imperialreign
New Member
- Joined
- Jul 19, 2007
- Messages
- 7,043 (1.13/day)
- Location
- Sector ZZ₉ Plural Z Alpha
System Name | УльтраФиолет |
---|---|
Processor | Intel Kentsfield Q9650 @ 3.8GHz (4.2GHz highest achieved) |
Motherboard | ASUS P5E3 Deluxe/WiFi; X38 NSB, ICH9R SSB |
Cooling | Delta V3 block, XPSC res, 120x3 rad, ST 1/2" pump - 10 fans, SYSTRIN HDD cooler, Antec HDD cooler |
Memory | Dual channel 8GB OCZ Platinum DDR3 @ 1800MHz @ 7-7-7-20 1T |
Video Card(s) | Quadfire: (2) Sapphire HD5970 |
Storage | (2) WD VelociRaptor 300GB SATA-300; WD 320GB SATA-300; WD 200GB UATA + WD 160GB UATA |
Display(s) | Samsung Syncmaster T240 24" (16:10) |
Case | Cooler Master Stacker 830 |
Audio Device(s) | Creative X-Fi Titanium Fatal1ty Pro PCI-E x1 |
Power Supply | Kingwin Mach1 1200W modular |
Software | Windows XP Home SP3; Vista Ultimate x64 SP2 |
Benchmark Scores | 3m06: 20270 here: http://hwbot.org/user.do?userId=12313 |
Been shopping for a new soundcard or speaker setup and wondered what all those specifications slapped on a package really mean? Read reviews about equipment and wondered the same thing? It can all be confusing if you've never read up on audio specs, and even then, a lot of technical jargon is thrown around. I'm going to try and describe what these specs mean in more laymans terms so as to make these specifications more accessible to those not versed in audio equipments, and also try to describe the differences between different speaker hookup connections and equipment, and give some advice to help make component selection easier, and some pointers on adjusting your audio setup as well. I've also included some RMAA output signal spectrums from testing on an unnamed HD audio card, an upper-end Creative X-Fi Elite Pro, and for comparison, the same results from testing the onboard HD audio of my motherboard; and I've tried to explain in a general manner how to interpret these signal spectrums. Signal spectrum tests for audiophiles are like 3DMark06 scores for GPU junkies - they're not meant to be taken as gospel to a components ability, what looks good in spectrum doesn't necessarily mean it will sound good; there are ways to clean up spectrum output that will negatively affect output quality - but we're generally given a good idea of what a component is capable of.
So, let's begin!
For reference, the discernable audio range is accepted as being 20Hz to 20,000Hz - as in, what range of sounds we as humans can hear. Some people can hear beyond this range, some can only perceive sounds below this range; with age, as well, the discernable range narrows. Lower frequency sounds tend to be more easier to hear (and many times even felt), and these sounds will also travel a further distance than higher frequency sounds. Lower frequency tones can cause reverbration, while high frquency tones tend to cause echo.
We can typically discern differences in volume of 3db on average at certain audio frequencies - these values are only accepted as average, and can be different from one person to the next.
Test result spectrum legends: green/white are from testing an unnamed HD audio card, blue/white are a tested X-Fi Elite Pro, and red/white are from testing a SoundMax HD AD1988B onboard chipset . . . all tests were performed at the same settings of 24bit at 96kHz playback in loopback mode using an 8" shielded cable.
I. Basic and common audio hardware quality measurements:
Frequency response: This measurement is a rating of what range of sounds the hardware can produce, from very low frequencies to it's highest frequencies. Ideally, you want the equipment to be able to produce a range exceeding what the human ear can percieve, which would be <20Hz and >20kHz. You will also typically see a db rating for frequency response, and this is a measurement of how much of a volume variation exists across the perceivable hearing range in a playback sample. Ideally, this rating shouldn't exceed a total range of 3db in either direction (plus or minus) from a 0db source.
In this example, you see an almost ideal frequency response - the device in question is able to playback well across the discernable audio range, and there are no spikes or troughs in the output signal within that range, either. In comparison, a perfectly ideal response curve would be flat from 20Hz to 20kHz.
Here we have an even better frequency response spectrum - notice how close the response curve is to being flat . . . a definite marked improvement over the last test.
