OK, I wasn’t sure where to put this, but since it’s related to my DirectStream use - I’ll put it here.
So I’m following the adventures of a friend who’s doing the budget version of Paul’s Mac Mini upgrade, and I’ve already done my upgrades to my Tiny Form Factor Lenovo PC (SSD drive, Linear PSU, shielding, software tweaks, etc). As we fine tune our source devices for digital media, I sometimes get flustered and go to Ted with a “why does this make a difference!?” question. Oftentimes, his answer boils down to many speculations and not much more…I’m unsure if this is a lack of time or patience with me (and I totally understand! LOL), trade secrets, or he really just doesn’t know because of so many variables…but the one thing these changes and Ted’s responses have in common, is noise transferred from source to DAC.
So I’ll get to the crux of my question - why haven’t we (by we I mean PS Audio or high end audio in general) moved to some sort of Gigabit Glass Fiber connection between our digital gear? Why aren’t we isolating the sources of noise between the gear with fiber? Or hell, is there a way to use fiber between sections of the DAC itself even?
Am I looking at this wrong? Is the “noise” we’re eliminating by tweaking our computers not carried by the copper, but in fact something happening to the transmission of the bits? I guess that’s one of the questions I didn’t feel really got answered in my many “why is this” to Ted. I’ve read articles that mention the leading edge of the bit causing problems, or the leading edge having some sort of an echo slightly +/- the original bit. But I’m no EE by any stretch, I’m just a smart guy who’s really good with concepts trying to wrap my head around all this, and maybe improve the state of the state while I’m at it.
As always, sorry so wordy, and I hope it makes sense what I’m getting at.
Higher speed optical transducers cost significantly more and there’s no particular standard so interoperability is an issue until enough companies settle on a single standard and a “critical mass” is reached. There’s a set of transducers for 100Mb optical Ethernet that aren’t horribly expensive, perhaps more people will be using them. They can carry multichannel quad rate DSD. For example Merging Technologies recently went to Ethernet connections for all of their new recording and editing equipment. That works very well and you can use copper or fiber to connect the units together.
Jitter is the deviation of the edge signal from it’s ideal point in time. In signal transmission theory (e.g. working with GPS, SONET, high speed Ethernet, etc) jitter is very well understood and people realize it’s something to take seriously. People don’t argue about where it comes from or that it shouldn’t be there: they understand the (many) causes and the (sometimes incontinent) solutions.
In audio people used to deny that it was a problem. These days many people that are used to digital transmissions sending essentially perfect bits don’t see how jitter could affect digital audio. Also non-engineers don’t understand the many sources of jitter. Paul has put up some talks I did about jitter up on the PS Audio DirectStream pages. If you haven’t watched them, please do.
We’ve taken some strides in that most people now accept that jitter can make a difference for audio. But the point I’ve been trying to make for about a nine months here is that everyone’s systems have different sensitivities, RF, groundloops, power line noise, etc. Even if all designers understood all of the pitfalls, the cost would be prohibitive to “harden” each component to RF, jitter, HF noise on the power lines, etc. and it the cost would be prohibitive to make each component a good citizen, i.e. to not radiate, not ad noise to the power lines, etc. For an example most people who incorporate linear power supplies actually add a lot of noise to the power lines because linear power supplies typically take huge gulps of current right at the tops of the AC sine waves. That ends up interfering with the other devices taking big gulps at the same time. With switching power supplies you have the choice of having smoother output power (but adding more noise to the input power) or being nicer to the incoming power (but having more noise on the output power.) You can guess the choices most designers make
People seem to think that there’s one answer that helps all systems. There isn’t. You need to learn about your system and the things that matter most for your situation. I keep baskets of power cords, interconnects, optical cables, etc. and every time I get a new component I try multiple versions of them till I find what works best: I’m getting pretty good at guessing what I need to do in my system, but I know that that probably doesn’t help most other people.
Here’s a diagram I did a while ago to show how various kinds of noise can become jitter. There’s no such thing as a clock with instantaneous rise times. Since the rise time has some slope something has to determine when the clock goes from, say 0 to 1. The exact place that decision is made depends on many kinds of noise in the system.
Since (with one bit audio at least) the energy in a bit depends on the level of the bit (i.e. the power supply noise) and the width of the bit (i.e. jitter) the 2nd half shows how, say, -120dB of jitter is “equivalent” to -120dB of noise. Having only -120dB of jitter or only -120dB of noise is vary hard.
stevem2 said
I always hated those "sometimes incontinent solutions" (and spell checkers).
My old Wadia 860/861 CD player has connectors for “AT&T” glass fiber but I never owned anything else that I could connect them to.
