Audio from the EE standpoint

[Crazy*Carl]

I have been a regular at head-fi for a while now, and have gone through a few of these outboard DAC/AMPs, but have never noticed any improvements in terms of sound quality, even with high end Sennheisers.

Seeing the EEV Audiphoolery blog got me thinking.  What exactly goes into good sounding audio from a EE standpoint?

Just some of the nonsense I have read on head-fi:

-solid state equipment vastly improves with burn in

-optical is better than usb

-high impedance headphones sound bad without expensive dedicated amplification

-onboard sound is bad

-a whole range of silly audio terms

[alm]

I have been a regular at head-fi for a while now, and have gone through a few of these outboard DAC/AMPs, but have never noticed any improvements in terms of sound quality, even with high end Sennheisers.

In my opinion, there are real differences between speakers. They are mechanical, and resonance and acoustics are real. Of course improvements are usually overstated, this is basic human psychology: whatever you have now is infinitely better than what you had before, especially if you just spent a lot of money. Plus the law of diminishing returns. The difference between $2 and $20 headphones is probably much greater (except if that $20 pair was just a re-branded version of the $2 pair) than between $50 and $500 headphones.

-solid state equipment vastly improves with burn in

Unlikely IMO, just like burning in cables. I've always considered burn in mainly as the listener getting used to the sound.

-optical is better than usb

If ground loops are a problem, or the equipment is not on the same potential, optical might be a solution. But optical for consumer audio is a solution in search of a problem, in my opinion. A simple differential cable (eg. USB) works just as well. And heck, for the low-bandwidth digital signals, something like an RS-232 cable would probably work just as well. One issue is that optical is a standard audio stream (S/PDIF), while USB is a much higher level protocol, so the DAC on the other end needs to be smarter.

-high impedance headphones sound bad without expensive dedicated amplification

You need more voltage swing for high-impedance drivers, I can see how an amplifier with a 3V LiPo power supply might not have the necessary output voltage swing. Line outputs (sound card) are usually designed for high-impedance (10kohm+), low-voltage (.7VrmS?) output, so headphones are a little outside their specs. Plus some consumer electronics might have amplifiers that were designed to be really cheap, giving them a high-impedance load and lowering the gain might improve things.

-onboard sound is bad

I'm waiting for someone to post a link to that motherboard with a tube in the amplifier, it must be good if there's a tube in it, right? There are exceptions to almost every rule (including this one ), but if you consider audio as low-noise design (the signal levels are fairly low to begin with, and the dynamic range of the human ear is fairly large, 120dB or so?), you might recognize that a box with a lot of high-speed signals and a power supply modulated by a CPU that draws a lot of power and rapidly changes between idle and full-power is not the ideal place, especially if you try to do it cheaply without any special shielding. A dedicated PCIe card or USB device is easier to isolate, plus usually more design effort and parts were spend on that $50 (?) sound card than on the audio part of a $75 motherboard. I have had onboard sound were the disk activity modulated my sound, my laptop does something like that. Front-panel outputs with an unshielded cable running through the case are especially likely to pick up noise.

-a whole range of silly audio terms

Likely to be crap. If you can't explain something in simple terms, you don't really understand it.

[Crazy*Carl]

Sorry, what i typed was a bit confusing.  What I meant was that AMP/DACs did not seem to change sound, while headphones or speakers do have an enormous impact on sound quality.

With the Sennheiser comment, what i meant was that even with 300ohm Sennheisers, I didn't see the benefit of DAC/AMPs.

Thanks for your reply though, its nice to hear different perspectives.  There is truth in the audio world, its just coated in so much garbage you have to sift through.

[mkissin]

Audio enthusiasts are very similar to car enthusiasts. Why bother owning a car that can do 250 km/h when the speed limit is 100? Why bother upgrading your audio equipment from someting with 0.05% THD to something with 0.0001% THD when you probably cannot hear the difference?

Because you can.

[Crazy*Carl]

Ya, its all of a question of where does the line get drawn between clear sounding audio and placebo.