This result spectrum simply blows my mind - so much so that I ran this test nearly a dozen times; first with the onboard chipset set to Azalia HD standards, then as AC97, I quadruple checked all output settings. I attempted to run the test at 16bit and 24bit outputs, 48kHz and 96kHz sampling rates, I tried using the DirectSound audio mapper, Windows default settings, and with the AD1988B set as default device. I even swapped the line-in plug to the mic-in setting . . . every result returned this ghastly spectacle. Needless to say, this frequency response curve is simply horrible. I am at a loss, if anyone could offer any insight as to the possiblity that I might have done something wrong, I'm more than happy to re-run this test and post the result; but I'm fairly confident that this is correct for this hardware.
Signal to Noise Ratio (SNR): This is a realtive measurement of the strength of an ouput audio signal to the strength of noise within that signal. That is to say, that if SNR is rated at 113db, then there exists a relative difference of 113db between the strength of the output signal to that of the noise signal. In general, anything over 100db is excellent, and the higher the db rating, the better and clearer the signal will be.
Dynamic Range: You will sometimes see this measurement given in place of SNR; dynamic range is a measurement of the signal ratio between the noise floor of an output with no signal to the loudest signal that can be output at a predetermined level, expressed also in terms of db. Again, the higher the db rating specified, the better.
In the example we have an excellent dynamic range spectrum. The spectrum itself is fairly flat through the audio range, with no massive spikes. The single spike at 1kHz is a result of the frequency of the test signal, and is considered normal. The few small spikes in the lower frequencies hint at the possibility of some slight amount of distortion, but lower frequency tones tend to be able to mask this.
Another excellent spectrum. Although we don't see the same flatness as we did with the last test, instead we see a uniform and consistent waveform here. The spectrum itself exists even lower down the scale than our last test, there are no spikes of major importance at any harmonic level.
Here, we see a decent dynamic range spectrum, again, the spectrum waveform is fairly flat through the audio range; although coming in rather high through the lower frequencies; the spikes towards the higher frequencies are not entirelly favorable, though. Although the spikes are not massive in nature, they are still there where ideally in a "HD" component, they should not be. These spikes alude to quite a bit of distortion being present.
Total Harmonic Distortion + Noise (THD+N): A measurement of the amount of unwanted impurities in a given signal. Whereas Total Harmonic Distortion measures unwanted signals in the output, THD+N measures everything that manages to interefere with the output signal, from EMI to harmonic irregularities. Ideally, THD+N would be 0% at any tested level, meaning that the signal being sent into the system is identical to the one being played out of the system; but we live in an imperfect world where our audio equipment will pick up interference and noise and pass it along through the output signal. Be careful reading these specs as well, as an example it's very easy to mistake 0.013% for being better than 0.000013% (the latter is much smaller and closer to ideal 0% than the former is).
In the example, we have an acceptable THD spectrum. It's by far not the greatest, but still acceptable. Of slight concern are the spikes above the 1kHz range - it's more preferable to see spikes like that below 1kHz, where impurities in the signal are harder to hear. Ideally, the THD spectrum would be as flat as possible, with very little spikes at each harmonic level, and no scaling of the spectrum. The massive spike at 1kHz is generally ignored, as the test signal used is 1kHz, the signal will be it's loudest at this harmonic. It's acceptable to see spikes at each harmonic level above 1k as well, but we don't want to see spikes in between these harmonic levels (for example, a spike a 2kHz is ok, while a spike at 1.5kHz would not be). The lower the spectrum exists on the graph as well is more prefered.
With this example, we have another really good THD spectrum. The primary waveform is not entirelly flat, but instead uniform and consistent, and runs the length of the audio range without any sloping. There are a few spikes at a few of the midrange harmonic levels, but acceptable.
Here, we have a decent THD spectrum; definitely not the greatest, as again we see quite a few spikes at each harmonic level above 1kHz, and, IMO, this is representative of too many irregularities for a "HD" component. The results seen above are more preferable when compared to this, are would exhibit fewer negative influences during real audio playback, especially at higher volume levels.
Stereo Crosstalk: This is a measurement of how much an output signal on one channel crosses into a seperate output channel. Typically measured in db as well, by applying a nominal signal level to channelA and reading for that signal on channelB, the db level that the signal is picked up at relative to channelA is taken as measurement. Usually, you'll see the rating as a negative db level, and the lower the better. The higher the crosstalk rating, the less actual channel seperation you'll have, as the two channels will sound more similar to each other as each signal is crossing between channels.
