我今天在才亲眼看见了maya，而以前我就是用的waveterminal，从我个人的感觉来看，二者有如下的区别：
1、焊接工艺：两个差不太多，焊点基本上比较均匀，圆滑，看得出来，都是大的生产线上下来的东东，而且两块卡好像都不是用的最好的6 层电路板（眼神儿不太好的原因吧，呵呵，不过就即使是非常专业的人士，区别起来也是很困难的），五五开吧。
2、用料：wt在用料上比较马虎，虽然两块卡都使用了大量的分离元件，集成度不高，但是wt的直立元件较多，而且居然电容是一般的电解电容。而m aya的电阻全部是贴片焊接，整个卡感觉很舒服，最重要的一点，maya全部是才用的钽电容，从这点上讲，起码可以肯定，maya的线路噪声在材料角度来看要小于w t，当然，这不排除由于设计的原因造成电路不合理而产生噪声的可能性。但是，maya的接插件比较粗糙，远远比不上wt。
3、驱动支持：wt目前有比较稳定的win9x下的asio、gsif驱动，而maya目前只能使用稳定的asio驱动，但是，maya支持gsif的驱动目前已经有b eta版，从使用的情况来看，效果不佳，不但对VIA芯片主板和AMD的芯片支持不太好，而且对系统本身也很挑剔，不过，maya有个很诱人的独立开发技术：G igaWire，它能使GigaStudio在使用中调用VST的效果器插件！想想那么多好用的VST插件……嘿嘿……同时，GigaWire技术还在GigaStudi o和Cubase VST之间在硬件层次上提供了一条Digital-Digital的音频通道，怎么说呢？以前的音频录制都是用音序器软件的midi信号触发音源，比如G iga发声，音频信号通过声卡的DA转换器和声卡上的AD转换器，最好才变成计算机里面的音频文件（当然，如果你有两台计算机，并且两台计算机上的声卡都有D igital的in/out通道，可以把由音序器（Cubase、Cakewalk等）触发的，由音源（GigaStudio）发出的Digital信号直接D-D、则另当别论），除非你使用 GigaStudio 自带的录音机，所以，音质必然有很大的损耗，而现在……不用我说大家都会明白这意味着什么。
4、主观印象：这个就不好说了，限于监听条件的限制，在听maya时用的音箱是用的haobo的监听一号，而用wt时用的NS10，但是整体感觉好像差不太多，至少主观感觉上没有差到1 000多大洋。在打开soundforge，去掉所有的外置设备以后，在用户状态下（采样率44.1，如果用48的话，噪声还要小），maya的噪声稍稍比w t高一点，大概5、6个db吧（之所以用soundforge而不用wavelab，是因为wavelab的电平显示是取的平均值，不太准确，而且，不要太相信技术资料上的什么－9 x db一类的鬼话，那是在实验室中，非联机状态下测试出的理想值）。
在同一种监听条件下，只听了Live（用aps驱动）和maya，形容词我就不多写了，反正一句话：没法比！
三块卡都使就放音而言，录音由于时间限制没有测试。
以上均为个人感受，呵呵，自己本来就是土人一个，所以结果嘛大家就爱看不看啦……
------------------
＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝
小的我不才，
做了Vegas板的板猪，
肚子里面的东东少啊……
心里面那个慌啊……
还请大家多多光临！！
多多指点！！！
[该信息已经被 fatdog 编辑过.(编辑日期:2月 20, 2001).]
[该信息已经被 糊格格 编辑过.(编辑日期:2月 20, 2001).]
