for op amps to achieve excellent CMRR and PSRR (Power Supply Rejection Ratio) performance. This helps their real world performance in audio applications because they can reject noise on the power supply, and inputs, much better than most any discrete circuit.
  • High Open Loop Gain – Op amps typically have higher open loop gain. This allows more feedback which in turn lowers distortion. There’s another audiophile myth high feedback is somehow bad but that’s the topic of another article. Look up Bruno Putzeys recent article on the topic. He pretty much busts all the feedback myths wide open with real science. He even explains how so many people got off track. Trying to get comparable open loop gain in a discrete design typically creates challenging stability issues.
  • Repeatability – Op amps have tightly controlled specifications and detailed information available on their performance. They’re typically individually tested when they’re made so you know exactly what you’re getting. You can buy a TI 5532 today and an On Semi 5532 a year from now, and both will perform very similarly. High performance discrete circuits often require matched or hand picked components to achieve their best performance. This makes them difficult to reproduce and their real world performance is more of an unknown and sometimes requires detailed testing of each implementation (something few DIYers can do). Discrete circuits also typically cannot hold as tight of performance over a wide temperature range.
  • Massive R&D – The big semi conductor companies compete against each other for design wins. They spend serious money trying to out perform each other and have million dollar labs full of advanced equipment. They design the ICs right down to the properties of each transistor and have at their disposal types of internal components that are not even available as discrete parts. It’s impossible to match all their capabilities with a discrete design.
  • Built In Protection – Many op amps have at least current limiting and some also have other forms of protection like thermal shutdown. This makes them more robust than a discrete circuit unless similar circuitry is added to the discrete design making it even more costly and complex.
  • Ease Of Use – Op amps are very well characterized and typically well supported by the manufactures with detailed specs, performance graphs, and even application notes and sometimes reference designs. Following their guidelines usually results in predictable performance. They’re also typically easier to “glue” together when using a proper bipolar power supply as the outputs are referenced to zero volts and can be directly coupled to the next stage. The discrete designer is largely starting from scratch and has far more hurdles to clear.
  • Lower Power – An op amp, because of all the advanced techniques available to IC designers, can operate its output stage in Class-B with vanishingly low levels of distortion. While discrete designs are often forced to use much more power hungry Class-A to even come close. Overall, an op amp typically needs substantially less power than a typical discrete equivalent. This is a huge advantage for battery powered gear or if you need lots of amplifier stages.
  • Lower Cost – You can get amazing performance for under a $1 with an op amp. You can’t even come close with discrete designs. In fact, discrete designs often end up significantly compromised to limit their cost and complexity to reasonable levels. So you end up comparing a highly optimized IC against a compromised discrete circuit that still costs a lot more and performs worse.

DISCRETE ADVANTAGES: There are a few circumstances where discrete designs generally make sense. One is a needing output voltages greater than about 8 V RMS. There are some high voltage op amps but the selection is limited and they’re expensive. And the same is true if you need peak currents much over about 300 mA. There are some high current op amps, but they can suffer from thermal issues and are also relatively expensive. Driving speakers, for example, requires both voltages and currents that exceed these ranges. That’s why most high power amps for speakers are discrete designs. There are also some decent “chip amps” but they’re not strictly op amps. They also have thermal limitations and are limited to relatively modest power outputs. They also tend to have rather invasive protection circuitry due to their thermal limitations.

FIRST GRAIN OF TRUTH: There are often grains of truth behind many audiophile myths. But, very often, those grains no longer apply, or audiophiles apply them in ways that are entirely invalid. The early op amps, like the 741, were rather substandard for some audio applications. They had poor slew rates and they couldn’t manage much gain while maintaining bandwidth to 20 Khz--let alone do so with low distortion. They also couldn’t drive much of a load. That didn’t stop some manufactures from using early op amps in 70’s and 80’s audio gear.

wax phonographWAX CYLINDERS: Many audiophiles condemn op amps based on the early examples. By that same logic, their $10,000 vinyl turntables should also be condemned because the Edison Wax Cylinder phonograph was horrible. It could barely reproduce intelligible human speech let alone do a good job with music. Obviously, it’s not fair to judge a technology only by the early examples. But that’s how op amps mostly got their reputation of shame among audiophiles, and it stuck. Op amps didn’t offer serious audio performance at reasonable prices until the 5532 was released around 1985—almost 20 years after the 741. And it took several more years for the 5532 to show up in much consumer gear. (photo: photopedia.com)

SECOND GRAIN OF TRUTH: Some op amps have been judged in the wrong applications. The Cmoy headphone amp is just such an example. Typical op amps are made to drive loads of several thousand ohms or higher. Even the more powerful ones are often only rated down to 600 ohms. But that didn’t stop anyone from using these same op amps to drive headphones in the 16 – 300 ohm range. And, not surprisingly, some of them audibly complain at their mistreatment. They simply weren’t designed for such a low impedance load. If you compare overloaded op amp A against overloaded op amp B you may hear some differences in their highly distorted performance. But that’s only because you’re stepping on their tails and making them screech in pain.