In the example, we have a good crosstalk spectrum. Although there is slightly more channel crossing taking place on the left channel as compared to the right towards the end of the spectrum, which tends to be normal, the actual db levels is still superb, and any actual channel "bleeding" during playback would go unnoticed on all but the most elite setups.
Here, we have an excellent crosstalk spectrum, a well defined climb in the signal which again tends to be normal at such high frequencies. Notice, also, that the spectrum results do extend further below the image on the graph, but I kept the image as it is considering that all these graphs are at the same formatting.
Surprisingly, this represensts a bit better crosstalk spectrum as compared to the first result above; neither channel appears to be bleeding over more. Although, the lower frequencies are a bit unprefered, but easy to live with as most of these frequencies in a multi-speaker setup would be passed on to a subwoofer rather than your mid-ranges.
Intermodular Distortion + noise (IMD+N): A measurement of all impurities in an amplified audio signal that are not harmonically related to the source signal. A little more in-depth than THD+N, this rating is more relative to audio quality as the impurities measured by IMD can cause an audio signal to sound unpleasant, garbled, harsh, distorted or otherwise - usually the result of poor internal or degrading electrical components. Unlike with THD, though, poor IMD can introduce unwanted sounds and tones into a playback, instead of just distorting the signal. The measurement is typically given as a percentage as well, and like with THD+N, the closer to 0% the better.
What we see in this example is a representation of decent IMD+N. The two large spikes, at 60Hz and 7kHz are the result of the two tones used to conduct the test. The spectrum waveform should most flat and even across the audio range, with ideally no other spikes present. The small spikes following the 60Hz and 7kHz range are not ideal, but acceptable, as the spikes themselves don't climb to high at all.
Here we have an excellent IMD+N result - the primary waveform is uniform and consistent, the test signals are clearly seen at 60Hz and 7kHz, and no noteworthy spikes at any of the harmonic levels. The couple of spikes that are apparent are rather small and don't represent too much of a chance of any disturbance to playback.
And this, IMO, about sums it all up - a very poor IMD+N result. Again, the 60Hz and 7kHz tones are easily defined, but every single spike at each harmonic level before, between and beyond the two test ranges bodes very ill of the output quality capabilites of this component. Even at low volume there stands the possibility of subjecting audio output to serious distortion levels.
====================================================================
So, then, how could we summarize all this as a rule of thumb?
The most important specs to look out for, though, are Frequency Response, SNR or Dynamic Range, and IMD
So, let's begin!
For reference, the discernable audio range is accepted as being 20Hz to 20,000Hz - as in, what range of sounds we as humans can hear. Some people can hear beyond this range, some can only perceive sounds below this range; with age, as well, the discernable range narrows. Lower frequency sounds tend to be more easier to hear (and many times even felt), and these sounds will also travel a further distance than higher frequency sounds. Lower frequency tones can cause reverbration, while high frquency tones tend to cause echo.
We can typically discern differences in volume of 3db on average at certain audio frequencies - these values are only accepted as average, and can be different from one person to the next.
Test result spectrum legends: green/white are from testing an unnamed HD audio card, blue/white are a tested X-Fi Elite Pro, and red/white are from testing a SoundMax HD AD1988B onboard chipset . . . all tests were performed at the same settings of 24bit at 96kHz playback in loopback mode using an 8" shielded cable.