1、焊接工艺：两个差不太多，焊点基本上比较均匀，圆滑，看得出来，都是大的生产线上下来的东东，而且两块卡好像都不是用的最好的6 层电路板（眼神儿不太好的原因吧，呵呵，不过就即使是非常专业的人士，区别起来也是很困难的），五五开吧。
2、用料：wt在用料上比较马虎，虽然两块卡都使用了大量的分离元件，集成度不高，但是wt的直立元件较多，而且居然电容是一般的电解电容。而m aya的电阻全部是贴片焊接，整个卡感觉很舒服，最重要的一点，maya全部是才用的钽电容，从这点上讲，起码可以肯定，maya的线路噪声在材料角度来看要小于w t，当然，这不排除由于设计的原因造成电路不合理而产生噪声的可能性。但是，maya的接插件比较粗糙，远远比不上wt。
3、驱动支持：wt目前有比较稳定的win9x下的asio、gsif驱动，而maya目前只能使用稳定的asio驱动，但是，maya支持gsif的驱动目前已经有b eta版，从使用的情况来看，效果不佳，不但对VIA芯片主板和AMD的芯片支持不太好，而且对系统本身也很挑剔，不过，maya有个很诱人的独立开发技术：G igaWire，它能使GigaStudio在使用中调用VST的效果器插件！想想那么多好用的VST插件……嘿嘿……同时，GigaWire技术还在GigaStudi o和Cubase VST之间在硬件层次上提供了一条Digital-Digital的音频通道，怎么说呢？以前的音频录制都是用音序器软件的midi信号触发音源，比如G iga发声，音频信号通过声卡的DA转换器和声卡上的AD转换器，最好才变成计算机里面的音频文件（当然，如果你有两台计算机，并且两台计算机上的声卡都有D igital的in/out通道，可以把由音序器（Cubase、Cakewalk等）触发的，由音源（GigaStudio）发出的Digital信号直接D-D、则另当别论），除非你使用 GigaStudio 自带的录音机，所以，音质必然有很大的损耗，而现在……不用我说大家都会明白这意味着什么。
4、主观印象：这个就不好说了，限于监听条件的限制，在听maya时用的音箱是用的haobo的监听一号，而用wt时用的NS10，但是整体感觉好像差不太多，至少主观感觉上没有差到1 000多大洋。在打开soundforge，去掉所有的外置设备以后，在用户状态下（采样率44.1，如果用48的话，噪声还要小），maya的噪声稍稍比w t高一点，大概5、6个db吧（之所以用soundforge而不用wavelab，是因为wavelab的电平显示是取的平均值，不太准确，而且，不要太相信技术资料上的什么－9 x db一类的鬼话，那是在实验室中，非联机状态下测试出的理想值）。
在同一种监听条件下，只听了Live（用aps驱动）和maya，形容词我就不多写了，反正一句话：没法比！
三块卡都使就放音而言，录音由于时间限制没有测试。
以上均为个人感受，呵呵，自己本来就是土人一个，所以结果嘛大家就爱看不看啦……
------------------
＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝＝
小的我不才，
做了Vegas板的板猪，
肚子里面的东东少啊……
心里面那个慌啊……
还请大家多多光临！！
多多指点！！！
[该信息已经被 fatdog 编辑过.(编辑日期:2月 20, 2001).]
[该信息已经被 糊格格 编辑过.(编辑日期:2月 20, 2001).]
我用的是LIVE的标准版,做工比数码版的要好.至于APS音质的问题,一方面硬件的质量有影响,在软件方面,由于创新公司的环境音效使LIVE WARE驱动发出的音质并不是硬件本来的声音(如果装了LIVEWARE,驱动会自动给你的声卡加上效果,只不过你不一定听的出这是加过效果器的声音) ,APS本质上说并没有提高硬件的品质,只不过是还硬件本来面目而已.什么样档次的卡要有相应的回放设备,其实,当今大多数人对高保真的声音并不在行, 就象看星球大战和天文记录片一样,味道是完全不同的.我自从用了APS后,就已经忘了那个LIVEWARE了.当然,我录音用的是WT2496.