YACA (yet another car analogy): Let’s take a Ferrari and a Lamborghini--both amazing cars with very similar performance on the road. But instead of driving them on pavement as their manufacture’s intended, let’s hitch them up to a plow and try to plow some corn fields. Suddenly the two very similar cars behave very differently. The Lamborghini has all wheel drive and likely does significantly better trying to pull a plow in dirt than the 2 wheel drive Ferrari. But so what? Who tries to plow a field with a sports car? This is analogous to misusing op amps and judging their performance based on an application they were never designed for.

op amp pinoutMYTH: MOST OP AMPS ARE COMPATIBLE - Op amps are complex devices. While many have the same pin configuration making it appear you can simply substitute one for another, that’s often not the case. They’re optimized for different input bias values, configurations, gains, feedback circuits, load impedances, quiescent currents, speeds/bandwidths, compensation values, operating voltages, noise trade-offs, etc. But some audiophiles can’t be bothered by, or don’t understand, all those details. So they simply swap them out without changing anything else in the circuit. Many have a favorite op amp or two they like to use in most everything without considering if it’s a good match. It’s like the old quote “when you have only a hammer everything looks like a nail”. But that just doesn’t work for op amps. And, not surprisingly, some op amps do better than others when you ignore their requirements.  Again, this is a lot like the Second Grain Of Truth above. If you use a given op amp incorrectly, it may well sound different. But it’s not the op amp’s fault. The person misusing it is creating the audible differences.

opamp upgradeMYTH: OP AMP UPGRADES ARE USUALLY WORTHWHILE - The forums are full of posts from people who buy some perfectly nice piece of audio gear, open it up, discover the manufacture used a “cheap 5532” op amp, and they promptly pop in something much more expensive and exotic to “improve the sound”. These are usually people who lack the test equipment to have any idea if their op amp swaps help, hurt, or are just a waste of money. They simply use their ears, in highly biased sighted listening, and draw all sorts of erroneous conclusions. I know of op amp swaps causing potentially harmful oscillation. Someone might hear ultrasonic artifacts from the oscillation as “newfound detail” when, in reality, they have unwittingly created a radio transmitter. And this isn’t as rare as you might think because of the “faster is better” mentality. (photo: photozou.jp)

YACA: Some of the car magazines have done objective tests of tire upgrades, shock upgrades, oversized wheels, etc. When they compare say a stock BMW to the same car with one or more of the “upgrades”, the factory set up usually posts the best overall performance, lap times, etc. While those massive 20 inch rims with ultra low profile tires might look cool they often perform worse than what the car came with. That’s because the manufactures, especially for performance-oriented cars, optimize the tires and suspension carefully. They understand all the trade offs better than anyone. And the same is often true with op amps and audio gear. In their attempt to “upgrade” the owner is often messing up a carefully engineered design and making it worse.

MYTH: GEAR COMES WITH LOUSY OP AMPS – Believe it or not, you can usually trust the bigger manufactures of audio gear to use an op amp that’s up to getting the job done. Why? Because for one thing such op amps are surprisingly inexpensive. So there’s little reason for a manufacture not to use an ideal op amp. For another they likely have $50,000+ worth of test instrumentation and can precisely measure differences between op amps. They also designed the circuit the op amp is in, so they better know what requirements matter most. But it seems a lot of audiophiles only see an inexpensive op amp on a circuit board that needs replacing. They may not understand the rest of the circuit, have any way to know if they’re creating new problems, etc.

op ampOP AMP STABILITY: Stability is critical for proper op amp operation. The gear makers tend to specify relatively stable op amps with sane bandwidths. So they don’t have to take many precautions in their circuits to keep such op amps happy. But when Joe Audiophile whips out his soldering iron and replaces the factory op amp with some high strung exotic part, what happens? The new part often requires much more attention to PC board layout, better power supply decoupling, different compensation, perhaps a more stable power supply, etc. In other words, without updating the rest of design (perhaps including even the PC board), the op amp may be at least somewhat unstable or otherwise unhappy. And because most instability is at ultrasonic or RF frequencies, the person doing the swap may not even know they just took a giant step backwards. Even RMAA can’t “see” ultrasonic and RF problems.

THE EASY LIFE: Most op amps in audio gear are not used in stressful ways. Many are simply buffers which means they don’t even have any voltage gain. And even in something like a headphone amp, the gains are relatively modest. And modern audio op amps are virtually never stressed for slew rate. This means the theoretical advantages of many expensive parts are completely useless in these “easy” audio applications. It’s like upgrading your lawnmower with a 400 horsepower V8 when the original 5 horsepower engine cuts the grass just fine.

THE WRONG LIFE: A lot of expensive op amps, including many favored by certain audiophiles, are optimized for entirely different applications—like precise DC performance, video use, etc. Using these for audio is like trying to plow a field with a Ferrari or drive a race car to work. The Ferrari’s main assets are useless in a corn field. For example, the OPA690 op amp used in the AMB Mini3 was never designed for audio use and has relatively horrible audio performance. If you stray too far from what the op amp is designed for it may well sound worse than a much cheaper part designed for audio use—like the 5532.

mixing consoleFOLLOW THE CHAIN: If you follow most music up the signal chain you’ll almost always find it’s already been through dozens, or sometimes even hundreds of op amps. Those big giant mixing consoles you see in the recording studios? Yup, most are filled with hundreds of op amps. Same story for the EQ, compressors, vocal processors, mic preamps, A/D converters, and much more. And do you think they’re all $10 Analog Devices or Burr Brown parts? Nope. They’re mostly inexpensive 5532s in the better gear, and even cheaper parts in the less expensive stuff. Sure music is increasingly produced in the digital domain, but I’ll bet before your music ever touches your headphone gear it’s already been through at least several inexpensive op amps. (photo credit Dennis AB) (photo: photopedia.com)

COMMON SENSE: If much of our most loved music has already been through dozens or hundreds of cheap op amps, is it reasonable to think that one more such op amp is going to make much difference? It’s not and that’s largely because op amps, properly used, are audibly transparent—i.e. you can’t tell they’re even there.

