USB / reclocker / regenerator / DDC (digital to digital convertor)

amit11

Well-Known Member
Joined
Jul 7, 2014
Messages
606
Points
63
Location
Pune
Hi Friends,
Over the few months I had tried to experiment with USB sound. In some cases it was bright and in some cases it was fine. I also did USB to SPDIF and tried.
Somehow I still preferred my original chain of SPDIF. Though USB had some good points, it was still missing the natural & lush mid-range of my spdif.
I see many FMs have tried by changing the source, adding reclockers, regenerator, DDC (digital to digital convertor) etc in the chain and had better experience. What I understand is if
we have a normal USB chain in our system (without reclocker, regenerator etc), it does not sound good, the reasons being mentioned as jitter due to packet loss, changes in packet timings etc which can be caused by electrical interference, impure power supply etc. I somehow am not able to grasp it. The reason is USB has become a matured technology over these years and the creators of USB standards and the electrical and elecronics technology around it, would have definitely considered it. The chips are so fast today that packet timings cannot be an issue. I also don't think that in USB packets are getting lost. I may be wrong but i think when we try to use any DDC (digital to digital convertor) or anything that purifies USB and we find that it sounds better, the reason for that might not be that it is reducing jitter and improving the packets timing etc. Just like jriver, foobar and many music players have their own sound signature even though each claims to be bit-perfect, similarly a DDC interceps the signal and via some sonic algorithms alters the final digital data, which modifies the sound and it might sound more different and better to us.
I know my understanding may not be correct since I have not tried these higher end things in my chain. However I would be happy to 'hear' if someone can post sound from their system which is BEFORE adding these reclockers/regenrators/DDC vs AFTER. Thanks


Amit.
 
Firstly my apology for not having any experience with USB audio, my amplifier has built in streamer and my other source is my CD transport. So, I am not qualified to attempt a reply based on experience but since I have a little experience in designing digital circuits, I couldn't resist a post.

Let's get a little deeper into the digital world. We will visit the USB interface later after taking a round of what's electrically going on inside a computer.

Almost all complex switching circuit based device has a dark area. The power consumption. It has been a highly debated subject for decades. TDP or Thermal Design Power, is one of the measurements used as an indicator to a digital circuits power requirement under "realistic" load. Mark the word "realistic" here, it's not "actual". The primary concern about power supply in a digital circuit is the moment the number of '1's increase in the operated data, the power requirement shoots up in a few pico seconds. This is coupled by number of '1's in the address bus and other peripheral connection like USB. The moment, a switching circuit demands sudden jump in power, and if the power supply isn't capable of supplying the required instantaneous power, it results in slow rising and falling edge of the 'bits', which can be seen in an "Eye Diagram". If a "bit" doesn't reach the minimum threshold voltage to be considered as '1' within it's clock cycle, then a data corruption occurs, which may force the system to retransmit and lose it's temporal resolution (jitter).

Armed with the above, it's easy to understand that a high speed serial transmission protocol like USB is very much prone to data retransmission due to, inadequately designed power supply, CPU load at that moment, shift or ocsliation in ground plane, conducted interference etc.

Although USB is designed to handle retransmission at the controller level but it has really just a small buffer to cache the travelling data. The cache falls short when retransmission happens frequently. The retransmission attemp timing interval rate is almost like Fibonacci series and is limited to 11 attempts. Once the statistical monitoring of the transmission performance is found below satisfactory threshold, the transmission speed is reduced to make the connection more reliable. Thus the throughput of USB isn't constant, infact it can momentarily change when another electrical load is switched on nearby, like an AC or Microwave, due to momentary drop in power supply.

All of the above factors combined can affect the performance of the USB interface a lot. The relockers receive and cache the "bits" then retransmits through a galvanically isolated circuit. The reclockers can look slightly below the USB specification based threshold voltage to "catch" the near miss '1's thus reducing retransmission. Reclockers also use much better quality clock to ensure even less jitter. That said a poorly designed reclocker can be catastrophic.

Hope I could make some points for your further thoughts.
 
Normal electronics of today's age like computers, laptops, mobiles etc, are pretty well designed keeping all these in mind. If we consider the normal circuitry of such things and the internal data flowing (non audio data, from processor to other chips etc), then a normal computer would go into error soon considering so much electronics is fitted in such compact volume. The RAM, internal data transfers which happen at even higher speeds than usb and are even far more sensitive, would get error prone in that case. But it hardly happens, then what makes usb different and more sensitive than the internal electronics which carries far larger and speedier amounts of data. This is what i fail to understand. Galvanic isolation and DDC are the new generation hype of marketing giving a promise of bit perfect output, which in fact modify the digital data at the hardware level instead of software level.

