找回密码
 -注册-
查看: 2493|回复: 26

老高发烧随笔——Once it is heard, it is hard to live without.二楼编辑部分完成

[复制链接]
发表于 2018-2-22 00:11 | 显示全部楼层 |阅读模式 来自 北京市
本帖最后由 andygaof 于 2018-2-22 23:59 编辑

Why ladder DACs?                                                

                                                                                                                                                                                                                                                                                                
                                                        Ladder DACs – Vince’s View!

The MSB Ladder DAC is a cutting edge, statement product technology designed by MSB’s fanatical engineering team. Unlike virtually all other DACs available today, the MSB DACs do not use any off-the-shelf DAC chips or digital filter chips. Everything is designed and built in house including digital filters, reclocking scheme and other special internal processes that takes advantage of the extraordinary performance of MSB’s proprietary hyper bandwidth DAC modules. The MSB DACs offer unprecedented sonic performance and value.
Digital since the mid 80’s
When CDs were designed in the mid 80’s, the Redbook standard was implemented that dictated a certain digital code written in bits consisting of 1’s and 0’s, that could be stored in media like the CD disc. The specific Redbook code would describe an analog music waveform and upon the conversion of the digital bits back to analog, that waveform would be decoded and reproduced. Lacking the inexpensive processing chips that are available today the conversion to digital was done by a resistor ladder. The bits were sent into a simple resistor ladder called an R2R DAC (resistor to resistor Digital To Analog Converter).
vinceLadderDAC.png