I am not an audiophile nutcase yet, but the difference between a 100 dollar pair of headphones and apple ibuds is massive.   But how much better can it really get?  I made and posted this simple graph on head-fi.



So where does the 100 dollar mark go on that graph?  I would guess at least 1/2 way, if not 3/4 down the x axis.

[A-sic Enginerd]

-optical is better than usb

If ground loops are a problem, or the equipment is not on the same potential, optical might be a solution. But optical for consumer audio is a solution in search of a problem, in my opinion. A simple differential cable (eg. USB) works just as well. And heck, for the low-bandwidth digital signals, something like an RS-232 cable would probably work just as well. One issue is that optical is a standard audio stream (S/PDIF), while USB is a much higher level protocol, so the DAC on the other end needs to be smarter.

mmmmyeahoksure.
Actually, this one could have merit. NOTE I'm saying could, not definitely does. Depends on some factors. It's all digital with either of these so ground loops and such......no. It's all about bits on the AD/DA, compression (if any), and link bandwidth. USB has a crapload of overhead on it and even though the wire speed might be 480Mbps (in HS mode), real through put is a fraction of that. If the bus isn't heavily loaded, and the devices have sufficient processing power on either end to handle the overhead, and all the other factors are equal (compression, AD/DA, etc.) then there could or would be no real difference. However, with what I know about USB....I would actually just go with optical. USB will use a type of transfer for audio streams that will allow frames to be dropped. I don't know if optical protocol does same, but even if it does, USB is such a hideous link that it wouldn't be hard to swamp things where frames will be getting dropped. That means you will have loss of data in the sound reproduced.

[EEVblog]

Sorry, what i typed was a bit confusing.  What I meant was that AMP/DACs did not seem to change sound, while headphones or speakers do have an enormous impact on sound quality.

Of course.
Speakers are doomed to almost always be the greatest cause of acoustic distortion in any given system. Not forgetting the room acoustics of course.
It's trivial to get <0.01% distortion or whatever over bandwidth in the electronics for only a few dollars. Try matching that with speakers over bandwidth, good luck.
That's why you'll NEVER hear the difference between good amps, cables, or whatever in any well designed and controlled double blind A-B test with the same speakers.
Anyone who thinks otherwise is clearly delusional.
The moral of the story is spend your money on good speakers or headphones.

Dave.

[alm]

Actually, this one could have merit. NOTE I'm saying could, not definitely does. Depends on some factors. It's all digital with either of these so ground loops and such......no. It's all about bits on the AD/DA, compression (if any), and link bandwidth. USB has a crapload of overhead on it and even though the wire speed might be 480Mbps (in HS mode), real through put is a fraction of that. If the bus isn't heavily loaded, and the devices have sufficient processing power on either end to handle the overhead, and all the other factors are equal (compression, AD/DA, etc.) then there could or would be no real difference. However, with what I know about USB....I would actually just go with optical. USB will use a type of transfer for audio streams that will allow frames to be dropped. I don't know if optical protocol does same, but even if it does, USB is such a hideous link that it wouldn't be hard to swamp things where frames will be getting dropped. That means you will have loss of data in the sound reproduced.

Yes, if the USB bus is saturated, you might hear it, for example if you had an external HD on the same hub. My comment about optical being a stupid choice for consumer electronics was mainly because S/PDIF over coaxial cable seems perfectly adequate to me, no need for fancy optical links. Grounding can be as much an issue with digital signals as with analog, that's why most modern buses like USB are differential. But that's only if you're pushing the bandwidth limit, not with 44k samples per second or so.

It's trivial to get <0.01% distortion or whatever over bandwidth in the electronics for only a few dollars. Try matching that with speakers over bandwidth, good luck.
That's why you'll NEVER hear the difference between good amps, cables, or whatever in any well designed and controlled double blind A-B test with the same speakers.

Absolutely, but the important keyword is 'good' (or at least decent), I'm sure there are plenty of cases were the designer was asked to make an amplifier for a few pennies, or just nobody cared. Probably the same reason why the majority of the consumer audio equipment (and probably even the majority of the audiophool stuff) uses single-ended signals instead of differential; it's not like the cables would be more expensive than the current monster crap.