TRANSPARENCY: If op amps really have a “sound”, as many audiophiles suggest, it would follow when you add op amps to the signal path the sound should change. Two guys named Meyer and Moran conducted a very interesting rigorous study. They played high resolution SACDs on a high end system and sometimes inserted an extra A/D and and D/A into the signal path to “down convert” the high resolution audio to CD quality (16/44) audio. After 500+ trials lasting more than a year, using audiophiles, recording engineers, and students as listeners, they found nobody could tell when the extra A/D and D/A was in the signal path. On top of demonstrating the supposed benefits of SACD are highly questionable they also managed to demonstrate that A/D and D/A converters can be audibly transparent as well. And, as you may have guessed, both the A/D and D/A add several op amps to the signal path. But nobody could tell they were even there. There have been many more blind tests that also demonstrate different op amps (and much more) indeed sound so much alike even audiophiles can’t hear the difference. See the Matrix audio test for another example.

SIGHTED LISTENING: So early op amps sometimes sounded bad and if you misuse an op amp you can make them sound different. But what about all the decent op amps in proper designs? Why do so many claim they sound different? The answer is they use sighted listening. Our brain filters what we hear using other knowledge—like which op amp you’re listening to. It’s an involuntary response that even the most skilled listener cannot “turn off”. Check out this short BBC video on the McGurk Effect for an example of how our brain influences what we hear. And if you find that interesting, there’s a lot more in my Subjective vs Objective Debate article. Basically when someone “upgrades” an op amp they’re often expecting to hear a difference based on the manufacture’s claims, other (similarly biased) subjective listening tests they’ve read about, audiophile myths, etc. So their brain serves up the expected difference much like in the video linked above. But if they let someone else swap the op amps, and they don’t know which is which, the differences always seem to disappear unless there’s another problem. And such problems can be revealed with proper measurements. So if two different op amps both measure reasonably well in a given piece of gear, the evidence strongly suggests they will impossible to tell apart in blind listening tests.

MYTH: FASTER IS BETTER - An often cited reason for expensive op amps, or upgrades, is to get more speed. Sadly, it seems those making these claims don’t understand how “speed” applies to audio. As can be demonstrated with some simple math (and has been verified by Doug Self and many others), any op amp with a slew rate of 3 V/uS or greater is fast enough for nearly any audio application on the planet. And op amps like the 5532 can easily have bandwidths out to 200+ Khz in most applications which results in negligible phase shift or “delays” in the audio band. Using an op amp rated for 20 V/uS or even faster is just asking for other problems and it probably performs worse in other ways (i.e. more noise and distortion). This will be discussed more in the next article.

opamp cmoy rollingOP AMP ROLLING: Swapping op amps out, often using a socket so they’re easy to switch, is known as “op amp rolling”. Countless words have been written describing one op amp as having more “depth”, another as having a “blacker background”, etc. The Tangent headphone amp site has a typical list of subjective comments. As explained above, sometimes the differences might be real because some of the op amps are seriously unhappy in that particular configuration. But, more often, it’s just the usual sighted listening bias described above. If the swaps are done blind, and they’re using suitable op amps operated correctly, the alleged differences seem to always disappear.

DISCRETE OP AMPS: There’s another tiny grain of truth here. There are some very expensive discrete op amps that were designed specifically to outperform IC op amps in very specific ways. Their highly specialized benefits, to my knowledge, don’t offer any real world advantages in typical audio use. And audiophiles seem to favor much less rigorously designed discrete replacements for IC op amps. These discrete substitutes typically lack any sort of valid test data to demonstrate their real performance. So how does anyone know they outperform ICs? They mostly use Sighted Listening which is completely invalid (see above). Audio-GD discrete op amps were tested by Samuel Groner using an Audio Precision analyzer and the results were horrible. See The Discrete Is Better Myth.

OP AMP MEASUREMENTS: To try and keep my articles to a more digestible size, I’m splitting this into two parts. This article has covered the history, applications, and myths. The next article will cover the more technical aspects including op amp parameters (what matters on a datasheet for audio) and some actual measurements using my dScope audio analyzer and a number of op amps that were evaluated for the O2 Headphone Amp.

61 comments:

  1. Another great read, very much looking forward to the measurements!

    ReplyDelete
  2. Another good topic.
    This too was enlightening , thanks!

    ReplyDelete
  3. Nice concise explanation. I can't wait for the numbers.

    ReplyDelete
  4. Hey, thanks for another great article. As someone relatively new to the higher-end audio world I appreciate this.

    Being a scientist, I am always skeptical of the outrageous claims made (and prices charged) relating to a lot of audio gear.