I. Basic and common audio hardware quality measurements:
Frequency response: This measurement is a rating of what range of sounds the hardware can produce, from very low frequencies to it's highest frequencies. Ideally, you want the equipment to be able to produce a range exceeding what the human ear can percieve, which would be <20Hz and >20kHz. You will also typically see a db rating for frequency response, and this is a measurement of how much of a volume variation exists across the perceivable hearing range in a playback sample. Ideally, this rating shouldn't exceed a total range of 3db in either direction (plus or minus) from a 0db source.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Ffreqresp.jpg&hash=c2fe846a01b71e128fe06caf3e12fda3)
In this example, you see an almost ideal frequency response - the device in question is able to playback well across the discernable audio range, and there are no spikes or troughs in the output signal within that range, either. In comparison, a perfectly ideal response curve would be flat from 20Hz to 20kHz.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Ffreqresp-blue.jpg&hash=e4b9a5acd239a94fe143f16fb1b01766)
Here we have an even better frequency response spectrum - notice how close the response curve is to being flat . . . a definite marked improvement over the last test.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Ffreqresp_red.jpg&hash=8a163b623a6bedfac49fbc5c7babb208)
This result spectrum simply blows my mind - so much so that I ran this test nearly a dozen times; first with the onboard chipset set to Azalia HD standards, then as AC97, I quadruple checked all output settings. I attempted to run the test at 16bit and 24bit outputs, 48kHz and 96kHz sampling rates, I tried using the DirectSound audio mapper, Windows default settings, and with the AD1988B set as default device. I even swapped the line-in plug to the mic-in setting . . . every result returned this ghastly spectacle. Needless to say, this frequency response curve is simply horrible. I am at a loss, if anyone could offer any insight as to the possiblity that I might have done something wrong, I'm more than happy to re-run this test and post the result; but I'm fairly confident that this is correct for this hardware.
Signal to Noise Ratio (SNR): This is a realtive measurement of the strength of an ouput audio signal to the strength of noise within that signal. That is to say, that if SNR is rated at 113db, then there exists a relative difference of 113db between the strength of the output signal to that of the noise signal. In general, anything over 100db is excellent, and the higher the db rating, the better and clearer the signal will be.
Dynamic Range: You will sometimes see this measurement given in place of SNR; dynamic range is a measurement of the signal ratio between the noise floor of an output with no signal to the loudest signal that can be output at a predetermined level, expressed also in terms of db. Again, the higher the db rating specified, the better.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fdynrang.jpg&hash=7b01b3cbdd925305cc6b82ae7ab6f62f)
In the example we have an excellent dynamic range spectrum. The spectrum itself is fairly flat through the audio range, with no massive spikes. The single spike at 1kHz is a result of the frequency of the test signal, and is considered normal. The few small spikes in the lower frequencies hint at the possibility of some slight amount of distortion, but lower frequency tones tend to be able to mask this.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fdynrang_blue.jpg&hash=a5fadb0abd40d6bdbe99866a0a7b70b1)
Another excellent spectrum. Although we don't see the same flatness as we did with the last test, instead we see a uniform and consistent waveform here. The spectrum itself exists even lower down the scale than our last test, there are no spikes of major importance at any harmonic level.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fdynrang_red.jpg&hash=a1c578dd93a0c150a665ebea0cf24539)
Here, we see a decent dynamic range spectrum, again, the spectrum waveform is fairly flat through the audio range; although coming in rather high through the lower frequencies; the spikes towards the higher frequencies are not entirelly favorable, though. Although the spikes are not massive in nature, they are still there where ideally in a "HD" component, they should not be. These spikes alude to quite a bit of distortion being present.
Total Harmonic Distortion + Noise (THD+N): A measurement of the amount of unwanted impurities in a given signal. Whereas Total Harmonic Distortion measures unwanted signals in the output, THD+N measures everything that manages to interefere with the output signal, from EMI to harmonic irregularities. Ideally, THD+N would be 0% at any tested level, meaning that the signal being sent into the system is identical to the one being played out of the system; but we live in an imperfect world where our audio equipment will pick up interference and noise and pass it along through the output signal. Be careful reading these specs as well, as an example it's very easy to mistake 0.013% for being better than 0.000013% (the latter is much smaller and closer to ideal 0% than the former is).
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FTHD.jpg&hash=2ced7e54a0f71fbf0304102209492db3)
In the example, we have an acceptable THD spectrum. It's by far not the greatest, but still acceptable. Of slight concern are the spikes above the 1kHz range - it's more preferable to see spikes like that below 1kHz, where impurities in the signal are harder to hear. Ideally, the THD spectrum would be as flat as possible, with very little spikes at each harmonic level, and no scaling of the spectrum. The massive spike at 1kHz is generally ignored, as the test signal used is 1kHz, the signal will be it's loudest at this harmonic. It's acceptable to see spikes at each harmonic level above 1k as well, but we don't want to see spikes in between these harmonic levels (for example, a spike a 2kHz is ok, while a spike at 1.5kHz would not be). The lower the spectrum exists on the graph as well is more prefered.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FTHD_blue.jpg&hash=5f223ae4701a73eb9daef54026b49c1a)
With this example, we have another really good THD spectrum. The primary waveform is not entirelly flat, but instead uniform and consistent, and runs the length of the audio range without any sloping. There are a few spikes at a few of the midrange harmonic levels, but acceptable.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FTHD_red.jpg&hash=73ede173fe33e0ff102bef477e596397)
Here, we have a decent THD spectrum; definitely not the greatest, as again we see quite a few spikes at each harmonic level above 1kHz, and, IMO, this is representative of too many irregularities for a "HD" component. The results seen above are more preferable when compared to this, are would exhibit fewer negative influences during real audio playback, especially at higher volume levels.
