TR Forums

sluggo wrote:

Headphone impedance - if you want the most efficient power transfer, you want your amp's output impedance to match the impedance of the load. But with headphones, efficient power transfer is not really all that important. What's important is that your output is not the limiting factor in your voltage swing. In other words, as long as the amp's output impedance is less than the headphone impedance, you've done about all you need to do. A 600 ohm headphone is a very easy load to drive and your choice of op-amps is less critical. A 32 ohm headphone is more difficult (electrically) and will probably show more differences with different op-amps. A motherboard bragging about having "a high-capacity amplifier which is able to drive 600Ω loads" is not saying much, really.

Yes, a higher impedance load is easier on the amplifier as, at any given gain, it draws less current. BUT a higher impedance load also requires more gain (voltage) to achieve the same volume (db) level as the amount of power being pushed by the amp is less. The problem is that most line outs on sound cards/sources are not designed to have enough gain to provide adequate volume for impedances much above 50 Ohms (but this also depends on driver sensitivity). So, basically a 600 Ohm solid state amplifier chip's main goal is to increase gain (voltage). I imagine these circuits must also have more power/thermal handling capability as they must also be designed to handle a wider range of impedance loads...

So in reality, these 600 Ohm designs DO have to be designed to provide more power than the regular line outs designed for lower impedance loads.

cynan wrote:

sluggo wrote:

Headphone impedance - if you want the most efficient power transfer, you want your amp's output impedance to match the impedance of the load. But with headphones, efficient power transfer is not really all that important. What's important is that your output is not the limiting factor in your voltage swing. In other words, as long as the amp's output impedance is less than the headphone impedance, you've done about all you need to do. A 600 ohm headphone is a very easy load to drive and your choice of op-amps is less critical. A 32 ohm headphone is more difficult (electrically) and will probably show more differences with different op-amps. A motherboard bragging about having "a high-capacity amplifier which is able to drive 600Ω loads" is not saying much, really.

Yes, a higher impedance load is easier on the amplifier as, at any given gain, it draws less current. BUT a higher impedance load also requires more gain (voltage) to achieve the same volume (db) level as the amount of power being pushed by the amp is less.

Uh, no. To say that increased volume always requires more voltage or power ignores a few things. Volume level as a function of input voltage depends on many variables, most of which get summed up into a stat called sensitivity.

The problem is that most line outs on sound cards/sources are not designed to have enough gain to provide adequate volume for impedances much above 50 Ohms (but this also depends on driver sensitivity).

A 3dB difference in sensitivity is not at all uncommon in the world of headphones, and it makes all the difference in the world in terms of perceived volume. To hand-wave it away as though it's secondary to the discussion makes for an unconvincing argument.

And are we talking about line outs? I'm pretty sure the OP was talking about "headphone" outputs (basically lower impedance line outs). Headphone outputs swing between +5v and ground with oodles of current available. This is plenty for any set of reasonably designed headphones.

So, basically a 600 Ohm solid state amplifier chip's main goal is to increase gain (voltage). I imagine these circuits must also have more power/thermal handling capability as they must also be designed to handle a wider range of impedance loads...

Every amplifier's main goal is to provide gain, reduce output impedance, or both. I don't know what you mean by "600 Ohm solid state amplifier chip", but if you mean an op-amp in a circuit likely to see a 600 ohm load, then no, they are not designed with higher power dissipation in mind.

So in reality, these 600 Ohm designs DO have to be designed to provide more power than the regular line outs designed for lower impedance loads.

I don't see any reasonable way to make that claim. A 600 ohm load is going to draw far less current than a 32 ohm load. If the output voltage is the same and the current demand drops by a factor of nearly 20, then so does the power output.

I was taking the 600 Ohm example from your "a high-capacity amplifier which is able to drive 600Ω loads" statement.

I already acknowledged that driver sensitivity is a variable. The fact remains that if you design an output/amplification stage to accommodate a large impedance range you need to also accommodate a larger range in gain.