    I have skimmed over some of your pages, and plan to read in more depth, but now that the O2 is coming out as a roughly $100 headphone amp, are you planning on anything similar relating to DACs? I currently use either my iPhone (pretty good sound quality, and enough volume for most of the time, but limited listening options) or my computer (more versatile, but has horrible audible noise from the rest of the system), and so I'd like to be able to get clean audio out of the computer, so any advice or recommendations would be great!

    ReplyDelete
  5. DAC recommendations on the way I think.

    Enlightening as always. :)

    -Satellite

    ReplyDelete
  6. I always though "opamp rolling" seemed like a stupid idea, as it works on the assumption you know more about the circuit and opamps than the circuit designer. Nice to see such an eloquent and comprehensive rebuttal of such ideas, from someone considerably more qualified to do so than me.

    ReplyDelete
  7. Another great article - and what I really like about this one, you need zero engineering knowledge or interest to read & understand it.

    -> I.e. this is an article for everybody who can read.

    Well done :)

    ReplyDelete
  8. Another great and entertaining read. I specially value what you have in one or two threads, referred to as the "activist" character of your blog.
    It is not only useful to be able to distinguish and filter the facts from all the BS out there in reviews, magazines and websites, but a truly sound -and sobering- practice to add solid proof to arguments and conclusions.

    Looking forward to reading the second part.

    Thanks

    ReplyDelete
  9. @Focal
    In other comment sections, NwAvGuy has recommended numerous DACS, off the top of my head the FiiO E7 (which he reviewed in an older post) and the Music Streamer II+ (the $125 one, i think its the II+?). He is going to be doing a review of the Asus Xonar DAC I believe, which is around $35 or so

    ReplyDelete
  10. The Music Streamer II is $150.

    Sounds good to me.

    -Satellite

    ReplyDelete
  11. If some people realized how much most opamp's schematics look like a traditional three stage discrete amp (as far as difference amp->VAS->output stage... with some current mirrors added in), then maybe they wouldn't be so hesitant to use/approve of them. Only things missing are some Miller caps for frequency compensation.

    ReplyDelete
  12. great article, as always.. but i'm still eagerly waiting for your measurements of asus xonar u3, haha..

    ReplyDelete
  13. Good piece. And by making your article a "more digestible size," I think your points become more clear and focused.

    ReplyDelete
  14. Good article. And I was planning to change TL082 in my NAD C352 amp to Burr Brown OPA2134. You saved me the labor.

    ReplyDelete
  15. There are still many areas of differences in performance exist among opamps such as driving capacitive loads like driving XLR cables with long lengths and driving low impedance loads down to 600ohms.

    Another very significant area of difference in performance of opamps is when you use them in active filters, some of them dont like over driven and tend to shift the phase excessively or phase reversal happens. TL072 is one such opamp, though this is not prevailent in NE5534/32 because it has back to back diodes at its inputs +/-.

    Yet another area of difference lies in headroom, discreet implementations offer much higher voltage swing and its not difficult to make it when you have matched pair of transistors available in form of differential pairs and current sources.

    my 2 cents,
    Kanwar

    ReplyDelete
  16. Kanwar, thanks for your comments. I plan to cover some of those issues in the next, more technical, article. I talk about phase reversal in the O2 design process article.

    I did credit discrete designs with higher voltage capability--although there are several op amps with around 60 volt maximum ratings that can work well if you just need some extra headroom. In practice, even in pro sound gear, common op amps typically have enough headroom as long the engineer is careful with the gain distribution. But it can make the design a bit more challenging.


    In

    ReplyDelete
  17. Regarding high PSRR,kindly note that most opamps have single ended differential input stages which gives different values of PSRR for negative and positive rails. In discrete implementation, one can opt for fully symmetric design in order to get balanced PSRR on both rails. NE5532 gives the PSRR curves w.r.t both negative and positive rails and you can see the difference yourself in the datasheet.

    There are certainly some benefits associated with opamps such as ease of use, small footprint and much less external components required and they perform good in most of the applications. NE5534/32 is versatile proven workhorse in many applications but it also has limitations of its own.

    Gain and repeatability is not an issue with today's advancement in available discrete building blocks which come in matched pairs. I have made discrete opamps and class-A opamps in lots of 100 in which the parameter variation was tightly controlled such +/-5% openloop gain variation. Temperature/bias was again not an issue.

    As i am from pro-audio field only, i have seen yamaha and studiomaster mixers using NJM2068 and JRC4558 in plenty but they fall flat when they have to drive more than 3 to 4 power amplifiers connected in parallel through XLRs. Replacing them with NE5532 instantly makes the difference felt in the sound itself.

    Kanwar

    ReplyDelete
  18. Kanwar, thanks for the added points. The PSRR I'll get into more in the technical article coming up. I'm aware of the dual transistors available but there's a relatively limited selection available (especially from mainstream distributors) that are highly linear, low noise, and well suited for high-end audio use.

    There's more than just openloop gain in terms of parameters. In my experience discrete designs made with unmatched parts tend to have rather variable CMRR/PSRR. And if they use a class B output stage they also have rather variable THD.

    I'm not saying you can't get impressive performance from a discrete design. But it's generally far more costly and far more difficult to do so. And for most audio applications where an op amp works well, I just can't see any real world advantages to discrete.

