DIYHiFi.org

Hi all,

I've got a question about jitter, usb and spdif. I'm aware of the jitter associated with USB in a typical PCM2707 (or similar) as a USB->I2S receiver chip. However I was wondering how a USB-I2S (via PCM2707)->DAC would compare to a USB->SPDIF (via PCM2707)-> SPDIF (via CS8414 or 8416)-> DAC? Is jitter additive, such that jitter from SPDIF would accumulate in addition to the jitter from USB?

A friend and I have some DACs that we're trying out - but to keep things simple, we're using SPDIF inputs for each of the DACs, and i'm going to use a USB to SPDIF transport to connect into each of these DACs. One of the advantages I can see about doing it this way, versus going USB->IS2->DAC, at least the way we're implementing it - the USB->SPDIF section, the CS841x and DAC chip, and the gain stage will each have independent power supplies.

Thanks,
Clutz

Hey, Clutz. Are you migrating over here from headfi? If so, I think you'll find this place very helpful. I'm not going to answer your question definitively, because I think there are much more knowledgeable people than me to help you. My impression, however, from this forum and others, is that jitter is definitely additive, unless you do some re-clocking.

_________________
He's not hi-fi, he's my stereo.
DIY projects:
http://www.ezdiyaudio.com

Jitter aside, an advantage of SPDIF is the easy possibility of transformer isolation between the computer and audio hardware. There are fiber optical solutions for USB, but I wonder about jitter issues despite the raves on another forum (no, not THAT forum - the other one.)

Jitter from a TOSLINK fibre optic receiver is worse than you would expect from a competent USB I/F.

The fibre attenuation will change every time you connect it or even move it, which in turn has major effects on the receiver pulse width distortion and hence jitter.

I guess I should've been more clear, I won't be using a TOSLink cable for SPDIF, it'll just be a standard coaxial cable. I know - there is a lot to do with SPDIF and cables - but I'm not planning on TOSLink.

Cheers,
Clutz

I am not referring to TOSLINK, but rather fiber optical USB cables. Could have similar issues as TOSLINK though.

Ah.......I can feel safe now to express the truth.

But I'll let him do it. I need a break.

Jocko

I was being lazy before by not posting a link. Here is what I'm referring to:

http://www.audioasylum.com/scripts/t.pl?f=pcaudio&m=11929

http://www.usb-shop.com/opticis.html

Some of the guys first bought this cable to solve ground loop issues and they and others have claimed improved sound over copper USB cables. I wondered aloud in one of the threads what might be happening to jitter with this cable given that opto-isolators don't have a good track record in this area. All I got in response is that the jitter must be improved because it sounds better. I'd rather know what's really happening.

I mentioned this in this thread originally as a possible way to achieve the advantage of galvanic isolation in a USB --> I2S converter arrangement. In a USB --> SPDIF set-up you can use a transformer.

More is better? Don't believe it. May sound different and better to some.

Short fibre optic links can be designed with low jitter, but no better than cable. Long links using lasers will have mode problems that just don't exist on copper.

I used to use led modulators with a linear pin receiver in a fibre video transmission system. I found that the sync pulse was obviously distorted. The led turned out to have a low frequency distortion problem similar to dielectric absorption - probably thermal in origin - but only affecting lower current levels. In a SPDIF application this would certainly cause jitter. Different manufacturers all had this feature.

TOSlink receivers are another story, they have a crude non linear amplifier and simple slicer, that gives bad pulse width distortion that varies with received signal level.

The short links also have mode problems, as too much light is reflected back to the source.

Jocko

Optical links are actually worse in almost every way than proper electrical links for this kind of application.

The advantages of fiber are low loss and wide bandwidth. Both are very dependent on the type of the fiber, the optical wavelength used, and the type of optical source used. There's all sorts of non-linear optical effects and system considerations that are way too complex to describe here. Just one example to keep in mind - I've learned through hard experience that you need better than 50 dB of return loss in an optical link to minimize distortions. Think about that. With an electrical link, you can probably get better results with only 35 dB return loss.

The one true advantage of fiber optics for audio applications is galvanic signal isolation. No ground loops with fiber, that's for sure.

A moderately bad optical link might be better than a monumentally bad electrical link, but that's no endorsement. Better to just get it right in the first place.

Can you expound on "proper" links for transitioning USB to a PCB?

Also, are there any non-optical alternatives that provide similar isolation?

Thanks,
Bryan

Laser sources are strongly affected by reflected signal, LEDs are more or less insensitive.

Very short fibres have a different problem with signal in very high modes and in the cladding getting through with a lot of delay.
Also many receivers, including TOSlinks will simply overload and enter a sort of latch up mode.

At my job we are using plastic optical fibers (POF) for internal communication inside our medical system for the last 15 years. We are using Infineon SFH series of receivers and transmitters which are fine for digital signal transmission with frequency of a few hundred kilohertz. For some time now I would like to improve quality of this link so that it might be used in SPDIF transmission and transmission of PWM signal in class D amplifier. While for SPDIF jitter is the main problem I guess that new breed of POF trancievers like this one would be better that standard toslink.

Pulse width distortion is the main problem for PWM signals. I would like to achieve PW distortion in a few ns range, preferable 1 to 2 ns. For the transmitter I would like to use SFH757V diode and for the receiver SFH250 photodiode.

Transmitter diode is quite powerful, with it's older cousin SFH756 I was able to establish a working link over 60m of unpolshed bundled POF with standard Toslink receiver. I tried to drive it with paralleled AC04 inverters in bridge configuration, but the result was not brilliant. Because of relatively high capacitance I guess it must be reverse biased during turn off.

Receiver is another (even bigger) problem. I am prepared to use 24V supply voltage to speed up the diode, but I have no experience with fiber optic receivers. I am familiar with HP (now Avago) schematics that uses cascaded ECL gates for preamplifier, but since input power will be quite high, maybe use of a single transistor or jfet buffer followed with a fast comparator with a constant fraction discriminator would be a better idea. I would be thankful for some hints or links.