[jahonen]

I think there is a reserved part in each USB frame for audio stream transport (so-called isochronous data), so that mass storage device transfers should not affect the audio.

Regards,
Janne

[chscholz]

Carl,

I am rarely involved in measurements of audio signals. Most recently this is mostly related to decoding of I2S signals which can be done on modern oscilloscopes. Much of the work I have been involved is was related to specialized audio filtering ("A-weighted filters").
 
24 bit samplers appear to be standard in audio analyzers, one good audio analyzer is spec'ed to have an amplitude level measurement accuracy of ± 0.05 dB @ 1 kHz. One approach to get a feel for what is real and what is "snake oil" might be to look at the published specs of audio analyzers. Hard to believe that even the most sensitive ears can hear issues that a high-end audio analyzer can not measure.

I have seen large differences in signal integrity of USB2 cables. Cheap cables that do not carry the USB2 logo (but are often sold as USB2 cables nevertheless), tend to fail high speed signal quality tests. I have no idea if this makes any appreciable difference in audio quality.

[Simon]

I have a cheap ebay amp that outputs a few watts (they never stated RMS but 180+180 W  ) I was running a pair of nice looking little 10 W speakers off it for my PC, then I thought, well I'm storing my 100 W speakers up there on the shelf in ready to use position, what the hell: now the sound quality is distinctly better and I'm still using that cheap ebay amp ! I do have an external car amp thats stuffed full of parts and to be honest I can hardly tell the difference but for the fact that it is more powerful

[xani]

Thing about digital audio signals is that if it "loses packets" somewhere you WILL bloody hear it, its not analog where when u get crap cable or w/e u get more noise + same 50hz from power network, in digital signal you would get all kinds of clicks and weird noises ;]. Wost thing that could happen to digital signal is bad resampling esp when it goes 44100->48000->44100 thru some fast (crap accurate) algorithm, and even then just getting better spearker would give bigger difference

[alm]

I believe that there's no separate spec for USB 2 hi-speed cables, so in theory any USB 1.x cable should be fine for USB 2. I have seen some crappy cables that failed with USB 2 hi-speed devices, but they probably didn't meet the USB 1.x spec either (although it did work because it wasn't as critical). And indeed, problems with digital communication are very different, so claiming that a different S/PDIF cable will give you better frequency response or lower noise are absolute garbage. I didn't use an USB cable in that statement, since USB contains power, so a crappy cable could cause too much IR drop or pick up noise (but any competently designed cable should be fine).

[Simon]

I haven't forgotten a rather informative and entertaining article I read once about speaker cables and how all of the mambo jambo stuff about special cable was just that

[alm]

I find it amusing that audio people often worry about high-frequency behavior, since their idea of high-frequency is 30kHz or so. For most electronics purposes, this is just barely above DC, and you have to try really hard to notice skin effects or inductance/capacitance at those frequencies.

[A-sic Enginerd]

A few responses / my $0.02 on USB specifics:
You don't need to hit "saturation"  point of the USB bandwidth to see loss. The bus bandwidth is broken up into timesliced chunks. Which brings up the next thing: yes, audio streams will use isoch transfers. Yes, the OHVI or UVI or whatever the hell the acronym is for the host side driver (sorry, it's been 5+ yrs since I worked on USB) is designed to ensure a certain bandwidth for isoch traffic. However, as I said, amongst other variables - there needs to be adequate processing power behind all that list processing. At that point it really doesn't have anything to do with the physical bus bandwidth capabilities. That's why I was mentioning before about using something like S/PDIF (be it optical, coax, whatever) could be a better option over USB. But if the host side isn't trying to talk to too many things at once or there's say a dedicated processor just for doing the host USB traffic, then yea it's a pretty good bet you'd get equal performance. However, there is still the issue that isoch is not a lossless channel. Oh, it's also worth mentioning that even though isoch is guaranteed its chunk of timeslice, not all of that isoch slice necessarily gets to be allocated to any one device.  