    That's especially true for DIY, where few can even properly measure the result. As long as the voltage/current requirements are not an issue, a good op amp will likely yield far better performance than an untested discrete design made with unmatched commonly available parts.

    You're correct about the drive capabilities of the 2068 and 4558 vs the 5532. That's all just proper engineering. If you don't know what the load is going to be it's good practice to make some fairly worst case assumptions.

    ReplyDelete
  19. The µA709 wasn't a vacuum tube OpAmp, and it wasn't invented in 1948 by Philbrick. It was a monolithic chip developed by Widlar as the second attempt after the µA702. The Wikipedia article on OpAmps gives the details.

    ReplyDelete
  20. So maybe I should switch back my MStage from the AD797 to the OPA2134 it came with :P

    ReplyDelete
  21. Thanks anon about the timeline error. I corrected the article. And the the next comment, I would not replace an OPA2134 with an AD797 unless you have a way to fully test the result. And even then, the OPA2134 is a fully transparent op amp when used correctly so there's nothing to be gained.

    ReplyDelete
  22. Quick op-amp question for you, NwAvGuy:

    Are you aware of any common op-amps which do not exhibit the ringing or bouncing-across-to-the-opposite-rail behaviours which you mentioned previously, and where the output voltage swings to within the same distance of both supply rails (i.e. Vop+ = Vcc - x and Vop- = Vee + x). I was auditioning parts a couple of months ago for an instrumentation project (not audio) and had a tough time finding something that would do all this without going to an expensive rail-to-rail part.

    ReplyDelete
  23. FakeMilkshake, symmetrical clipping isn't that important for most applications. The opposite rail issue is addressed by many parts. And the voltage swing is very load dependent. Without knowing more details, even the $0.39 NJM2068 meets your criteria.

    ReplyDelete
  24. Symmetrical clipping was important in this case. I will grab some extra 2068's when I put in my O2 order and audition them. Thanks.

    ReplyDelete
  25. Hey, thanks to the one or two people who responded in regards to my DAC question. I know NwAvGuy has said he likes the Fiio E7, particularly its amp, but I recall him saying the DAC section on it was okay but not stellar. I guess I was wondering on two things in particular:

    1. Does he plan to review other DAC or particularly DAC/Amp options in that rough price range (like the Audinst HUD-Mx1, for instance)?

    2. Does he plan to come up with a DIY DAC design, like the O2, which is based purely on the specs of the unit rather than listening test?

    I don't know if DACs are as susceptible to the design issues that he finds with some amps, or if a DIY would be worthwhile compared to what is out there already.

    Thanks again for the help.

    ReplyDelete
  26. Focal, I'm open to suggestions for reasonably priced DACs to review that are $150 or less but most everyone seems to want me to review something different and I'm looking for something with hopefully broad appeal. The only two with much consensus so far are the Xonar U3 and Creative X-Fi HD.

    DACs are in many ways more susceptible to DIY problems than amps. It's challenging to achieve the specs on the datasheet if you don't use the chip maker's reference design (and I've yet to see a DIY designer that has). Even commercial companies, like NuForce, get it wrong (see my uDAC-2 review).

    ReplyDelete
  27. thanks for the post, very informative I'm looking forward for the op amp measurements XD


    I agree not all the op amps are good for all applications I remember when I was on school op amps were a headache on certain of the lab experiments and I remember how some (sorry I don't remember the exact op amps) performed better at higher frequencies and some performed better at points reaching the reference voltages.

    ReplyDelete
  28. And Focal, I believe you have it backwards on the E7 review - the amp part is lacking, and the DAC part is great. The dac uses a well respected chip, however the amp is a little underpowered. This is intentional though, because the amp is portable and thus in FiiO's mind, only to be used with headphones that don't need as big of a boost. If you own an E7 and need to drive bigger headphones, you can buy the E9 desktop amp and plug in the E7 to act only as a DAC.

    ReplyDelete
  29. You're welcome Amanuense. And to anyone looking at the E9, make sure you can live with it's high output impedance. It's a poor match for headphones in the 16 - 50 ohm range including AKG K701s, Denons, Grados, planars (Audeze, HiFiMan), etc. But it works fine with 300 ohm HD600s for example.

    ReplyDelete
  30. If the E7 has a good DAC section, yet certain limitations when it comes to its amplifier driving headphones with higher impedances, then it would be great to specify -more or less- what headphone impedances it is capable to handle, and which it definitely isn't made for.
    That would be a great reference you could add to your extraordinary review of the E7. The topic is certainly brought up by your review, but you only mention experiencing the E7's limitations while listening to headphones with a 80 and 250 Ohms impedance.
    Of course the sensitivity of headphones also plays an important role, but it would still be a great reference to have more clear spectrum of the headphones -or impedances- the E7 is really capable of delivering with.
    I personally own a pair of HFI 2400 Ultrasones, which have an impedance of 70 Ohm and a sensitivity rated at 94dB, and I would like to know if the E7 would be a good option for them, and I guess many other readers have the same question regarding their own headphones.

    Sorry to be asking for more, when your posts are already genuinely terrific, but I got really enthusiastic about the E7 and I would really get one if it does the job with my cans.

    Thanks for your great -and so necessary- blog, and for taking the time to answer everyone's questions.