For conceptual purposes let’s look at the simple R2R ladder diagram. Each bit has a certain voltage or “push” into the ladder. Each bit (pulsed voltage) is a small part of the original analog sine wave signal. The bits are sent into the rungs of the ladder (D0-D7), all bits blending their individual values into the resistor network. Each of these bits are driven into the resistor ladder on the individual rungs contributing its voltage or “push”. These individual bits with their short burst of voltage blend together in the network of resistors resulting in a new continuous voltage which is intended to be the exact copy of the original analog signal that is tapped off at the top and bottom of the ladder. To convert the digital bits to analog precisely, the value of each resistor has to be the exact correct ohm-value requiring very high precision in the manufacture of the ladder. The ladder DAC like the original Philips TDA 1541 was of limited accuracy and very expensive to make. Yet its design is potentially superior because the conversion is done through simple resistors (passive) with no processing. Because the ladder contains only passive resistors, the speed can be extremely high.
Later when the cost of processing chips came down, The Delta Sigma DAC was introduced using a 1 bit method and a form of computer micro processing to decode the digital signal. The Delta sigma is most accurate at the top of its dynamic range but losing resolution at low signal levels (softer sounds). Because the Delta Sigma design converts by processing, extensive analog filtering is required. This additional analog filtering introduces its own set of inaccuracies like phase shift in the output signal.
MSB Sign Magnitude R2R DAC
MSB has always known the ladder was a superior conversion method and introduced the world’s first discrete 24 bit Sign Magnitude R2R Ladder DAC. The term Sign Magnitude describes the special architecture we use that dramatically improves the sound of low level signals. Instead of always starting at the lower limit of the signal and adding voltage to reach the music signal, we start at the midpoint, or zero crossing, where music is quiet, and we either add or subtract voltage to get the required signal. Because this requires a much smaller addition or subtraction on average, it can be done much more accurately.
MSB has designed and built a new proprietary R2R architecture that far exceeds the performance of the original ladder DAC design. The performance of a Ladder DAC is defined by the precision of the resistors. There are hundreds of very expensive aerospace grade resistors on each MSB module producing a DAC with a level of precision that is unheard of. The noise floor (the lowest sound that can reproduced), is much lower than most test systems can even measure. But most important to MUSIC rather than TEST SIGNALS, and very different from Delta Sigma DACs the MSB DAC module are most accurate with signals crossing zero, where music actually exists.
Sample Rates
When we talk about digital sample rate we mean the speed of the bits in kHz. CDs are 44.1 kHz (44.1 thousand times per second), and higher resolutions typically go up to 192 kHz with the next generation of hi-res recordings just now available at 384 kHz. MSB DAC modules can operate beyond 5 mHz (5 million times per second), so these modules can receive and reproduce all current formats and conceivable future formats for many years to come. Unlike Delta Sigma DACs the performance of the MSB ladder DAC actually gets better with more bit depth and recording resolution. Low level resolution is recovered to an extraordinary degree. Please see How DACs Work for more information.
MSB Input Processing
The amazing performance of these DACs would be under-utilized if they did not receive the best high-resolution digital signal possible. The front end of MSB’s DAC IV series starts at the input receiver where MSB’s proprietary reclocking scheme reduces incoming jitter to under 7 Pico seconds and in some cases less than 2 ps. Since there was no source that could challenge the MSB DAC IV’s resolution and jitter, MSB designed an analog-to-digital converter (ADC), that was at least as good as the DAC. This later became a product – the MSB Studio ADC. Then MSB found it was it was necessary to design and build an in-house jitter measurement system to measure the final jitter performance. While some published specs show very low numbers the results are limited by the measurement equipment available.
After the jitter is removed the data signal agrees in time with the clock signal. Then the datastream is sent to MSB’s proprietary Intersample Harshness Correction. This circuit was designed in response to the universal complaints that digital sound was harsh with complex overtones on instruments like multiple horns, multiple violins, massed voices, etc. It is even more harsh when those multiple instruments or voices get louder as the artists express themselves. It was discovered that virtually all factory made discs and recordings we could find violated industry standards resulting in digital “clipping” and the resulting harshness. MSB designed the Intersample harshness correction as a no-compromise correcting process that brings the bit values into the correct signal level freeing up the top end of the signal and associated harmonics to be accurately converted and further improve dynamic range.
The next process is the digital filter necessary to remove artifacts above the audio range that are not related to the analog signal. The implementation of this filter is critical and there are none available off the shelf that come close to MSB’s requirements. Digital filters are written in house and installed on a SHARC DSP chip big enough to contain at least 4 of MSB’s proprietary filters. These are the fastest DSPs available and run a single-stage 80-bit fixed point FIR Filter, resulting in very fine resolution that MSB’s DAC modules can take full advantage of.
So what is a digital filter all about? Because of the very large scale SHARC chips, the digital filter can be written to a much higher resolution (like 32 X) performing a function similar to an upsampler. When done correctly with high accuracy this high resolution digital filter can make an extraordinary improvement in the sound. It is important to understand that the digital bits are nothing more than samples that were taken every so often from the original recording waveform. How frequent these samples were taken determine the original resolution so if they were taken at 44.1 thousand times per second (44.1 kHz), we don’t actually know exactly what the waveform looked like between the samples. This is also true of higher resolutions. This is the one place in the digital conversion where art enters in, and that is in digital filter design. Lets look at the process. We start out with a continuous analog stream at the studio. Every so often we record a single voltage point on the analog wave form. These points enter our DAC and our job is to figure out what was between the points.