LIVE的输字输入是假的，要通过他内部的一个频率CONVERTER。LIVE内部是用48KH工作的，而他的输字输入口可接收多种频率，无论是48K H还是44.1KH都要通过这个频率转换器和它内置的数字MIXER，所以它的数字输入的电平竟然是可调的。而专业声卡的输字输入一般都是直通（T RANSFER），不可调的。原来什么样，进去还是什么样。而LIVE是“主流”产品（大众产品），要考虑到多种需要和价格，它的频率转换器不可能非常昂贵和专业，因此L IVE的输字输入是有点变形的，特别是高音，发尖。对这个问题，98年LIVE刚推出是，CREATIVE的网站上专们有讨论，EMU的首席设计师戴夫也是承认的。原文如下：
CREATIVE AUDIO EXPERTS COMMENT ON SOUND BLASTER LIVE!'S AUDIO SPECIFICATIONS
To clarify certain confusion over audio specifications published by Creative, the editorial team interviewed a bunch of Creative audio experts and engineers in a technical session. We asked them to comment on why there are differing audio specifications of Sound Blaster Live! published by individuals on the Internet and to provide specific details.
SUMMARY
Creative clarified the discrepancies between the measurements posted on the www.sblive.com site and those done by end users. Basically, the discrepancies resulted from using different test standards and conditions. Certain users seemed to have confused digital with analog performance. Creative confirmed that the measurements published were based on the IEC publication 268-3 (1988), which is widely used in the audio industry. Users who want to get similar results in their measurements of Sound Blaster Live! can refer to Creative's guidelines.
EXCERPTS FROM THE INTERVIEW
Did Creative Technology publish the specifications of the Sound Blaster Live! during its launch?
HG Tan, Vice President, Audio & VLSI Product Group, AVP:
Yes, we distributed two types of technical specifications.
We highlighted the typical value of a small but important part of the pure analog performance in the product package, marketing materials and on our website.
To help the industrial reviewer appreciate the excellent performance of Sound Blaster Live!, a full measurement report was distributed together with the review samples. Test method and procedure were also provided as an appendix in the full measurement report. The first 50 units of samples and their measurement reports were serialized.
What is the reason for highlighting the pure analog performance?
Well, first of all, the pure analog performance is an important indicator of the end user's direct impression of the baseline quality of a PC sound card. When the PC sound card is not playing anything and the volume control of the amplifier or power speakers is set to its maximum position, the pure analog performance determines the level of "hiss" and "hum" the end user is going to get.
Secondly, all other sound card functions: playing back of wave files, synthesizing of MIDI, mixing of input sources as well as generation of Sound Effects are all built and reside on top of this vital baseline performance.
We understand that there has been confusion in the Internet regarding the Sound Blaster Live!'s audio specifications. Certain individuals have published specifications that differ. Can you clarify?
Basically there are three possible situations:
The correlation of the measurement: The measurement method, test conditions, the setting up of the sound card as well as the equipment used, are all factors that affect the integrity of the test results. We discovered that some of the measurements published by individuals on the Internet were not done on industrial standard test equipment.
Our measurement results (the figures) were not fully and correctly presented when other parties reproduced or cited our specifications. For example, if someone just mentioned that our sound card has a Signal-to-Noise Ratio (SNR) of 96 dB without mentioning that it is the pure analog performance, it may be misinterpreted by another person as the Digital-to-Analog or Analog-Digital-Analog performance.
Some people may have incorrectly compared the published analog performance with the test results published by certain websites. Those tests were done on full loop-back (Analog-Digital-Analog), Analog-to-Digital and Digital-to-analog operation modes that yield a lower reading, but still good enough for Sound Blaster Live! to be rated as "Very good" by the website. In fact, based on those reports published by the same website in an apple-to-apple comparison basis, we can proudly say that there is no competitor card at the same price point that comes close to the quality of Sound Blaster Live!.
Achieving correlation between measurements by different teams at different locations is difficult under the best of circumstances. We are always happy to discuss measurement techniques with independent labs and help them find these subtle sources of error to ensure that our measurements can be reproduced whenever possible.
Dr. Lim, what type of measurement method or standard does Creative Technology adopt?
Dr Lim Jit Wee, Product Development Director, AVP:
It should be stated here that sound card measurements are still not very well established in the industry. There have been several proposals brought out in the last 18 months but no standard has been agreed upon by all the sound card manufacturers to be used in their benchmarking practices. Only recently has some very basic measurement method defined by PC98 under Audio Quality Test Methods section been gradually gaining industrial acceptance.