So you are correct that the 600 ohm load is not the issue. But rather, it's driving a lower impedance load at the higher gain, that was implemented for the higher impedance load that, on average (variations in driver mechanics, etc, that result in varying sensitivity aside), that requires more power handling. In other words, you need to build in some degree of protection for people who crank the volume with a lower impedance load.

cynan wrote:

I was taking the 600 Ohm example from your "a high-capacity amplifier which is able to drive 600Ω loads" statement.

I already acknowledged that driver sensitivity is a variable. The fact remains that if you design an output/amplification stage to accommodate a large impedance range you need to also accommodate a larger range in gain.

So you are correct that the 600 ohm load is not the issue. But rather, it's driving a lower impedance load at the higher gain, that was implemented for the higher impedance load that, on average (variations in driver mechanics, etc, that result in varying sensitivity aside), that requires more power handling. In other words, you need to build in some degree of protection for people who crank the volume with a lower impedance load.

This could be accomplished passively by inserting a moderate amount of series resistance in the output circuit. It will have little effect when driving the higher impedance load since most of the output voltage will still appear across the load, but will significantly drop the amount of power delivered when the load impedance is small.

The one thing I'm not sure of is whether damping factor (ratio of output impedance to transducer impedance) matters for headphones; additional series resistance degrades the damping factor. For speakers you are supposed to have a much lower output impedance than the transducer impedance, to allow the amp to control unwanted motion (overshoot/ringing) caused by the inertia and mechanical resonances of the transducer itself. I suspect this doesn't matter nearly as much for headphones because of the small transducer mass, but I'm not a headphone expert.

I don't know a lot about PC sound. I just know that the analog inputs, probably from even the best motherboards, won't match a discrete sound card. I know some people bash Creative, but I've have really enjoyed my Creative X-Fi. I went for the external, USB option so I can swap it between machines easily. It delivers solid sound at a pretty reasonable price. The included software's faux 5.1 sounds amazing and always warns me when a pilot is dancing around on top of my titan in Titanfall.

Probably not what you want, but I wanted to toss this out in case a lower - mid range user wanted to see some more options.

southrncomfortjm wrote:

I don't know a lot about PC sound. I just know that the analog inputs, probably from even the best motherboards, won't match a discrete sound card.

Yeah, this is the area in which onboard still tends to be really weak. If you plan to do any recording from line or mic level inputs, and care about the sound quality, a discrete card is still an absolute must. For most other things, a reasonably well-implemented onboard solution is generally sufficient.

Yeah I have the little jewel I described near the top running my sound while my onboard audio handles my mumble client. Works well.

With the Muse DAC being so tiny I have plans to use it as an audio front end, apparently my Sony Experia Z Ultra will use it and then all I need is an amp and 12v for the DAC.

just brew it! wrote:

cynan wrote:

I was taking the 600 Ohm example from your "a high-capacity amplifier which is able to drive 600Ω loads" statement.

I already acknowledged that driver sensitivity is a variable. The fact remains that if you design an output/amplification stage to accommodate a large impedance range you need to also accommodate a larger range in gain.

So you are correct that the 600 ohm load is not the issue. But rather, it's driving a lower impedance load at the higher gain, that was implemented for the higher impedance load that, on average (variations in driver mechanics, etc, that result in varying sensitivity aside), that requires more power handling. In other words, you need to build in some degree of protection for people who crank the volume with a lower impedance load.

This could be accomplished passively by inserting a moderate amount of series resistance in the output circuit. It will have little effect when driving the higher impedance load since most of the output voltage will still appear across the load, but will significantly drop the amount of power delivered when the load impedance is small.