Yes, I would be sure to use strictly cables with the USB logo embossed on them. The standards body will not issue a company the right to use the logo if the cabling doesn't meet spec. And yeah, I wouldn't be surprised that if you used a cheap cable that doesn't meet spec you're going to experience a lot of loss which means things sound like a CD that's been run across the gravel parking lot. Also, even with all things being good (and this is where those with much more analog experience than I can help out) if there's sufficient noise I could imagine: is it possible to have a noisy enough environment that even with properly (spec) shielded cables and such that enough noise would be induced to cause issues? If there is, because isoch traffic is used, the data is gone, history, no longer recoverable. I don't know how noisy that environment would have to be and how likely that is with a typical home AV setup. And it's not a matter of maybe just a bit here or there getting flipped and ok we live with that. If a bit gets flipped in the packet header, for instance, that prevents the device from decoding the information on what it's supposed to do with it, the entire packet gets dumped.

So you might be able to make that cheap cable work for something like your mouse or keyboard - because they are low speed devices. Or it might even work for something like your printer - because it uses bulk transfers which is guaranteed delivery (i.e.: there's a confirmation handshake). But I would avoid using it for anything that will use isoch traffic.

So like I said initially - there's a lot of variables at play that can make USB behave just fine, or it could be crap. Could easily go either way.

BTW: give pause when using the phrase "USB bus". It's the same thing as saying "ATM machine".  

[jahonen]

Here are some my random thoughts about experiences with audio stuff:

I hate when audio marketing people explain something using engineering terms, without having any facts supporting their statement. I often tend to say that something may sound different without having any difference on the signal at all. Why not leave it just that way. I mean something like this. Anybody with a calculator can easily calculate the impedance of the RCA connector; it is near 30-40 ohms, not the mentioned 200 ohms! That makes RCA a poor choice at video or coaxial SPDIF interfaces.

Cable parasitic properties matter because way the audio interfaces are made, relatively high output impedance causes quite easily roll-off near 20 kHz. That brings me another question, why not make output impedance 50 ohms and terminate the inputs too to 50 ohms. That way one could get several GHz of bandwidth practically for free (or not free, one would have to use decent connectors, RCA's are a mechanically a joke compared to SMA or even BNC). Thus it would make the parasitics irrelevant, instead they become an essential part of the system.

I'm not sure that SPDIF is any better than USB. The fact that DAC master clock has to be recovered from possibly bad signal for both USB and SPDIF may in fact lead to measurable differences in the DAC clock phase noise (or jitter). You can think that as a frequency modulation of your sampled data. Clock timing is always essentially analog variable and it suffers easily. It is easy to get zeros and ones correctly but getting them exactly at right time is not so easy. The timing issue is completely irrelevant for computer storage media, as long as zeros and ones come through correctly. It would be technically much better if the receiver (DAC) would clock the source. That way the critical master clock would not have to be recovered via PLL from incoming data stream. I once measured SPDIF output from my DVD-player and it looked quite horrible. Bounces and rings. That measurement is not surprise when I looked inside, SPDIF is routed through large single-sided PCB. It made a signal integrity engineer in me to cry.

For some years ago, I designed and built DAC-headphone amplifier combo (TI TPA6120A2 as an output amplifier/buffer), using Jung super regulators in the power supply. I don't think that super regulators really made the difference, but hey, why not. At least they have measurable much less output noise and lower output impedance than common jellybeans, like 78/79xx or 317/337. As I write this text, I'm listening this thing. Sounds still nice.