The standard that Creative has adopted is a "best of breed" based on the open standard from IEC Publication 268-3 (1988). In some cases where the IEC standard may not be totally relevant to our situation, we conducted objective measurements with the intention of attaining meaningful results.
Was the measurement method also adopted for the Sound Blaster AWE64 Gold? How do you compare the Sound Blaster Live! to the Sound Blaster AWE64 Gold which has won many prestigious awards and favorable press reviews?
Yes, the same measurement method has been used for many years on many of our products including the Sound Blaster AWE64 Gold. The Sound Blaster Live! is at least 6 dB better on SNR with compared to the Sound Blaster AWE64 Gold. That means Sound Blaster Live! produces 50% less noise than the Sound Blaster AWE64 Gold.
How did your team achieve this?
This was possible because of our expertise in PCB layout and silicon chip design. Our strategic relationships enable us to work closely and exclusively with our key vendors to fine-tune the performance of the chips to meet our requirements and yet keep it at a reasonable cost.
Edward, a noise floor plot has also been published. How is this noise floor plot and Signal-to-Noise Ratio (SNR) related?
Edward Law, Technical Marketing Manager, AVP:
The relationship between the noise floor plot and Signal-to-Noise Ratio (SNR) is quite complex, the only similarity is that both are measurements of noise. The differences are:
Noise floor plot is constructed by the measured noise figures at each sampling frequency across the frequency spectrum, while SNR is a single accumulated noise figure measured for the entire frequency band.
Noise floor is measured without referring to any input signal, so it provides an absolute amplitude across the spectrum, while SNR is a measure of signal clarity with respect to noise.
Noise floor is not A-weighted, while SNR is A-weighted to represent ear sensitivity to frequency spectrum.
The SNR measurement of two different sound cards can be the same, but the shape of the noise floor can still be different. A sound card with a flat noise floor without any spurious tone (spikes) throughout the spectrum is far more desirable than the one with one or more spurious tones within the audible frequency range, although adding all the noise levels might give the same noise in the SNR measurement.
As usual, we measure separately the SNR performance from each input path to the line output. As for the noise floor plot we published, we have done the specific measurement with all the input amplifiers and mixer turned ON. From the flat noise floor, one can easily appreciate that the card is quiet throughout the whole audible frequency range.
How are the SNR and dynamic range related?
Basically, dynamic range differs from the Signal-to-Noise Ratio in the way the noise is measured. In SNR, the noise is taken when the signal is shut off, while in dynamic range, the noise is taken when the signal is present but filtered out by a notch filter. In analog measurement, both figures will be similar, but in terms of digital measurement, they may differ substantially. The problem lies with the digital-to-analog converter. If there is no signal at all, the noise is the intrinsic noise of the converter itself. When there is a signal, the quantization will come in and contribute very much to the noise.
There are different opinions in interpreting the noise caused by quantization. Some measurement methods will term the dynamic range to be the same as SNR. A more informative term may be "SNR in the presence of signal".
There are some concerns about the frequency response in Digital Playback and Recording, do you have any comment?
The design and operation of our system is optimized and locked at a sampling frequency of 48KHz. With the input sources and digital S/PDIF output both fixed at a sampling frequency of 48KHz, the typical frequency response is 20 to 22KHz at +/- 1dB.
Our system has the ability and flexibility to handle input sources with different sampling rates. In such operation mode, the Sampling Rate Converter (SRC) will be involved, a slight reduction in high frequency response is expected but it will not affect the listening experience at all.
Dave, let's talk about the S/PDIF output of Sound Blaster Live!. Why is the S/PDIF output locked at 48kHz?
Dave Rossum, Chief Scientist:
It's important to remember that the Sound Blaster Live! is much more than just a wavetable synthesizer or a CD playback device. At its heart is the EMU10K1 effects engine, a powerful DSP performing, among other things, mixing of all the various functions in the digital domain. To mix digital audio signals, they must all be at EXACTLY the same sample rate, even deviations of a few parts per million must be eliminated. So when we went to design the EMU10K1, we had to choose a single master sample rate at which the mixer would operate, and of course we had to design sample rate converters to change any incoming audio to match this sample rate. This is the technology required to achieve digital mixing - that's why it's first featured by Sound Blaster Live!.