Hmm. In that case, variation in sensitivity aside due to other factors, statements like "a high-capacity amplifier which is able to drive 600Ω loads" really has nothing to do with current supplied. Rather, its all about gain - so that, on average, enough will be supplied so that higher impedance headphones will be comparatively as loud as lower impedance on "regular" line outs.

cynan wrote:

It may be harder to distinguish between OP-AMPs in some implementations vs others. And some implementations are better than others. I have sound card that uses four separate OP AMP chips for right and left stereo channels - one for each polarity. I could definitely tell the difference when changing them. Now, of course it wasn't a major difference, and likely not one that would matter to 99% of people

Again, you knew the difference, and that colored your perception.

It's always been the same. Any good system will audibly change if you change anything in the signal path. The problem has always been, does it sound better or worse?

PenGun wrote:

It's always been the same. Any good system will audibly change if you change anything in the signal path. The problem has always been, does it sound better or worse?

I think people should be thinking "does it sound better or worse to me"

chµck wrote:

PenGun wrote:

It's always been the same. Any good system will audibly change if you change anything in the signal path. The problem has always been, does it sound better or worse?

I think people should be thinking "does it sound better or worse to me"

It has also been demonstrated that people will convince themselves that they hear a difference even when there is none, if they *believe* there is supposed to be a difference. Even in a blind test, tiny (sub-decibel) differences in signal level can color perceptions. This is why any data that was not obtained via an ABX double blind test with carefully matched signal levels is automatically suspect.

just brew it! wrote:

chµck wrote:

PenGun wrote:

It's always been the same. Any good system will audibly change if you change anything in the signal path. The problem has always been, does it sound better or worse?

I think people should be thinking "does it sound better or worse to me"

It has also been demonstrated that people will convince themselves that they hear a difference even when there is none, if they *believe* there is supposed to be a difference. Even in a blind test, tiny (sub-decibel) differences in signal level can color perceptions. This is why any data that was not obtained via an ABX double blind test with carefully matched signal levels is automatically suspect.

Human hearing is a subtle and powerful sense. If you can't hear a change, in signal path changes I think your system is not revealing enough.

That tiny signal levels make such a difference should be a clue.

It is in phase analysis that human hearing can discern such subtle things. You are equipped to hear an object flying through the air and touching a leaf or branch, here or there, and move your head out of that path. Pretty large areas of the brain are applied to processing audio signals by humans. Because of this humans are very sensitive to distortion of any kind and easily detect the common problems audio reproduction presents.

just brew it! wrote:

chµck wrote:

PenGun wrote:

It's always been the same. Any good system will audibly change if you change anything in the signal path. The problem has always been, does it sound better or worse?

I think people should be thinking "does it sound better or worse to me"

It has also been demonstrated that people will convince themselves that they hear a difference even when there is none, if they *believe* there is supposed to be a difference. Even in a blind test, tiny (sub-decibel) differences in signal level can color perceptions. This is why any data that was not obtained via an ABX double blind test with carefully matched signal levels is automatically suspect.

A small nitpick, but I think you are referring to a single blind test, not a double blind test. This terminology gets misused all the time when taken out of medical related research, in my limited experience. A double-blind test would mean the both the listener and the individual switching between the equipment do not know which is which.

@PenGun - I think you're at least partially missing the point. It has been shown that people will hear differences when the signal path is EXACTLY THE SAME, provided they are convinced that something has been changed.

@cynan - Read the link I posted. They do in fact suggest using a true double-blind setup.

Edit: And yes, I am familiar with the difference between the two. Although I do not do audio (or medical!) testing, I run homebrew competitions. The beers are judged double blind, i.e. neither the person serving them nor the judge tasting them knows who brewed them. My unpacking/sorting team and I spent several hours this past weekend replacing the entrants' bottle labels with anonymized judging number stickers, on over 1,200 bottles of beer!