If we hear clicks or pops occassionally, then in that case i can understand that the particular usb has some transmission issue. But if no such clicks or pops then yes all data is arriving correctly. ( there might be some occasional timing differences of the packets bursts).
 
Last edited:
What I understand is that it won't be correct to say that internal electronics doesn't matter, everything matters, but we can't modify that and test. we can try different external cables, try reclockers, isolators, regenerators and they surely make improvements in one way or another. In digital audio transmission, correct timing is most important so that receiver receives the bits accurately and noise also can be an important component that affects this and needs to be kept out, that's where isolators help. Its not just about speed of transmission, its the timing of the bits that's so much important and usb is not an easy interface from which to extract audio bits with accurate timing. Below link has some interesting discussion.

https://www.psaudio.com/pauls-posts/is-it-possible/
 
Normal electronics of today's age like computers, laptops, mobiles etc, are pretty well designed keeping all these in mind. If we consider the normal circuitry of such things and the internal data flowing (non audio data, from processor to other chips etc), then a normal computer would go into error soon considering so much electronics is fitted in such compact volume. The RAM, internal data transfers which happen at even higher speeds than usb and are even far more sensitive, would get error prone in that case. But it hardly happens, then what makes usb different and more sensitive than the internal electronics which carries far larger and speedier amounts of data. This is what i fail to understand. Galvanic isolation and DDC are the new generation hype of marketing giving a promise of bit perfect output, which in fact modify the digital data at the hardware level instead of software level.

If we hear clicks or pops occassionally, then in that case i can understand that the particular usb has some transmission issue. But if no such clicks or pops then yes all data is arriving correctly. ( there might be some occasional timing differences of the packets bursts).

With regard to the internal communications, a few things are done which are not possible for external communications. For example, internal signalling is done at much lower voltage considering the extremely small distance of a maximum of few centimetres. There are elaborate CPU sections dedicated to monitor the TDP to keep the whole system in sync. Moreover, most of the signalling is in differential mode. Finally all internal communications are synchronised by the same master clock inside and the whole system has the same ground reference.

Galvanic isolation is actually a necessity, the electronic subsystem of one component is bound to have a slightly different ground plane that other connected component, with day and night temperature variations the ground plane can swing a few Mili volts that can affect the data transmission between equipments. Galvanic isolation is nothing but a miniature transformer in the Tx&Rx line which does two things, it breaks the direct electrical connection and it creates differential signal from ground referenced signal. At the recieving end its easy to do a common mode noise rejection to recover the differential data with much less error caused by any external interference.

Finally, since USB signalling is done around 2.4 GHz which makes a single a bit to travel around 1.5 meter before the next bit arrives on the same wire, thus a standing wave is can be avoided by keeping the USB cable length below 5 ft or 1.5 mt.

A generic point to remember is any cable is susceptible to interference. The higher the speed of communication the faster the switching, the lesser the voltage and subsequently much more prone to EMI and RFI. Try copying data from an external hard drive while the USB cable and any power cable run parallel and next to each other. Perform the same without the power cable around, check for transfer speed differences.

As we all know but few accept that almost everything in audio is hyped up and made to look like the audio companies are the flag bearer of science and technology. For example, for decades galvanic isolation is mandatory in communication between medical equipments i.e. between a CT or MRI Scanner and a computer. But when it comes to audio it is projected as if it's just invented for us audiofools. Split phase power is mandatory in UK for over half a century for industrial machinery where a human operates, to reduce the effects of electrical shock in an eventuality. This same thing is reinvented by audio companies and branded as Balanced Power Supply.

One worthwhile reading would be to go through the failure of the first Trans atlantic telegraph line. The dots and dashes were all mixed up. At that time the long distance effects of standing wave, cable impedance, velocity of propagation etc weren't taken into consideration. As frequency rises the cable resistance rises and the only available path for conduction remains the skin of the conductor. The impurity at the surface matters most at high frequency transmission. Incase a USB cable has been manufactured properly then its characteristic impedance can be different from the protocol defined impedance and hence slow down the communication speed.


Best option always is to measure and proceed. The most affordable measurement is through our ears.
 
Last edited:
HDMI which transmits audio and video data, data amount which requires much higher speed than simply audio data over usb; that works fine for most of the home applications. E.g. tata sky set top box to tv. That must be even more suspectible to noise but works fine. We rarely see or hear the noise in our tv even when it transmits 4k high definition data.

Secondly some usb cables come with a lump of ferrite choke for negating noise... that can also take care.

If hdmi works fine, then why usb is so much hyped.
 