It was obvious that either 44.1 kHz, the CD standard, or 48 kHz, the professional audio and DVD standard, were the only possible choices for the master sample rate. We picked 48 kHz for a variety of reasons. First, if we were processing incoming audio at nominally 48 kHz, use of 44.1 kHz would lose information. Second, even in preparation of 44.1 kHz CDs in professional studios today, 48 kHz is the preferred standard, with a final conversion to 44.1 kHz as the last step. Third, the nearly twice larger "guard band" (the difference between 20 kHz and the Nyquist frequency) of 24 kHz means virtually every audio process and effect performs much better at 48 kHz than at 44.1 kHz. And finally, the economical AC-97 CODEC operates at 48 kHz.
The S/PDIF outputs provide the exact EMU10K1 effects engine outputs, so they operate at this same 48 kHz rate. So why didn't we add sample rate converter to each S/PDIF output to convert the 48 kHz signal back to 44.1 kHz? The primary reason is cost. Since these outputs are "master" signals, they should be treated as very high quality and as such the sample rate converters, to be useful, would be fairly expensive in silicon. Furthermore, since the EMU10K1 is clocked off the AC-97 CODEC master clock which is based on 48 kHz, a separate clock and crystal would be necessary to support 44.1 kHz. At least, if audio quality were important, since a standard phase locked loop or divider system would introduce too much jitter. And since the job could be done externally fairly straightforward, we felt making all users bear the cost burden of this feature for the few who would use it was a poor trade-off. Also, we are trying to promote the uniform system design which will be found in the modern digital studio in which the digital inputs bear the burden of sample rate normalization. For those few users who for some reason seem to need a 44.1 kHz S/PDIF output, they'll either have to purchase a sample rate converter box (Analog Devices makes a <$20 chip which will do the job nicely), or (as I would tend to recommend) convert their operation to 48 KHz throughout and if 44.1 kHz is needed for CD, have this jobbed out at CD production like most people do.
Why must all the sound sources pass through the sampling rate converters?
It's not actually true that every sound source must pass through sample rate conversion. If the Sound Blaster Live! can serve as the master 48 kHz clock for the sound source (as provided for in the newest operating system), then the source need not be converted. But any external source can't possibly know our EXACT 48 kHz rate, and so its sample rate must be normalized to ours, even if the deviation is very small. This is a fact of digital mixing.
Once again, those most familiar with practices in older digital studios will ask why we didn't make accommodations for supply and receiving "AES black" synchronization information to eliminate the necessity of sample rate conversion. The reason is that studio synch is simply too complex an issue for most consumers to grasp - sample rate conversion (at least by E-mu) sounds great and makes the digital patch cord behave just like an analog one.
Why didn't Creative publish the digital performance specifications?
Edward Law:
Measuring digital performance is not as standardized as measuring analog performance. Take Dynamic Range as an example, some measurements quote it as SNR, but there are also different opinions on how to characterize the signal. Some think that quantization noise is more related to THD, and should be dithered to minimize the THD effect. Then there are differing opinions in dithering methods. Another question is what frequency should be used. Following the analog field it should be 1000 Hz, but some also suggest that 997Hz should be used instead.
Another factor affecting this is the level of the signals. If it is too high, then it will introduce harmonics due to saturation. If it is too low, the contribution due to quantization noise is quite substantial. Finally, the use of the weighted filter will further complicate the measurement. While A-weighted filter is commonly accepted in the analog field, some found that other filters can eliminate the coupling of noise from digital signal to analog signal and is more suitable for digital performance.
We would like to emphasize that Creative Technology has internally performed all varieties of measurement on Sound Blaster Live! and the measurement results meet and exceed the measurement objective and expectation. Creative believes that publishing digital audio specifications in general without first stating the influence of the above mentioned factors would probably generate even more confusion and controversy because of the lack of standard digital measurement methods. Based on PC98 / Audio Quality Test method, we have attached the digital performance in the following table.