Amazon has the Sound blaster Z with the beam forming microphone and the EMI shield for $86 and free shipping along with 241 customer reviews with a 4.5 star rating.
Well worth the price. Just about as cheap as the OEM version with no EMI sheild or mic that works great.
http://www.amazon.com/dp/B009ISU33E/?tag=pcpapi-20

Also the more expensive ZX version that comes with the volume control knob and headphone jacks ETC. actually reduces headphone output quality. Besides there is no need for it since you can switch from speaker to headphones in the SBlaster CTRL panel so you do not have to reach around the back of the computer to plug and unplug your headphones. Yes the plain SB Z has a headphone amp along with 600ohm support.

just brew it! wrote:

@PenGun - I think you're at least partially missing the point. It has been shown that people will hear differences when the signal path is EXACTLY THE SAME, provided they are convinced that something has been changed.

@cynan - Read the link I posted. They do in fact suggest using a true double-blind setup.

Yeah. You're right. I missed the asterix at first. It seems that they are recommending something analogous to true double blind. But often people use double=single blind when discussing these sorts of comparisons.

But then, I fail to see the benefit of double blind testing for this purpose. Single blind should be sufficient. With something like the testing of pharmaceuticals, the administration of the actual treatment can be a lot more involving than just flipping a switch, allowing for much more opportunity for administrator/tester bias. And the large distinction is that in a medical trial, in general, the same individual is not getting the intervention and the control. In audio, double blind testing seems like overkill. The good thing about having the testing automated is not so much to make it double blind, but so that each sample gets played identically, for the same length of time, etc.

If you read further, the explanation also recommends having the listener leave the room between switching equipment. This, IMO, is counterproductive. The whole point is to compare subtle differences, known not to stick in memory very well. If there is too much time between A/B/X than of course you're not going to be able to pick out subtle differences. In this way, these tests are almost designed to fail (accept the null hypothesis). What you want is to switch as quickly as possible. And then, of course, do the set, A/B/X multiple times, in random order, until you have enough statistical power. Whether or not you have one or multiple listeners depends on the objective: whether you are interested in determining which piece of equipment listener A prefers or if they can tell the difference (which I think is the most interesting insofar as this discussion) or whether people, on average, can.

I think I know the origin of the "leave the room" recommendation. I remember reading an article about a test where it turned out the listener was subconsciously picking up on subtle differences in the sound of the switches/relays which were being used to select between the A and B sources. I agree if you can completely eliminate all such cues from the environment then leaving the room should not be necessary (and is probably counter-productive).

Let me be clear: IF there is a difference within the range of human hearing, THEN it would be audible. The problem is that self-professed audiophiles get hung up on the notion that they can "hear" differences between op-amps that produce technically-indistinguishable results, when blind testing has shown that they cannot actually hear such a difference. Because our sense of hearing is subjective, the notion that an improvement has been acquired yields a dopamine rush and makes the experience more pleasurable. Problem is, dopamine rushes are transient, and so you have to go spend another chunk of money on another alleged improvement to get it again. It is essentially the psychological component of addiction, and at the extreme end of such silliness it can be just as preoccupying and expensive as an actual drug habit -- e.g. you have people paying hundreds or thousands of dollars for specialty power cords and the like.

Novices should not be encouraged to buy into such things in a fruitless chase for improvements that don't technically exist, when their money could be invested in things that do have wide variability, such as a good DAC (although even these are increasingly becoming indistinguishably good) or a nice set of speakers/headphones (which vary widely in design and quality, and make an enormous difference in both measurable and perceptible audio output). If someone wants an option for swappable op-amps simply for the joy of tinkering, more power to them; some people like tinkering, myself included.

ludi wrote:

If someone wants an option for swappable op-amps simply for the joy of tinkering, more power to them; some people like tinkering, myself included.

Indeed. Just don't deceive yourself into believing that just because you buy more expensive op amps, the sound will be better. Or that a card with swappable op amps automatically sounds better than one without.

I love tinkering too. If you don't believe me, go look up my recent thread about the DIY repair on the RK-9000 keyboard I fried a few weeks ago!