My another audio experience was the rebuild of my Tripath(rip)-based amplifier. That thing was riddled by hilarious amounts of EMI noise and stability problems. The ground system was done exactly like Tripath recommended, separate grounds for power and signal, connecting them to the TA0104 module corresponding pins. After the amplifier destroyed itself with somekind of spurious oscillation, I decided to design a new PCB just against the manufacturer recommendations, I used a multilayer PCB (yes, it has SMA connectors as audio inputs!) with single solid ground layer in it. This has been proven to be successful approach at my work in countless EMC testing situations. The result was that not only the EMI and stability problems were gone, but the amplifier is much more silent than before. My conclusion is that RF-approach seems often work best, even at low frequencies. Although the circuit itself operates at relatively low frequencies, the noise in the environment may be high frequency stuff. My friend also rebuilt his amplifier and got similar results. I wonder why Tripath didn't mention this approach in their application notes

Regards,
Janne

[Simon]

that article you linked to is the biggest load of bullshit i ever read, if the imput impedences is several Kohms if not a mohm or more: so what ! I never thought RCA connectors were frowned on so, they have been an industry standard for audio and vide for years, as for the grounding issue he went on about - the guy needs putting in a mental institute 

[Crazy*Carl]

In terms of high voltage swing needed for high impedance headphones, I take it the goal is to move the transducer to its maximum or ideal position for a given recording?  My ipod can get lots of volume even out of 300ohm headphones, but I guess what it lacks is the voltage swing and lack of movement of the transducer.  Does this make sense?

[Simon]

A higher impedence headphone working on the same output designed for lower impedence headphones will not deliver as much power but you only need mW in headphones so you will probably not have a problem until you try to get high volume from it.

[jahonen]

Actually, it is not delivered power what matters, conventional speakers (including dynamic headphones) are really a voltage driven devices, i.e. constant voltage = constant sound pressure level (in mid-band). Impedance will vary considerably along the frequency, and thus the power delivered.

I have Sennheiser HD600 headphones, which has impedance of 300 ohms. Line-level voltage results so loud sound that it must be attenuated quite a bit to be reasonable. For my self-built DAC+headphone amplifier, I typically attenuate the internal DAC output (which is quite near the line level) about 35 dB from line-level.

Regards,
Janne

[Simon]

well I think most amps are actually controlling current aren't they so the same milliamp output will produce a higher voltage swing which is not above the available voltage will make a loud output. the magnetic field in the coil will sure be as much about the amps as the volts ? like i said people probably do not know how small a power a headphone needs to deliver. I think mine are 1 W thats insane for a headphone but sure give nice sound as the drivers are never "over driven"

[Crazy*Carl]

Actually, it is not delivered power what matters, conventional speakers (including dynamic headphones) are really a voltage driven devices, i.e. constant voltage = constant sound pressure level (in mid-band). Impedance will vary considerably along the frequency, and thus the power delivered.

I have Sennheiser HD600 headphones, which has impedance of 300 ohms. Line-level voltage results so loud sound that it must be attenuated quite a bit to be reasonable. For my self-built DAC+headphone amplifier, I typically attenuate the internal DAC output (which is quite near the line level) about 35 dB from line-level.

Regards,
Janne

How would an ipod insufficiently drive the HD600 then? (I have HD580)
It sure is not lack of volume, my shuffle can get the 300ohm HD580s louder than anyone would ever listen.

[Zad]

Janne mentions a very important point, USB systems recover their clock from the USB signal. In all but one system I have seen, they re-sample with a clock locked to USB. This clock is absolutely horrible, and results in the DAC output jittering like someone shivering in the snow. Doing it asynchronously is *not* easy. Or at least, doing it on a budget commensurate with commercial equipment isn't. Buffering SPDIF, on the other hand, can be done with relative ease.

Similarly, PC audio cards are supplied by the same power supply that is feeding heavy duty current to multiple switched mode regulators, and hundreds of near-by digital components. All this generates power line noise, and RFI too. You just cannot expect to feed a DAC and some op-amps in that sort of environment without compromising the analogue signal. Fortunately, most PC audio is listened to in close proximity of loud fans and hard drives!

I run my PC into a cheap amplifier ripped from an old pair of Creative Labs active speakers. However, it drives a pair of Technics 60W speakers from the early 1990s. In terms of quality, it utterly out performs products from people who really should know better, because they *still* churn out horrible Chinese speaker drivers in plastic enclosures.

[Simon]

well when it comes to plastic PC speakers the other problem is this craze over tiny speakers which are horrible