We are equally proud of our digital signal processing technology within the Sound Blaster Live!. Its performance certainly has exceeded end user's expectations.
CREATIVE AUDIO EXPERTS COMMENT ON SOUND BLASTER LIVE!'S AUDIO SPECIFICATIONS
To clarify certain confusion over audio specifications published by Creative, the editorial team interviewed a bunch of Creative audio experts and engineers in a technical session. We asked them to comment on why there are differing audio specifications of Sound Blaster Live! published by individuals on the Internet and to provide specific details.
SUMMARY
Creative clarified the discrepancies between the measurements posted on the www.sblive.com site and those done by end users. Basically, the discrepancies resulted from using different test standards and conditions. Certain users seemed to have confused digital with analog performance. Creative confirmed that the measurements published were based on the IEC publication 268-3 (1988), which is widely used in the audio industry. Users who want to get similar results in their measurements of Sound Blaster Live! can refer to Creative's guidelines.
EXCERPTS FROM THE INTERVIEW
Did Creative Technology publish the specifications of the Sound Blaster Live! during its launch?
HG Tan, Vice President, Audio & VLSI Product Group, AVP:
Yes, we distributed two types of technical specifications.
We highlighted the typical value of a small but important part of the pure analog performance in the product package, marketing materials and on our website.
To help the industrial reviewer appreciate the excellent performance of Sound Blaster Live!, a full measurement report was distributed together with the review samples. Test method and procedure were also provided as an appendix in the full measurement report. The first 50 units of samples and their measurement reports were serialized.
What is the reason for highlighting the pure analog performance?
Well, first of all, the pure analog performance is an important indicator of the end user's direct impression of the baseline quality of a PC sound card. When the PC sound card is not playing anything and the volume control of the amplifier or power speakers is set to its maximum position, the pure analog performance determines the level of "hiss" and "hum" the end user is going to get.
Secondly, all other sound card functions: playing back of wave files, synthesizing of MIDI, mixing of input sources as well as generation of Sound Effects are all built and reside on top of this vital baseline performance.
We understand that there has been confusion in the Internet regarding the Sound Blaster Live!'s audio specifications. Certain individuals have published specifications that differ. Can you clarify?
Basically there are three possible situations:
The correlation of the measurement: The measurement method, test conditions, the setting up of the sound card as well as the equipment used, are all factors that affect the integrity of the test results. We discovered that some of the measurements published by individuals on the Internet were not done on industrial standard test equipment.
Our measurement results (the figures) were not fully and correctly presented when other parties reproduced or cited our specifications. For example, if someone just mentioned that our sound card has a Signal-to-Noise Ratio (SNR) of 96 dB without mentioning that it is the pure analog performance, it may be misinterpreted by another person as the Digital-to-Analog or Analog-Digital-Analog performance.
Some people may have incorrectly compared the published analog performance with the test results published by certain websites. Those tests were done on full loop-back (Analog-Digital-Analog), Analog-to-Digital and Digital-to-analog operation modes that yield a lower reading, but still good enough for Sound Blaster Live! to be rated as "Very good" by the website. In fact, based on those reports published by the same website in an apple-to-apple comparison basis, we can proudly say that there is no competitor card at the same price point that comes close to the quality of Sound Blaster Live!.
Achieving correlation between measurements by different teams at different locations is difficult under the best of circumstances. We are always happy to discuss measurement techniques with independent labs and help them find these subtle sources of error to ensure that our measurements can be reproduced whenever possible.
Dr. Lim, what type of measurement method or standard does Creative Technology adopt?
Dr Lim Jit Wee, Product Development Director, AVP:
It should be stated here that sound card measurements are still not very well established in the industry. There have been several proposals brought out in the last 18 months but no standard has been agreed upon by all the sound card manufacturers to be used in their benchmarking practices. Only recently has some very basic measurement method defined by PC98 under Audio Quality Test Methods section been gradually gaining industrial acceptance.
The standard that Creative has adopted is a "best of breed" based on the open standard from IEC Publication 268-3 (1988). In some cases where the IEC standard may not be totally relevant to our situation, we conducted objective measurements with the intention of attaining meaningful results.