I like tinkering when I have the time, but am not the most handy with electronics as I don't have too much practice. I fried a Creative X-Fi board when I tried to replace all the capacitors, for the heck of it

I agree largely with ludi's sentiment. Of course it's not in one's best interest to waste their money on higher end parts/equipment when the differences may subtle, esoteric, or even non-existent, particularly if budget is a factor.

Also, I'm inclined to agree that most SMD op-amps that are intended for analog audio probably produce audible results that are very similar audibly, and many may be indistinguishable to most people. On the other hand, there are OP-amps that have been designed with the purpose of imparting different tonal flavors... Not all OP-amps are created equal.

And one thing to reiterate about blind testing from the above discussion: Not all blind testing is equal. This can make all the difference.

If an op amp is imparting a "different tonal flavor" while being operated within spec (i.e. not being driven beyond its output voltage swing limits or faster than its rated slew rate), it is pretty much by definition a defective op amp.

just brew it! wrote:

If an op amp is imparting a "different tonal flavor" while being operated within spec (i.e. not being driven beyond its output voltage swing limits or faster than its rated slew rate), it is pretty much by definition a defective op amp.

How can you call something a defect that's imparted by design?

Here'as an example of a couple of op-amps designed to serve this purpose.

just brew it! wrote:

If an op amp is imparting a "different tonal flavor" while being operated within spec (i.e. not being driven beyond its output voltage swing limits or faster than its rated slew rate), it is pretty much by definition a defective op amp.

*Disclaimer: I have not read every post in this thread.*

This is correct. Op amps are supposed to be "perfect" voltage amplifiers (at least if you intend to use one as an amplifier). So anything you feed into the op amp should come out with virtually zero distortion.

superjawes wrote:

just brew it! wrote:

If an op amp is imparting a "different tonal flavor" while being operated within spec (i.e. not being driven beyond its output voltage swing limits or faster than its rated slew rate), it is pretty much by definition a defective op amp.

*Disclaimer: I have not read every post in this thread.*

This is correct. Op amps are supposed to be "perfect" voltage amplifiers (at least if you intent to use them as an amplifier). So anything you feed into the op amp should come out with virtually zero distortion.

Yeah, But now we're confusing the definition of what makes a perfect op amp from an engineering principles perspective with the utility of different op amps that impart slight variation in distortion or gain at different frequency ranges, etc, that may be interesting for tinkering with when messing about with analog music reproduction.

The most extreme example is with speakers/headphones. Not all are engineered to be perfectly flat across their frequency range. While some would argue that this detracts from their performance (which is probably the case if using them for sound engineering), others appreciate emphasis imparted in mid frequency range, etc...

A piece of sound reproduction equipment is not necessarily designed with the same goals as, say, a test/measurement device (where accuracy always trumps everything else).

All op amp circuits heavily utilize negative feedback. The input-output transfer function has a rigorous mathematical definition based on the values of the resistors and other passives connected to the op amp. While I'll agree there can be differences in S/N performance, any other deviations (assuming the circuit is being operated within spec) are an indication of poor design, or poor component selection.

I'm also finding it rather difficult to take the page you linked seriously, for at least a couple of reasons:

1. The "extend lead" kit is going to be terrible from a noise pickup standpoint. 5" (give or take) unshielded analog audio connections? Riiight.

2. Any description of audio equipment which uses the phrase "soft and slightly thick, and full of emotions", while also claiming a 0.0008% THD figure, is, quite frankly, blowing smoke up your posterior.

just brew it! wrote:

1. The "extend lead" kit is going to be terrible from a noise pickup standpoint. 5" (give or take) unshielded analog audio connections? Riiight.

I'm sure one could easily trim the leads as needed...