AT&T Glass was used by a number of higher end machines like Wadia back in the day, including EAD, Levinson, Madrigal, and Theta. I used Glass between my Madrgial transport and EAD DAC. My friend still uses it on his Wadia gear. We ran many experiments back in those days and we both concluded AT&T Glass was superior to any of the other media transmission options available. I frankly thought the industry made a mistake when they started dropping this, but it was an expensive option compared to the standard forms. But the industry has a habit of using whatever is cheap and handy, even if it isn’t technically correct - classic example is the use of the completely inappropriate RCA connector for digital data transmission instead of BNC.
Gigabit ethernet to optical converter can be bought for like $50 bucks (the component probably cost less than $10). With the 10gbit fiber cable at $10 at monoprice.
Considering the DirectStream sells for $6000, you guys really need to look into implementing optical connections. Electrical noise from the computer is really the biggest problem.
Could you point out a cheap Ethernet to optical converter? You’d still need something like a PS Audio Bridge or other Ethernet streaming device to talk to a DAC. Like I said earlier the transducers aren’t cheap in the quantities we’d be buying them at.
If you were talking about devices with optical audio outputs: I don’t know which devices in particular you are talking about but 99.44% of the optical audio devices out there output TOSLink which we already support. Tho TOSLink is only speced to handle 96k audio with a good cable it will do 192k audio. Double rate DSD however takes a 352.8k pipe. Using something other than TOSLink has it’s own problems: what would you use as a source?
Ground loops are never a problem with Ted’s newest DAC. He does a flawless Pagliacci but tends to lose focus and sound, well, cheesy during the chorus. He will not do the soundtrack from Cats at any bit rate, which is fine by me.
Ted, and esteemed forum members, please forgive me for taking a slight refractive bend to this thread, but it is related to digital inputs.
Ted, Paul indicated that one of the improvements made in Pikes Peak was better performance of the “other” inputs besides the I^2S. My question relates to 75 Ohm coaxial S/PDIF standard. Some of my reading suggests that using a true 75 Ohm cable is essential to optimize the coaxial S/PDIF interface. My understanding is that a regular RCA connector is never a true 75 Ohm connector due to geometry and spacing of the RCA. WBT and others have designed RCAs to come up with 75 Ohm result. Nordost, which used to produce digital coax cables using a WBT RCA connector, now only produces cables with BNC connectors on both ends, and provides BNC-to-RCA adaptors.
Nordost’s FAQs states this: “The correct termination for true 75 ohm impedance is a BNC connector due to the mechanical spacing of the center conductor relative to the outer insulation. We cannot alter the laws of physics. We have found that keeping the integrity of the 75 ohm cable with a BNC termination and using an RCA adaptor always provides better sonic results.”
I was told that even if one device (the digital source, for example) has BNC terminals, and the connected device (DirectStream) has RCA terminals, rather than having a digital cable with BNC on the source end and RCA on the DAC end (no adaptors needed in this case) a better result would be achieved having BNC on both ends of the cable and using one or two adaptors to mate with any RCA terminals when BNC is not provided. Generally it would seem avoidance of adaptors is a good thing, but not in this case, according to Nordost.
Ted, have you observed any difference with BNC versus RCA connections on 75 Ohm S/DIF inputs? Other though for coaxial inputs?
The point is that any RCA in the chain (for example, any kind of adapters from/to RCA) from the source of the S/PDIF connection to the destination will add an impedance discontinuity: i.e. a reflection point for the signal. This will cause ripples in the level of the signal which for many receivers will add jitter as they extract the clock from the S/PDIF signal. The DS doesn’t use a normal S/PDIF receiver nor does it get a clock from the S/PDIF signal. It uses software to look for patterns in the incoming signal and then puts the data in a buffer. The ill effects of an RCA in the circuit are ameliorated.
With the DS the magnitude of any differences caused by RCAs will be much smaller than the magnitude caused by various shielding choices in the cabling involved or ground loop issues, etc. If you happen to have BNC/RCA mismatches feel free to put a RCA/BNC adapter wherever you need it or want it.
Ted, did you ever try replacing the RCA connectors on the DS and source with BNC to see if the result was audible? Seems like an easy fix if you are proficient with such things as you are.
Ted Smith said
If you happen to have BNC/RCA mismatches feel free to put a RCA/BNC adapter wherever you need it or want it.
I have a number of pieces of kit with BNC connectors. I mix and match with adapters and make cables with BNC on one end; RCA, the other. All with no issue. I have yet to hear a difference.
In general BNC’s are definitely the way to go over RCA for digital: But you need to replace the connectors on both the source and the destination (and not use any RCA/BNC convectors anywhere on the connection) to get the advantage. On the DS is doesn’t matter so why use the (much) rarer connector?
When I factor ground loops out, the termination of the cables doesn’t make any difference (S/PDIF, AES/EBU or I2S) but most of the time when people are comparing inputs they use inputs from different sources and hence aren’t comparing apples to oranges with respect to ground loops. Yes, I have RCA/BNC adapters and cables and use them as I need them regularly, I also at times just use an alligator clip. If you have mismatched cables for some reason, you can use RCA/BNC adapters… As I said they just don’t matter as much on the DS.