Was the measurement method also adopted for the Sound Blaster AWE64 Gold? How do you compare the Sound Blaster Live! to the Sound Blaster AWE64 Gold which has won many prestigious awards and favorable press reviews?
Yes, the same measurement method has been used for many years on many of our products including the Sound Blaster AWE64 Gold. The Sound Blaster Live! is at least 6 dB better on SNR with compared to the Sound Blaster AWE64 Gold. That means Sound Blaster Live! produces 50% less noise than the Sound Blaster AWE64 Gold.
How did your team achieve this?
This was possible because of our expertise in PCB layout and silicon chip design. Our strategic relationships enable us to work closely and exclusively with our key vendors to fine-tune the performance of the chips to meet our requirements and yet keep it at a reasonable cost.
Edward, a noise floor plot has also been published. How is this noise floor plot and Signal-to-Noise Ratio (SNR) related?
Edward Law, Technical Marketing Manager, AVP:
The relationship between the noise floor plot and Signal-to-Noise Ratio (SNR) is quite complex, the only similarity is that both are measurements of noise. The differences are:
Noise floor plot is constructed by the measured noise figures at each sampling frequency across the frequency spectrum, while SNR is a single accumulated noise figure measured for the entire frequency band.
Noise floor is measured without referring to any input signal, so it provides an absolute amplitude across the spectrum, while SNR is a measure of signal clarity with respect to noise.
Noise floor is not A-weighted, while SNR is A-weighted to represent ear sensitivity to frequency spectrum.
The SNR measurement of two different sound cards can be the same, but the shape of the noise floor can still be different. A sound card with a flat noise floor without any spurious tone (spikes) throughout the spectrum is far more desirable than the one with one or more spurious tones within the audible frequency range, although adding all the noise levels might give the same noise in the SNR measurement.
As usual, we measure separately the SNR performance from each input path to the line output. As for the noise floor plot we published, we have done the specific measurement with all the input amplifiers and mixer turned ON. From the flat noise floor, one can easily appreciate that the card is quiet throughout the whole audible frequency range.
How are the SNR and dynamic range related?
Basically, dynamic range differs from the Signal-to-Noise Ratio in the way the noise is measured. In SNR, the noise is taken when the signal is shut off, while in dynamic range, the noise is taken when the signal is present but filtered out by a notch filter. In analog measurement, both figures will be similar, but in terms of digital measurement, they may differ substantially. The problem lies with the digital-to-analog converter. If there is no signal at all, the noise is the intrinsic noise of the converter itself. When there is a signal, the quantization will come in and contribute very much to the noise.
There are different opinions in interpreting the noise caused by quantization. Some measurement methods will term the dynamic range to be the same as SNR. A more informative term may be "SNR in the presence of signal".
There are some concerns about the frequency response in Digital Playback and Recording, do you have any comment?
The design and operation of our system is optimized and locked at a sampling frequency of 48KHz. With the input sources and digital S/PDIF output both fixed at a sampling frequency of 48KHz, the typical frequency response is 20 to 22KHz at +/- 1dB.
Our system has the ability and flexibility to handle input sources with different sampling rates. In such operation mode, the Sampling Rate Converter (SRC) will be involved, a slight reduction in high frequency response is expected but it will not affect the listening experience at all.
Dave, let's talk about the S/PDIF output of Sound Blaster Live!. Why is the S/PDIF output locked at 48kHz?
Dave Rossum, Chief Scientist:
It's important to remember that the Sound Blaster Live! is much more than just a wavetable synthesizer or a CD playback device. At its heart is the EMU10K1 effects engine, a powerful DSP performing, among other things, mixing of all the various functions in the digital domain. To mix digital audio signals, they must all be at EXACTLY the same sample rate, even deviations of a few parts per million must be eliminated. So when we went to design the EMU10K1, we had to choose a single master sample rate at which the mixer would operate, and of course we had to design sample rate converters to change any incoming audio to match this sample rate. This is the technology required to achieve digital mixing - that's why it's first featured by Sound Blaster Live!.