So let’s take a simple example. Here are three dots. . . . Now they can be connected a lot of different ways, with a straight line, with a square wave, or with a sign wave. Each will sound very different but all are technically correct possible interpretations of what might have been present between the original data points. This is what a digital filter does. It tries to make an intelligent guess what the original music looked like. One of the digital filter techniques looks way back in the past, and way forward in the future, and uses this data, plus what we know about the nature of music to make a better prediction about what was between the data points. This is where the art and many years of knowledge and experience come into play making the best digital filters possible. The “algorithms” are the predictive math that is based on what came before the sample we are looking at and what came after the sample we are looking at. The algorithms are designed to go back and go forward thousands of samples just to decide what was most likely in between the two original bits we are trying to fill in at that particular time. It is easy to see why such large extensive high-speed processing is needed. We fine tune these algorithms by testing, and then listening, to make a filter that guesses better. We have tried hundred, if not thousands before arriving at our current filter suite.
The whole rest of the audio chain just applies good engineering to be as accurate and noise free as possible. But in this case we are artists, trying to paint the best interpretation of the same data points all our competitors are also looking at. That’s what makes our product sound different and why we offer so many filters and upsamplers. None are perfectly right or wrong, they are just different interpretations of the same data, that may match the studio conversion better for one recording than another. This very high-density data is easily resolved and converted by MSB’s Platinum, Signature or Diamond DAC modules since they are capable of converting at more than 5 mHz.
You will remember from the earlier discussion about the original Philips ladder DACs that the bits have a voltage and that voltage blends into the resistor ladder to make the analog signal, the final continuous voltage that is now the analog waveform (music). A happy result of the MSB ladder DAC design is the output voltage of the DAC is in excess of what is needed to drive an amplifier. Therefore all that is required to achieve a very high performance volume control is a stepped resistor attenuator. After the stepped attenuator there is a single obsessively designed output buffer to be sure the impedance at the output jacks stays very low. The result is that this DAC is able to drive any amplifier no matter how difficult.
The Sound
The performance of DAC IV is amazing. Immediately you’ll notice the beauty of the voices and the lack of congestion and harshness that often accompanies massed voices and massed instruments with complex harmonics like violins and saxophones. This new level of clarity is also noticeable in instruments of all frequencies. This DAC also dramatically increases the resolution of the performance. Fine detail that was once hidden and “smeared” in the background is now clear and obvious. There is a perception that each fine harmonic is individually revealed giving new definition and separation to the instruments. The sense of space where the recording was done is intricately revealed with wider and deeper soundstage. Each instrument is more singly placed and positioned. Our engineering and listening team have always noticed that when genuine fine detail is further revealed (without harshness), there always seems to be an increase in the sense of space. This makes sense because spatial cues are the original sound of the voice or instrument with the reflections from the walls of the recording space added, and are therefore complex harmonics themselves.
The dynamics are significantly improved giving the drummer’s expression a convincing “hit” that cannot be ignored. Bass attack is convincing and scaled to what the artist intended. This new level of dynamic reproduction is a critical ingredient in convincing us we are listening to the live performance, and drawing us deeply into the music. Once it is heard, it is hard to live without.
One theory of what is going on with system matching
Now that we have some experience with the sound quality of the DAC IV including comments from around the world, we feel that the DAC IV achieves a new paradigm in sound performance. We also have realized it has had a profound effect on system matching. It is our opinion that for over 25 years now, audiophiles (ourselves included), have been tuning their systems to reduce the harshness and grain that they hear in their systems. Often in addition to the digital front end, the preamplifier, amplifier, cables, or wall power are blamed. While any of these components can show an improvement and can have better performance on their own, we also suspect that certain components can sonically “slow” the signal making digital more tolerable with decreased harshness and increased clarity. Because digital is math, and crimes (errors) in that math can have incredible speeds and rise times these can be transferred to the analog signal. These errors may not pass through certain components downstream like colored tube or solid state equipment, certain cable designs, or even some speaker crossovers with their necessary inductance and capacitance. So if digital has been harsh all these years and certain components “slow” the signal and have a tendency to subdue the really fast and sharp rise times and let the more rounded sine waves through, maybe this sounds more like the original analog wave form and the natural sound of the instruments. If this is true then it is easy to see how complex it would be to correlate a good sounding result with more accurate components vs colored components.
So if previous digital has been harsh and we have been tuning our systems for it, and the DAC IV is really a new level of fine harmonic detail lacking harshness, then the front end may well be the most important component in the system. As audiophiles we have found that the introduction of the DAC IV in an otherwise known system changes the game. As an example, certain components like a well loved preamp that was once a stable and treasured part of the system can be found to be in the way of the sound. We have also noticed that changes downstream are now laid clear for what they are – good or bad. There is little or no “waffling” about how valuable a change is like: Did the cymbal have a little more “air” with this amplifier vs that amplifier, or cable, or speaker, or wall power treatment? Not only is it clear that a change downstream is good or bad but it is also clear why. The soundstage may be better, or better pace and rhythm, or natural female vocals, etc. As audiophiles we hate to think about all the sideways moves we made over the years that were expensive and time consuming. We are thrilled to know that a better source like the DAC IV is true to the music, making system matching an easier and more efficient process.
We invite the user to peel back the layers of correction and listen to the DAC IV with internal volume control in the simplest of configurations. With a good bit-perfect source, and fast and accurate amplifier and your favorite speakers. We are confident you will hear a new level of realism never before experienced with your system.