And audio-gd is a respected brand among various audio circles. Their product descriptions might be exaggerated, which is typical of many asian companies, but once you get past that, many people enjoy their products.

just brew it! wrote:

All op amp circuits heavily utilize negative feedback. The input-output transfer function has a rigorous mathematical definition based on the values of the resistors and other passives connected to the op amp. While I'll agree there can be differences in S/N performance, any other deviations (assuming the circuit is being operated within spec) are an indication of poor design, or poor component selection.

I'm also finding it rather difficult to take the page you linked seriously, for at least a couple of reasons:

1. The "extend lead" kit is going to be terrible from a noise pickup standpoint. 5" (give or take) unshielded analog audio connections? Riiight.

2. Any description of audio equipment which uses the phrase "soft and slightly thick, and full of emotions", while also claiming a 0.0008% THD figure, is, quite frankly, blowing smoke up your posterior.

For the first point, these aren't designed to be used for computer audio. But primarily to be used in devices where power supplies and analog stages are already separated/shielded from each other. And that extension is optional, only for enclosures that don't have enough clearance in the vicinity of the socket.

I agree with you on the second point. This is definitely a marketing job. This is aimed to help them sell to a certain (gulp) target audience. Also, the English is a factor in the poor selection of buzzwords...

However, I find it hard to believe that he's offering these 2 or 3 op amp varieties purely to scam customers. If that was the game, why not have 10? If the designer really believed that there was no difference in tone, why bother offering more than 1? And they're only about $25 each - and are significantly more expensive to make relative to the SMD op amps Gigabyte sells in their overpriced upgrade kit.

It's not like op amps are his only business, he has a whole line of DACs and amps that run the price range gauntlet. The one thing I do respect about this company, is that, in most cases, you can literally see the component cost that goes into the more expensive equipment, whether it be redundant power supplies and filtering, to multiple DAC chips per channel. That said, I obviously can't vouch that all (or most) of the more expensive products are worth the money (i.e, that all of the redundancy really makes a difference in the sound), but it's hard to say it's all blowing smoke... (like you can easily with some other companies).

But I'm getting off topic. I only wanted to show an example of op amps that are actually designed to perform differently. But I can't argue that I have the experties to validate these difference purely based on the designs.

just brew it! wrote:

All op amp circuits heavily utilize negative feedback. The input-output transfer function has a rigorous mathematical definition based on the values of the resistors and other passives connected to the op amp. While I'll agree there can be differences in S/N performance, any other deviations (assuming the circuit is being operated within spec) are an indication of poor design, or poor component selection.

And to quickly respond to your first point, yes. This is the function of the actual op amp. But these modules have additional circuitry added (one of the most substantial being actual coupling capacitors) that might be thought of as extemporaneous to the op amp at their core (eg, the gain stage, by which I believe he means the power supply to the actual op amp circuit that allows it to perform its voltage balancing duties, and a sort of pre output stage). I believe it's here where the differences that (may or may not) impact tonal flavor derive.

cynan wrote:

Yeah, But now we're confusing the definition of what makes a perfect op amp from an engineering principles perspective with the utility of different op amps that impart slight variation in distortion or gain at different frequency ranges, etc, that may be interesting for tinkering with when messing about with analog music reproduction.

The most extreme example is with speakers/headphones. Not all are engineered to be perfectly flat across their frequency range. While some would argue that this detracts from their performance (which is probably the case if using them for sound engineering), others appreciate emphasis imparted in mid frequency range, etc...

A piece of sound reproduction equipment is not necessarily designed with the same goals as, say, a test/measurement device (where accuracy always trumps everything else).

We aren't confusing definitions (I put "perfect" in quotations for a reason). As JBI pointed out, op amps use negative feedback, and as long as you stay within the voltage rails and below the maximum slew rate, the internal components are basically irrelevant to the voltage response of the output. This is what I mean when I say they are designed to be "perfect." If the frequency response is anything but flat (in the audio range), then the design engineer has failed.

And the "op amps" we're talking about are packages. The components in the negative feedback loop are on the board and will not change if you swap the op amp.