It was obvious that either 44.1 kHz, the CD standard, or 48 kHz, the professional audio and DVD standard, were the only possible choices for the master sample rate. We picked 48 kHz for a variety of reasons. First, if we were processing incoming audio at nominally 48 kHz, use of 44.1 kHz would lose information. Second, even in preparation of 44.1 kHz CDs in professional studios today, 48 kHz is the preferred standard, with a final conversion to 44.1 kHz as the last step. Third, the nearly twice larger "guard band" (the difference between 20 kHz and the Nyquist frequency) of 24 kHz means virtually every audio process and effect performs much better at 48 kHz than at 44.1 kHz. And finally, the economical AC-97 CODEC operates at 48 kHz.
The S/PDIF outputs provide the exact EMU10K1 effects engine outputs, so they operate at this same 48 kHz rate. So why didn't we add sample rate converter to each S/PDIF output to convert the 48 kHz signal back to 44.1 kHz? The primary reason is cost. Since these outputs are "master" signals, they should be treated as very high quality and as such the sample rate converters, to be useful, would be fairly expensive in silicon. Furthermore, since the EMU10K1 is clocked off the AC-97 CODEC master clock which is based on 48 kHz, a separate clock and crystal would be necessary to support 44.1 kHz. At least, if audio quality were important, since a standard phase locked loop or divider system would introduce too much jitter. And since the job could be done externally fairly straightforward, we felt making all users bear the cost burden of this feature for the few who would use it was a poor trade-off. Also, we are trying to promote the uniform system design which will be found in the modern digital studio in which the digital inputs bear the burden of sample rate normalization. For those few users who for some reason seem to need a 44.1 kHz S/PDIF output, they'll either have to purchase a sample rate converter box (Analog Devices makes a <$20 chip which will do the job nicely), or (as I would tend to recommend) convert their operation to 48 KHz throughout and if 44.1 kHz is needed for CD, have this jobbed out at CD production like most people do.
Why must all the sound sources pass through the sampling rate converters?
It's not actually true that every sound source must pass through sample rate conversion. If the Sound Blaster Live! can serve as the master 48 kHz clock for the sound source (as provided for in the newest operating system), then the source need not be converted. But any external source can't possibly know our EXACT 48 kHz rate, and so its sample rate must be normalized to ours, even if the deviation is very small. This is a fact of digital mixing.
Once again, those most familiar with practices in older digital studios will ask why we didn't make accommodations for supply and receiving "AES black" synchronization information to eliminate the necessity of sample rate conversion. The reason is that studio synch is simply too complex an issue for most consumers to grasp - sample rate conversion (at least by E-mu) sounds great and makes the digital patch cord behave just like an analog one.
Why didn't Creative publish the digital performance specifications?
Edward Law:
Measuring digital performance is not as standardized as measuring analog performance. Take Dynamic Range as an example, some measurements quote it as SNR, but there are also different opinions on how to characterize the signal. Some think that quantization noise is more related to THD, and should be dithered to minimize the THD effect. Then there are differing opinions in dithering methods. Another question is what frequency should be used. Following the analog field it should be 1000 Hz, but some also suggest that 997Hz should be used instead.
Another factor affecting this is the level of the signals. If it is too high, then it will introduce harmonics due to saturation. If it is too low, the contribution due to quantization noise is quite substantial. Finally, the use of the weighted filter will further complicate the measurement. While A-weighted filter is commonly accepted in the analog field, some found that other filters can eliminate the coupling of noise from digital signal to analog signal and is more suitable for digital performance.
We would like to emphasize that Creative Technology has internally performed all varieties of measurement on Sound Blaster Live! and the measurement results meet and exceed the measurement objective and expectation. Creative believes that publishing digital audio specifications in general without first stating the influence of the above mentioned factors would probably generate even more confusion and controversy because of the lack of standard digital measurement methods. Based on PC98 / Audio Quality Test method, we have attached the digital performance in the following table.
We are equally proud of our digital signal processing technology within the Sound Blaster Live!. Its performance certainly has exceeded end user's expectations.