link to page:http://www.msbtechnology.com/faq/why-ladder-dacs/



 楼主| 发表于 2018-2-22 00:11 | 显示全部楼层 来自 北京市
本帖最后由 andygaof 于 2018-2-23 00:18 编辑

占一楼,做个翻译,同时发表一下自己的看法

第一点(
Digital since the mid 80’s)
,MSB在讨论Delta Sigma DAC时认为Delta Sigma 的主音解析非常精确,但是在回放低音量的信号时由于模拟滤波器的能力问题导致了一定的相移,所以在回放乐器的自身谐波这些小音量信号时不准确。这对多乐器的声音影响尤其大。

老高观点->这一段讨论准确的说明了使用90x8/449x芯片的解码器,其中一个核心是模拟滤波器的性能,模拟滤波器不好的,弱音表现力会很差,更差的模拟滤波器连大音量都有些乱。所以不同厂家的使用相同解码器的设备,但是模拟滤波器是自己做的,声音水准根据滤波器的能力差异会是巨大的。以老高本人的听音经验,很多delta sigma 解码器存在的问题是缺少弱音表现力,但是主音表现很不错。即使是很便宜的解码器在表达简单音乐时都是如此。这对于很多音乐类型,尤其是电子合成摇滚音乐这种本身就没有细节的音乐类型,低端解码器甚至会特别给力,缺少弱音,使得声音更加干净,通透,速度也显得很快。尤其是现代录音,其录音音量很高(专业术语叫采样电平),这时差一点的设备听感也还是过得去的。但是对于高水准的乐器尤其是多件乐器录音,一方面,这些乐器动态大,乐手演奏时小音量跟大音量的差距巨大,多乐器时就更加厉害。以克来伯指挥的贝多芬第五为例,弱音几乎不可闻,强奏则排山倒海。穆特卡门幻想曲第一乐章,2:40多秒的高音弱奏,细若游丝却细节丰富。这时候这些低端解码器就非常困难了。这些解码器会因为抹去了弱音细节,而使得声音干,硬,刺激。但是这不代表Delta Sigma 解码器都差,只要解决了模拟滤波器的性能问题,声音是可以不错的。但是模拟滤波器都不便宜,所以指望便宜好声,这基本是不现实的。

第二点(MSB Sign Magnitude R2R DAC):MSB说他的R2R是全新设计的,使用了Sign Magnitude技术,原理是使用音乐的中点作为一个电压值,然后音量大的加电压,音量小的减电压。这样可以很好的还原弱音细节。


老高观点->厂家的话听听就好,用5毛钱就能出十万的声音那才是牛逼。只要声音好还比别人便宜那就是好产品了。无所谓用什么技术。



第三点(Sample Rates):MSB说他的DAC最高可以处理5MHz多bit的音频,CD录音是44.1KHz,现代录音棚可以做到192KHz,有些已经达到了384KHz,所以MSB认为他可以又巨牛逼的未来兼容性,同时他又在黑delta sigma是1bit的,说MSB的DAC比delta sigma的弱音好特别特别多