Hello,

Consider this as a discussion rather than knowledge sharing as I am not an expert in digital electronics.

If I broadly look at the problem on hand, I see two groups:
Group 1: 3.5mm headphone jack, RCA, SPDIF (optical and co-axial) and even HDMI
Group 2: USB

The first group was meant primarily for audio signal transport but can be used for data
However, USB was meant for data transport but can be used or audio.
To me THIS is a BIG difference!

Basics from communication theory says that low frequency audio signals (not bass) can easily get lost or corrupted when passing through a medium that contains high frequencies.
To prevent this, we need to modulate audio using a high frequency carrier.

I'll start with the worst in the first group: HDMI (@amit11 posted post # 6 as I was typing this)

HDMI has two highways, within the 19 pins, for audio transport.
The first where audio is modulated (mixed with high frequency carrier and put in an envelope so that it does not get lost) and sent along with video signal. At the receiving end, modulated audio is separated from video, de-modulated and passed to audio sections.
The second is audio return channel (ARC) which enjoys two dedicated lines that carry differential digital audio that easily converts to SPDIF.
The differential nature helps in rejecting any common mode noise it picks up along the way.

Here, point to note is that audio is transformed from its basic form into a modulated signal for transportation along with the much higher frequency video. So it is easily possible to retrieve it from other high frequency signals and noise, at the receiving end, and de-modulate it to its original form.
The same concept used by radio stations where the modulation carrier frequency (name of the envelope) determines which station the modulated signal belongs to.

Now lets shift focus to Group 2: This is an area where I have little to no knowledge
There is audio and there is the USB clock. So there is a need for modulation at the transmitting end.
Does USB modulate audio before putting it on the cable?
Is re-clocking a means to shift audio into a higher frequency spectrum or what ever and can it be considered as a robust modulation method?

Whatever modulation technique is used before putting audio on USB cable, it needs to have a fixed standard so that de-modulation can be easy and exact.
Else, USB audio will always work well with a software media player that will modulate the audio and ITS dedicated hardware that will demodulate it at the other end of the PC.

Lastly, I would like to add one thought: Keep the number of audio conversions from one form (format) to another to a bere minimum for best signal integrity, else you will always be doing the job of chasing to clean it back to its original form and shape.

Regards,

Ravindra.
 
HDMI which transmits audio and video data, data amount which requires much higher speed than simply audio data over usb; that works fine for most of the home applications. E.g. tata sky set top box to tv. That must be even more suspectible to noise but works fine. We rarely see or hear the noise in our tv even when it transmits 4k high definition data.

Secondly some usb cables come with a lump of ferrite choke for negating noise... that can also take care.

If hdmi works fine, then why usb is so much hyped.

There are few points to consider before getting the full picture.

First, every modern TV has enough image processing algorithms to reconstruct and denoise each frame or to even handle missing frames.

Second, our visual hysterisys allows us to distinctly register about 24 frames per second only that is 41.6 mili second, whereas auditory hysterisys is about 11 micro second.

Third, we have the ability distinctly indentify only 256 shade gradation of a single colour.

Fourth, our visual processing is in spatial domain unlike our hearing which is in frequency domain.

Fifth, our ability to identify a single pixel of a TV is about the same distance as the diagonal size of the display.

There are definite losses in HDMI transmission, that's why HDMI has limitation about the distance of HDMI cabling. This is where optical HDMI steps in.

If you put forth all of the above points to visualise the combined effect of a missing bit in HDMI, it would be quite easy to see why audio has become so highly debated industry rather than the video. Our auditory system is much more evolved than our visual system.
 
Hello,

Consider this as a discussion rather than knowledge sharing as I am not an expert in digital electronics.

If I broadly look at the problem on hand, I see two groups:
Group 1: 3.5mm headphone jack, RCA, SPDIF (optical and co-axial) and even HDMI
Group 2: USB

The first group was meant primarily for audio signal transport but can be used for data
However, USB was meant for data transport but can be used or audio.
To me THIS is a BIG difference!

Basics from communication theory says that low frequency audio signals (not bass) can easily get lost or corrupted when passing through a medium that contains high frequencies.
To prevent this, we need to modulate audio using a high frequency carrier.

I'll start with the worst in the first group: HDMI (@amit11 posted post # 6 as I was typing this)

HDMI has two highways, within the 19 pins, for audio transport.
The first where audio is modulated (mixed with high frequency carrier and put in an envelope so that it does not get lost) and sent along with video signal. At the receiving end, modulated audio is separated from video, de-modulated and passed to audio sections.
The second is audio return channel (ARC) which enjoys two dedicated lines that carry differential digital audio that easily converts to SPDIF.
The differential nature helps in rejecting any common mode noise it picks up along the way.