    ReplyDelete
  31. This discussion really belongs in either the E7 comments or perhaps the Headphone Amps/DACs comments. There is zero issue with the output impedance with the E7. It's below the 2 ohm rule so it's a non-issue with any headphone. What matters are the sensitivity and voltage requirements. They are discussed in the Headphone Amps/DACs article.

    ReplyDelete
  32. This should be an interesting read. Even with my tin ear I can hear the difference between a OPA2134 and OPA627 in the same circuit.

    ReplyDelete
  33. The OPA627 isn't part of my tests because I was evaluating only dual op amps in DIP8 packages. But, that said, I'm fairly confident in a typical configuration with the circuit optimized for both parts, and in a blind test you wouldn't hear a difference unless it was a really high gain circuit and the rather mediocre noise performance of the 2134 was audible.

    The key word above is BLIND. It's not easy to to do blind comparisons of op amps without a fairly specialized set up, or long times between shutting down the gear, having someone else swap out parts (or pretend to), and powering it back up again. But I've done them and will be doing more now that I have multiple identical O2 boards laying around :)

    The OPA627 typically sounds better because it's expensive and everyone else said it sounds better so that's the widespread expectation. It's no different than the Matrix Audio test, or Meyer & Moran, etc. Or, it's possible it's ringing or otherwise has some problem making it at least sound different.

    ReplyDelete
  34. Problems with op amps made with ic or discrete is the topology itself. An op amp use a ton of negative feedback to adjust it's gain. Buffers with a gain of 1 are worse. Huge amount of negative feedback is the best recipe to get mediocre or bad sound.

    ReplyDelete
  35. @anon, where's your proof high amounts of NFB equals bad sound? All you're doing is propagating a myth with nothing to back up your claims.

    Why doesn't the "bad sound" show up in blind comparisons? I would also strongly suggest reading Bruno Putzeys recent article on negative feedback. He very clearly explains why "high NFB is bad" is a myth.

    Finally, I'm hoping to compare the O2 to a fully discrete amp like the HeadAmp GS-1 (a Kevin Gilmore design) in blind testing. It should be interesting! :)

    ReplyDelete
  36. I just read this on a well-known DIY forum, thought I'd share it:

    "Please immediately throw the supplied NE5532 in the bin and install your favorite dual op-amp in its place. The supplied opamp sounds like junk, and will give you an unfair opinion of the sound."

    A perfect example of the McGurk Effect ?

    ReplyDelete
  37. Yeah, that's likely typical biased sighted listening--i.e. the McGurk Effect. As I've explained, the majority of music people listen to, including some treasured audiophile recordings, has already been through multiple 5532 op amps. They're the mainstay in high-end pro sound gear.

    My blind challenge stands. Please direct whoever made that claim to step up. If they really believe the 5532 sounds obviously bad they can get $500 donated to the charity of their choice if their ears can support their claims.

    ReplyDelete
  38. I've rolled almost every op-amp you can but/think of. I've always come back to the Texas NE5532AP.
    Taking everything into consideration, it sounds the best, has excellent drive and thankfully is inexpensive. I agree with NWAVGUY - do a GENUINELY blind test between 5532 and other op-amps, with 10 minutes hearing rest between each chip change - I think you'll be surprised at what comes out on top.

    ReplyDelete
  39. Ohhh God... I would like to throw away all my SUPER EXPENSIVE op-amps and stay with the original ones. Great reading....

    ReplyDelete
  40. Hi NwAvGuy!

    I'm currently using NE5532P, OPA2134PA and OPA627AU(x2) in my DAC, I have a few more on the way now such as AD8620AR(x2), LT1028CSW(x2), AD797AR(x2), OPA2111KP & OPA552AP.

    Which of these would be suitable for a blind test in the objective2 amplifier?

    Thanks in advance if you think some of these are suitable to alternate in the objective2 design, then I'll black out the lettering, write on them underneath with ultraviolet ink, and practice to see whether I can identify any difference, in pursuit of audio truth.

    Kindest Regards,

    Kiteki

    ReplyDelete
  41. In the O2 it's fair to compare the OPA2134, NE5532 and NJM2068 in blind testing. But to do that right, you need two otherwise identical O2 amps, fed with a "Y" cable from the same source, volume matched to exactly the same level, a way to switch the headphones between both of them, and someone else to do the switching behind your back in such a way you have no idea which is which. The brain's memory of subtle audio details is very short. So during the time required to swap op amps you will forget a lot. It's best to rapidly switch between "A" and "B".

    Some of the other op amps you listed may be stable in the O2 as well, but I haven't tested them. I would stick with the three above.

    If you can really make them all look the same, that could work too, but that might be harder than you think. The color of the packaging material (shades of dark gray), style and plating on the pins, etc. may still give you clues as to which is which. Better is to have a friend swap the op amps in a way you can't see which one is installed. And sometimes have them only pretend to switch the op amp but leave the same one in place. You also have to make sure the volume setting does not change even a little between op amp swaps.

    ReplyDelete
    Replies
    1. Thanks for this idea.

      Hmm, right so three O2's with different opamps 'behind a curtain', all connected to the same source, with someone quickly switching them say 30 times+ and writing down every switch, while the person on the other side has to identify the sound and write down 30 times+ which opamp it was?