老高观点->在delta sigma巨大的成本优势面前,R2R面临巨大的风险,一旦解决了滤波器的问题,R2R势必因为成本过高快速衰落。目前同级别下R2R声音好于Delta Sigma基本上是事实,但是未来未必是事实。一切只能看结果。



第四点:(MSB Input Processing):MSB说没有好的、精确的数字过程,不管多牛逼的解码都是白费。然后他说通过专利技术他们把jitter降低到了7ps,在某些条件下降低到了2ps。然后说没啥玩意可以测量我的jitter了,所以又做了个ADC,还计划出一个测试仪。(从字面理解丫还没测试仪,那7ps怎么测出来的?)接着MSB说他有个什么巴拉巴拉的专利,可以在多乐器的情况下又更加精准的表现,单个艺术家表现的时候,某个乐器更加响可以很轻松的听到,然后进一步增加了动态。然后MSB开始发飙了,公开挑战了奈奎斯特定理和香农定理。说在44.1KHz甚至更高采样率下,我们不知道sample点之间的具体波形是什么样的。所以FIR filter,数字滤波器就是DAC厂家的艺术表现力之所在了。然后巴拉巴拉说了一堆他们用了什么硬件,怎么搞数字滤波器的。然后说他们很牛逼的可以很接近真实的还原。然后又说他们是真正的艺术家,提供了很多的滤波方式,每一种都各有风格,都好听。然后说没有那个选择是绝对的对与错,一切看口味。最后说他们用了什么技术保持输出阻抗如何如何低,如何如何适用所有的放大器。



老高观点->先科普一下采样率的滤波,然后在说明我的观点(技术帝请绕路,为了让大部分同学看懂,里面有些小误差)



幻灯片1.JPG

原始的波形被采样后就留下了这些点存储下来

幻灯片2.JPG

这些存储的点变成01000101001........可能在CD里,也可能在电脑文件里

幻灯片3.JPG

解码器如果电平值错误带来的危害

幻灯片4.JPG

解码器如果时钟精度错误带来的危害

幻灯片5.JPG

幻灯片6.JPG

不同滤波算法对最终的声音产生的影响,下面的比较极端,没人真的这么写滤波算法,因为这个声音肯定巨难听,但是大概说明的滤波的作用。

所以解码过程三个核心关键点,两个精确一个牛逼,时钟精确,电压幅值精确,滤波算法牛逼。然后我们就尴尬了,因为这三个指标,我们当前的测试技术,一个都测试不出来。现在的设备能测出来的时候这个机器就很难听了,尤其是数字滤波算法,几乎完全无法测试。









未完待续

回复

使用道具 举报

发表于 2018-2-22 00:32 | 显示全部楼层 来自 北京市
干货 硬货
回复

使用道具 举报

发表于 2018-2-22 00:39 | 显示全部楼层 来自 挪威
太贵了。有没有便宜点的Ladder DAC?
回复

使用道具 举报

发表于 2018-2-22 00:44 | 显示全部楼层 来自 美国
Holypal 发表于 2018-2-22 00:39
太贵了。有没有便宜点的Ladder DAC?

R2R解码很多啊,Aqua, Metrum 等。国产的泉,终结者等也是。
回复

使用道具 举报

发表于 2018-2-22 00:55 | 显示全部楼层 来自 挪威
tom133 发表于 2018-2-22 00:44
R2R解码很多啊,Aqua, Metrum 等。国产的泉,终结者等也是。

泉和终结者对比如何?
回复

使用道具 举报

发表于 2018-2-22 01:28 | 显示全部楼层 来自 美国
Holypal 发表于 2018-2-22 00:55
泉和终结者对比如何?