Stereo Crosstalk: This is a measurement of how much an output signal on one channel crosses into a seperate output channel. Typically measured in db as well, by applying a nominal signal level to channelA and reading for that signal on channelB, the db level that the signal is picked up at relative to channelA is taken as measurement. Usually, you'll see the rating as a negative db level, and the lower the better. The higher the crosstalk rating, the less actual channel seperation you'll have, as the two channels will sound more similar to each other as each signal is crossing between channels.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fxtalk.jpg&hash=394aff940581321db771e475186635cb)
In the example, we have a good crosstalk spectrum. Although there is slightly more channel crossing taking place on the left channel as compared to the right towards the end of the spectrum, which tends to be normal, the actual db levels is still superb, and any actual channel "bleeding" during playback would go unnoticed on all but the most elite setups.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fxtalk_blue.jpg&hash=8f32d3d1add98fe69e27a3070af8125d)
Here, we have an excellent crosstalk spectrum, a well defined climb in the signal which again tends to be normal at such high frequencies. Notice, also, that the spectrum results do extend further below the image on the graph, but I kept the image as it is considering that all these graphs are at the same formatting.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2Fxtalk_red.jpg&hash=8b8dc38165320410031d6bf8d20d4f03)
Surprisingly, this represensts a bit better crosstalk spectrum as compared to the first result above; neither channel appears to be bleeding over more. Although, the lower frequencies are a bit unprefered, but easy to live with as most of these frequencies in a multi-speaker setup would be passed on to a subwoofer rather than your mid-ranges.
Intermodular Distortion + noise (IMD+N): A measurement of all impurities in an amplified audio signal that are not harmonically related to the source signal. A little more in-depth than THD+N, this rating is more relative to audio quality as the impurities measured by IMD can cause an audio signal to sound unpleasant, garbled, harsh, distorted or otherwise - usually the result of poor internal or degrading electrical components. Unlike with THD, though, poor IMD can introduce unwanted sounds and tones into a playback, instead of just distorting the signal. The measurement is typically given as a percentage as well, and like with THD+N, the closer to 0% the better.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FIMD.jpg&hash=5d02682b16224212433b2816d26ef1ed)
What we see in this example is a representation of decent IMD+N. The two large spikes, at 60Hz and 7kHz are the result of the two tones used to conduct the test. The spectrum waveform should most flat and even across the audio range, with ideally no other spikes present. The small spikes following the 60Hz and 7kHz range are not ideal, but acceptable, as the spikes themselves don't climb to high at all.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FIMD_blue.jpg&hash=cacd09e1f02e8ed1e4a396bac06b58ce)
Here we have an excellent IMD+N result - the primary waveform is uniform and consistent, the test signals are clearly seen at 60Hz and 7kHz, and no noteworthy spikes at any of the harmonic levels. The couple of spikes that are apparent are rather small and don't represent too much of a chance of any disturbance to playback.
![](/forums/proxy.php?image=http%3A%2F%2Fimg.techpowerup.org%2F080705%2FIMD_red.jpg&hash=6c621c81bd191e5556fc39518a8123c5)
And this, IMO, about sums it all up - a very poor IMD+N result. Again, the 60Hz and 7kHz tones are easily defined, but every single spike at each harmonic level before, between and beyond the two test ranges bodes very ill of the output quality capabilites of this component. Even at low volume there stands the possibility of subjecting audio output to serious distortion levels.
====================================================================
So, then, how could we summarize all this as a rule of thumb?
- *Frequency response <20Hz and >20kHz
- *SNR and/or Dynamic range >= 100db
- *THD+N as close to 0% as possible
- *IMD+N as close to 0% as possible
- *Crosstalk <= -90db
The most important specs to look out for, though, are Frequency Response, SNR or Dynamic Range, and IMD