      Seems like a fair test to me. I hope someone takes it up if not myself.

      Delete
    2. Generally it's best to just compare two of anything at a time. So you compare A and B. If you can't tell them apart, you can then replace one with "C" and repeat the test. If you can tell A & B apart, then you first compare A to C, and then in a different series, B to C. There's a lot of info on the web regarding ABX, DBT (Double Blind Testing), etc. There are some links in my Subjective vs Objective article to more resources.

      Delete
    3. If there are three objective2's, and it's only a matter of deciding which two share the same chip in a volume matched ABX, however not _identifying_ which chip that is, that sounds too easy!

      I'm quite perplexed now, because I've even read comments on diyaudio from someone working at National saying which chip he thought sounded the best in their listening room (LME49713).

      I'll try assembling two objective2's with different opamps then.
      At the moment I'm perplexed because the OPA2134 and OPA627 honestly sound different in my DAC, perhaps it's a power issue or one of them is fake...

      Delete
    4. ABX/DBT testing is all about determining if a person (or ideally several people) can reliably tell A from B. They have to score significantly better than just random guessing over several trials.

      If they can reliably hear a difference, ABX/DBT doesn't help much with deciding which one sounds better--that's a subjective choice. With op amps, when properly implemented (power supply, compensation, drive capability, closed loop bandwidth, stability, sufficiently low noise and distortion, etc.), the issue usually isn't deciding which one sounds better, or which is which, it's that they don't have any "sound" at all.

      As explained in this article, lots of people who "roll" op amps really have little idea what they're doing. In some cases it's a lot like dropping a lawn mower engine in your car, or a big V8 engine into your lawn mower--neither is a good match. So it's no wonder you read all sorts of things on diyAudio, Head-Fi, etc.

      But it's fair to compare the 3 op amps I mentioned above without changing the O2's gain stage. There are others that work OK as well--see the next article on this blog regarding Op Amp measurements.

      Finally, if the differences are so obvious, why hasn't one single person come forward to accept my blind op amp listening challenge?

      Delete
    5. "Finally, if the differences are so obvious, why hasn't one single person come forward to accept my blind op amp listening challenge?"

      That's a very good question.

      Delete
  42. Hi kiteki

    If you find that changing the interconnect cable changes the sound of your DAC then it is could indicate a deficiency in the output buffer.

    I had an OPA2134 in the I-V section and an OPA2134 in post filter/buffer section of my Denon DCD825. Luckily Denon did not use 1 chip for the left channel and 1 chip for the right channel. I did think that the 'dark' sound of the OPA2134 was a myth, as the sound I was hearing was pretty punchy and forward. However I think that the OPA2134 has problems driving cable capacitance and the top end loses detail and the bass become boomy with even interconnect capacitance.

    With the OPA2134 it was quite easy to distinguish between a coax Radio Shack cable and a twisted pair + screen Ixos cable. The cheaper Radio Shack cable was felt to sound more neutral on this CD player.

    Swapping to an NJM4556 in the buffer position the sound is clearer, especially the top end. For example it is easier to identify the various cymbals in cluttered rock music. Switching the interconnects the difference is no longer so obvious. I think that this chip is just about impervious to any reasonable cable capacitance - NwAvGuy can prove me wrong - as it drove 500pF through 10r perfectly to my ears.

    I'm guessing that the difference in sound that you hear is dominated by you choice of buffer amp.

    ReplyDelete
  43. Just found this whole blog...and have been enamored reading it. As someone who has spent many years of her life recording/mixing on SSL 4k and 9k consoles full of 5532s and 5534s, it really is nice to see someone finally standing up and saying "hey the 5532 is a great opamp that most of your music has already been through". Was on the edge about building an O2... might just do that now that I see you're a smart man :-)

    ReplyDelete
  44. In my last comment I suggested that the differences that kiteki heard were due to the ability of the OPA2134 to drive cable. Now I think that the cable sound was a secondary effect and I am not even sure that the NMJ4556 made much difference.
    I put a NE5532 in the buffer position and tried other opamps in the I-V position. Here is the conclusion of sighted/blind testing using Vivaldi 4 seasons CD. (I knew which chips were which but my friend did not)
    OPA2134 - nice euphonic sound but violins sounded synthetic and a little bright. Harpsichord was lost.
    LME49720 - Not so bright sound but now we could hear the correct timbre of the violins and the harpsichord suddenly appeared.
    NE5532 - Clearly more accurate than the OPA2134 but violin tone was not quite as convincing as the LME49720. On the plus side it was also easier to pick out the rhythm of the bass than with the LME49720.

    The LME49720 gets my vote for the seductive top end and the mid detail. It just sounds right.

    Of course this adds nothing to the argument as to whether the same chips sound different in the simpler role of the gain stage of the 02...

    ReplyDelete
    Replies
    1. Thanks for the added comments. If you're talking about listening, killing the power, prying a chip out, replacing it, powering back up, and listening again you're at a big disadvantage. It's been shown we humans don't do very well at remembering subtle differences in things we hear. It's way too easy for our memories to slightly alter or skew what we thought we hear.