不清楚。6moons对终结者评价很高。泉价格较低,用户很多,和船界面的组合很受欢迎。
回复

使用道具 举报

发表于 2018-2-22 09:34 | 显示全部楼层 来自 山东省青岛市
MSB好像刚出了2款入门解码,最低端的也要1W美元
回复

使用道具 举报

发表于 2018-2-22 10:03 | 显示全部楼层 来自 安徽省亳州市
MSB的这几篇东西也没讲什么太有用的,因为R-2R要面对的主要问题,不是电阻精度,而是交越Glitch,对付这个东西才是最大的难点,不过眼下也有很多小公司用各种办法解决了。

还是那句话,要讲清楚这个玩意就要讲理论,画图,论坛猪友们没有那么高的智商,所以也不爱看,看不懂,说了也白说。
回复

使用道具 举报

发表于 2018-2-22 10:05 | 显示全部楼层 来自 四川省成都市
wansien 发表于 2018-2-22 10:03
MSB的这几篇东西也没讲什么太有用的,因为R-2R要面对的主要问题,不是电阻精度,而是交越Glitch,对付这个 ...

瞎说什么大实话
回复

使用道具 举报

发表于 2018-2-22 10:37 | 显示全部楼层 来自 安徽省亳州市
snova 发表于 2018-2-22 10:05
瞎说什么大实话

明摆着吗,一开始贴图算数给他讲,就“我不听我不听我不听,你骗我你骗我你骗我”,一如琼瑶剧里的弱智女主角。
回复

使用道具 举报

发表于 2018-2-22 10:56 | 显示全部楼层 来自 四川省凉山州西昌市
额。。。
回复

使用道具 举报

发表于 2018-2-22 14:45 | 显示全部楼层 来自 四川省成都市都江堰市
学习 了解一下
回复

使用道具 举报

发表于 2018-2-22 15:47 | 显示全部楼层 来自 新疆乌鲁木齐市
本帖最后由 okcomputer 于 2018-2-22 15:55 编辑

我听过的akm细节是要少一点,不过比听过的9018那台细节好,9018主干强,细小信息少的比较多,听过的cs4398细节更丰富点

补充一点,9018那台模拟输入放大后也是感觉细节少,所以到也许不能怪9018解码的错
回复

使用道具 举报

 楼主| 发表于 2018-2-22 17:50 | 显示全部楼层 来自 北京市
Fermat 发表于 2018-2-22 15:33
linearX不好卖吗,这么闲

吴总黑我无极限啊,托您的福,LinearX卖的还可以。正式上市三个月,20多台,远远没有您的A100销量高
回复

使用道具 举报

 楼主| 发表于 2018-2-22 23:58 | 显示全部楼层 来自 北京市
二楼今天更新到此结束,未完待续
回复

使用道具 举报

发表于 2018-2-23 00:04 | 显示全部楼层 来自 湖北省武汉市
高总辛苦。。
回复

使用道具 举报

发表于 2018-2-23 00:12 | 显示全部楼层 来自 广东省江门市
用啥吹啥定理?
回复

使用道具 举报

发表于 2018-2-23 08:34 | 显示全部楼层 来自 山东省
耳机俱乐部有方波帝10多页的高楼谈R2R,方波帝被封了就看不到了。

高总既然主打PC Hifi为什么不试试传说中打total six的T+A DSD8 1bit true DSD配合HQ转DSD512?
回复

使用道具 举报

发表于 2018-2-23 09:10 | 显示全部楼层 来自 浙江省舟山市
“还计划出一个测试仪。(从字面理解丫还没测试仪,那7ps怎么测出来的?)”

MSB很长一段时间直接用Lynx two声卡测量数据,早就在明眼人那儿沦为笑话了  卖 傻 B诚不我欺
回复

使用道具 举报

您需要登录后才可以回帖 登录 | -注册-

本版积分规则

Archiver|手机版|粤icp备09046054号|耳机网-耳机大家坛

粤公网安备 44030602000598号 耳机大家坛、www.erji.net、网站LOGO图形均为注册商标

GMT+8, 2024-3-29 17:13

Powered by Discuz! X3.2

© 2001-2013 Comsenz Inc.

快速回复 返回顶部 返回列表