Here, point to note is that audio is transformed from its basic form into a modulated signal for transportation along with the much higher frequency video. So it is easily possible to retrieve it from other high frequency signals and noise, at the receiving end, and de-modulate it to its original form.
The same concept used by radio stations where the modulation carrier frequency (name of the envelope) determines which station the modulated signal belongs to.

Now lets shift focus to Group 2: This is an area where I have little to no knowledge
There is audio and there is the USB clock. So there is a need for modulation at the transmitting end.
Does USB modulate audio before putting it on the cable?
Is re-clocking a means to shift audio into a higher frequency spectrum or what ever and can it be considered as a robust modulation method?

Whatever modulation technique is used before putting audio on USB cable, it needs to have a fixed standard so that de-modulation can be easy and exact.
Else, USB audio will always work well with a software media player that will modulate the audio and ITS dedicated hardware that will demodulate it at the other end of the PC.

Lastly, I would like to add one thought: Keep the number of audio conversions from one form (format) to another to a bere minimum for best signal integrity, else you will always be doing the job of chasing to clean it back to its original form and shape.

Regards,

Ravindra.


Ravindra, data passing through USB is indistinguishable, be it audio or data, it goes through the same sequence of operations. Line level voltage modulation (effectively amplitude modulation) and applicable multiplexing as per relevant USB standard are the only things applied. Multiplexing is needed to address the Plug n Play protocol implementation to support multiple USB endpoints connected to a single hub.
 
Ravindra, data passing through USB is indistinguishable, be it audio or data, it goes through the same sequence of operations. Line level voltage modulation (effectively amplitude modulation) and applicable multiplexing as per relevant USB standard are the only things applied. Multiplexing is needed to address the Plug n Play protocol implementation to support multiple USB endpoints connected to a single hub.

Then, I question the robustness of this digital - to - digital conversion/ modulation process, both, its reliability as well as its reputability and could well result in hit or miss as far as quality is concerned.

Multiplexing again means scanning and settling times and that will not be consistent with simultaneous data transport as it may have some priorities accociated with it.

Regards,

Ravindra.
 
Then, I question the robustness of this digital - to - digital conversion/ modulation process, both, its reliability as well as its reputability and could well result in hit or miss as far as quality is concerned.

Multiplexing again means scanning and settling times and that will not be consistent with simultaneous data transport as it may have some priorities accociated with it.

Regards,

Ravindra.

Absolutely correct. Its more convenient than robust in terms of constantly sustainable throughput. This is where things move into a different world of Asynchronous Transfer Mode, Frame Relay etc. where Permanent Virtual Connection ensures minimum latency.
 
Absolutely correct. Its more convenient than robust in terms of constantly sustainable throughput. This is where things move into a different world of Asynchronous Transfer Mode, Frame Relay etc. where Permanent Virtual Connection ensures minimum latency.

I feel like I reinvented the wheel after you mentioned asynchronous transfer.

But now we are back to square one unless the solution is to ensure use of USB gear that has asynchronous transfer protocol implemented. Unless there is more to it.....

Regards,

Ravindra.
 
Hi @Otpidus

Excellent explanation given by you

I have also read on many forums that these reclockers and regenerators of USB do bad rather then good as far as muisc is concerned and when usb transfer is asynchronous it relies on clock of dac rather then source

Correct me if i am wrong on the explanation of asynchronous because I am also confused with that

So now why do I get improvement in sound when I use a USB regenerator even if my dac supports usb asynchronous transfer ?

It's all confusing for me

Is it because since it's regenerating the usb signal the noise associated with power from source equipment gets discarded ?

I agree that even usb regenerators are in general used with a switching power supply and that too a cheap one which is known to induce noise in power
But is that still better then noise generated by say a pc with all its various parts and switching volatge regulators?
The usb regen like uptone has lienar regulators

So if that's true our main aim should be to remove power induced noise from usb stream ? And that will improve sound quality?

I am really confused about all this since many years and nor getting a sure shot answer

I even asked these same questions to a guy working for DCS when he wrote the same as you about usb reclockers and regenerators..but he has not replied back to me
Maybe he has not seen my message or not allowed to answer

So if you can help me it will be great
 
Asynchronous USB is better in terms of jitter and reduced retransmission but maintaining a steady throughput still relies on the data source. Asynchronous mode benefits from various factors including usage of Fixed Frequency Clock instead of PLL etc. A well implemented regenerator relies on a few things, a good master clock, a buffer and galvanic isolation. A galvanic isolation alone can put a stop to all unwanted switching noises that travels through the USB cable. If any dac has implemented a well isolated USB, has Asynchronous Transfer, has a good quality Master Clock then it already has a regenator built-in. I dont forsee any benefit of an additional regen then. This would then boil down to the source's CPU load and congestion at the North & South Bridge to cater to the demands of the Dac in real time.