      It's a lot like if I show someone a brown house color chip from a paint store, take it away for several minutes, and then present them with half a dozen very similar shades of brown and ask them to pick out which one I just showed them 5 minutes ago. They're apt to get it wrong if the shades are relatively close.

      But when you put two color chips side by side it's easy to determine if they're the same or different. The same is true in blind listening. You need to switch nearly instantly (ideally in under 0.1 seconds) to perform a direct comparison that's not tainted by "selective memory". This has been demonstrated with multiple studies (some of which are referenced in my Subjective vs Objective article).

      I believe in a proper blind ABX test with immediate switching between two otherwise identical pieces of gear you would not hear any differences between those three op amps unless one or more of them is being operated in a way that's stressing the part. In fact, I'll give $500 to the charity of your choice if you really can hear a difference in a proper blind public listening test (see the Op Amp Measurements article).

      Delete
  45. Unfortunately, logic and rationality are extremely rare commodities.
    And more so in the audiophile domain, which is being run by big money in niche domain, peer pressure, egos and placebos.

    *Sigh* but the people reading these posts would already know about this, and people who need to read and understand this, will never try to ...

    ReplyDelete
  46. I'm interested - as a musician and an EE student - to hear what you think about comparative testing by inverting the polarity of one amp and then summing its output with the other so as to attempt to get them to cancel. This kind of thing gets done often when people are evaluating for instance DAW plugins. I can imagine there's some challenges in getting this to work, but couldn't it help to support your arguments in some ways?

    ReplyDelete
    Replies
    1. This is discussed in the What We Hear article and also Music vs Sine Waves. Audio differencing is a very useful tool.

      Delete
  47. I'm curious most people usually 'upgrade' their Op-amps to some exotic amp never intended for headphones, which sounds completely stupid for obvious reasons, but it seems to me that it's unclear that perhaps say picking up an Op-amp like the NJM2068 to replace a stock NE5532 on say a soundcard which you do yourself mention in O2 design as being fairly comparible but noiser, I could understand in a time-pressured development that they might have made scarfices for reasons of supply and testing time which makes sense since someone mentioned that the NJM2068 is rather rarely mentioned (perhaps being Japanese and not very good at over seas marketing) would in fact degrade performance to get the product to ship on time. plus poorly implemented soundcard drivers leads me to think this (my onboard has far better features and stablity)

    I'm still skeptical of this though, as the design is still unknown and side-effects of using a NJM2068 can't be accurately predicted without it.

    ReplyDelete
    Replies
    1. Some devices, especially ones made in large volumes, may be "cost engineered" in ways that might compromise the choice of op amps. But it's not very likely in the sort of gear audiophiles are usually trying to "upgrade". The NJM2068 is cheaper than the NE5532 and performs better in some applications.

      The amount of time spent on design is a very different issue. Generally for large volume products, some extra development costs won't make a huge difference in the profits. But there often is a rush to get products to market for competitive reasons. With low volume audiophile products, all bets are off. Some are even largely "designed by ear" and/or designed by people who don't even understand proper engineering.

      Delete
    2. Thanks for the reply,

      My Soundcard is a low-end card being a Xonar DS, it comes with a Op-Amp Socket, and a NE5532, it fits the bill of a cost engineered entry level audiophile product. and 5.1 is pretty much broken as Windows seemly never advertises the correct configuration to any apps (Linux on the other hand makes this card far better).

      I'll probably just pick up an O2 amp kit to clean up the impedance problems I assume it has and open the possibility to some less efficient headphones, when I buy a decent pair, instead of fiddling with Op-Amps and the like.

      But either way, I like learning about stuff, so keep up with the awesome articles.

      Delete
    3. The NJM5532 is most likely used as a buffer on the front channel output of the card, and its advantage over the NJM2068 is that it can handle low impedance loads better. It probably also does not contribute much to the overall noise floor of the card, as most of it probably comes from the WM8776.

      Delete
  48. The weapon of choice to drive h/p seems to be JRC4556 (usually SMD type). I’ve found them in Creative Soundblaster Live series & quite a few motherboards (operating at +/- 5V), also in old Marantz & Philips CD players. The JRC4558 (SMD type) can be found in almost every line out driver section of consumer DVD players (+/- 12V). JRC4580 or NE/ NJM5532 (again SMD type) are also used in some soundcards to drive h/p. But never have I seen 4556 in parallel to drive h/p (like O2). Looks like these opamps from JRCs has got a huge fan followings amongst designers & engineers(incl. NWAVGUY & yours truly) but often getting slammed by subjectivists in forums.

    ReplyDelete
    Replies
    1. Thanks for the comments. The cool thing is the O2 has been praised by a lot of subjectivists, and those who have bothered to do blind testing, have found it sounds every bit as good as even much more expensive headphone amps. Those slamming the 4556 have even less credibility now.

      Delete

PLEASE NOTE! If you see "LOAD MORE..." OR "LOADING..." just above this paragraph, you need to click that link to load the newest comments. It may take several seconds. When all the comments are displayed the Load More/Loading link disappears.

COMMENT APPROVAL: Comments may take anywhere from a few minutes to a few days to be approved. You might want to save a copy (highlight Ctrl-C) of your comments before submitting in case something goes wrong. The approval process is necessary to prevent obscenity, etc. Please keep comments on topic. Comments promoting commercial interests will not be approved and may be reported as SPAM.