Sometime back, I read that a new USB class implementation is going to be proposed, which will make a local copy of the entire file and play from local temporary storage. It's almost similar to what happens in DLNA, where the chunk is so big that a normal Redbook track is completely transferred to the renderer before it starts playing.
 
Last edited:
Hi
To isolate the pc from dac I and many others use a NAA like maybe a raspberry pi, Allo usbridge and the SOTM and sonore which are really high end with well implemented deisgns

I completely agree with you that if a dac itself has a well executed usb section with all aspects, a regen and reclocker will show least impact

So that does mean that with mid range dacs which obviously can't have very well executed usb section due to cost factor can still benifit from a reclockers / regens

But then all becomes relative with what is the rest of the chain

But I will also like to add that a very good source does improve the quality of sound or you can say it makes a quiet a big impact on overall sound of a setup..

That's why if budget permits a good streamer like SOTM is the way to go

The new USB implementation looks very interesting

The roon guys also claim that their RAAT protocol is very different from DLNA as DLNA was made ages ago and they felt a new protocol is very much needed to transfer audio in a network streaming kind of set-up
 
Last edited:
The concept of mid range priced equipment is so skewed in Audio industry that it really needs government regulation to control the pricing.

I have only heard of the word "Roon", know nothing about it. So, I can't comment anything about it Rikhav.
 
@Otpidus :

By Asynchronous transfer we mean we need to use the DAC clock with an ASRC block, right?
The USB signal will still need to be converted to I2S by the USB protocol controller and these I2S lines will interface to the DAC I2S lines either directly or through the ASRC block.

Would it be correct to say that the discussion so far has only been to explain the USB transport.
However, we have not been able to identify and area or a step that will explain how USB audio will be superior.
e.g. If I have SPDIF or assume if my computer motherboard has 32bits/ 384KHz DAC in-built, the same file played played will not sound any better on USB. At best, it will sound the same!

In fact because of at least two additional conversions the USB transport is having to use, the audio will always be inferior in terms of re-constructed audio (signal integrity).

What do you think?

Regards,

Ravindra.

P.S. @Otpidus : I can't thank you enough for your explanations that has brought clarity to the topic of USB audio. The signal chain is now clear to me which otherwise was confusing because of the commercial terms like re-clockers, re-generators etc. Good job Sir, and thank you all over again!!

@amit11 : Thanks for the thread as it has help me to gain knowledge in an area I otherwise would not have trodden into.

For the DIY minded members, this is a link to the USB to I2S audio IC:
https://www.silabs.com/products/interface/digital-audio-bridges/usb-audio-bridges/device.cp2615
http://www.ti.com/product/pcm2704c/description

It supports 24bit/48KHz audio.

XMOS has platforms that support 32bit/384KHz, but need licensing. Here is the link: https://www.xmos.com/support/boards?version=6.5.1rc7.a&product=17498&component=14442&page=0
 
Last edited:
Right Ravindra, Asynchronous USB is "pull" type transfer. The USB endpoint will specify when and how much to send and the source complies. Rest everything remains same, except that data is flagged by the source. To and fro I2S conversion is unavoidable.

A direct I2S transfer would be beneficial in many ways but it's not meant for transferring data to more than a few centimetres, so error correction isn't a strength of I2S. Don't you think I2S is very similar to a low voltage AES/EBU?

@Ravindra Desai, Thank you for your appreciation Sir, but my knowledge is extremely limited and shallow. Moreover this isn't exactly my area of expertise. I just tried to put in words some of my limited understanding.
 
Last edited:
@Otpidus : you are very modest.
As you rightly said, I2S is for short distance communication.
AES/ EBU has two versions: Professional & Commercial. The pro version supports higher voltage levels, but most importantly, both support differential/ balanced mode of signalling making it robust in noisy environments.

Regards,

Ravindra.
 
Exactly thats what I meant Ravindra, Differential I2S and AES/EBU are fundamentally similar, major difference being the voltage level. A common mode noise rejection would be genuinely beneficial with higher voltage levels where signal to noise ratio is clearly demarcated, where as in Differential I2S we wouldn't have that advantage, making it more error prone in longer distances.
 
Last edited:
Purchase the Audiolab 6000A Integrated Amplifier at a special offer price.
Back
Top