TSS Two Chassis Super DAC

High end audio is a lot like higher end cameras and projectors. The really nice lenses that get placed inside these devices do not usually benefit from economies of scale. The products they go into are already expensive for other reasons. The same thing applies with high end audio. These lenses take longer to polish, they’re typically placed in better (metal) housings, have special ED elements, have high quality (expensive) coatings and due to their larger relative size literally require a larger amount of everything to manufacture. High end audio is similar in that we don’t typically build high end components in large enough quantities to benefit from economies of scale. Higher quality components are used as well as larger components (think oversized power supplies and magnets). These raw materials simply cost more money due to their being more it used to manufacture the product. You want nice wood veneers? High gloss polish? Not made in China? A shipping container that will hold up against UPS delivery men? An aluminum chassis cut from a single piece of metal? I think you get my point. Some of this “over-built-ness” adds to the sound quality. So while I think there will be clever ways to get the budget stuff to sound better and better over the years (as it always happens) there are very real tangible reasons why high end audio (like other enthusiast hobbies) will always remain high in price.

Maybe. But Ted’s brilliant conversion algorithms can be replicated at zero marginal cost. This inherent property of all things digital drives down price / performance dramatically.

And what would that beautiful code that runs the digital side of things benefit from with an analog section inside an entire DAC that can be manufacturered for $300? How would it sound? It wouldn’t sound anything like the DS Sr. That is the main difference between the Sr. and Jr. and the audible difference was easy to hear from my experience.The beauty in the Sr (and this new TSS) is in the analog section.

1 Like

The algorithms alone aren’t what make the DS special. It’s the whole-system design, and it’s critically dependent on the very expensive physical things like power supplies, clocks and transformers.

And the main point is that you have to use high quality components (which are not normally cheap) to get Ted’s code to sound great.

I like to tell myself that the software of the DS is like the hardware of the DS. People may see it, they may understand it, but I’m not worried about them “ripping it off” because they will never use the expensive hardware parts nor will they do the expensive things in the software that make critical differences. Upsampling filters have been around for a long, long time - but many think that some mathematical identities work in filter design where they really don’t. (Since there isn’t a perfect filter, simplifications based on the assumptions that a filter is perfect don’t necessarily have the desired effects (or lack of effects.)) For example look at the disbelief that Rob Watts gets for claiming he needs 360dB S/N. I just say 60 bits, but you do need very high precision at certain points to keep from loosing some of the “audio goodness.”

3 Likes

So I suppose it goes both ways. What good is the excellent software if your hardware sucks and it works vice-versa.

I’ve found that most people can’t get their heads around the fact that DSD is just a special case of the more general PCM form, and therefore can’t understand how you can claim to preserve every bit of input whether DSD or PCM even through upsampling to 56.448MHz and performing digital volume adjustment.

Let alone why on Earth you’d want to do such a thing :slight_smile:

I just did some research on old Panasonic “MASH” sigma-delta chips. The concept is not dissimilar but the quality of implementation is worlds apart. The modern successors are said to be the PCM1794 et al. Anybody who’s content with the sound that produces is welcome to save their dollars for something else.

TEd

Is there anything different from the DS in the TSS that will alleviate the need for many hundreds of hours of Burn in to sound good

magicknow

There should be no burn-in issues for the digital box which means that there very probably won’t be the short term changes just after an update or power on of the digital box.

The analog box will still need some time to sound it’s best: I have no idea if the Jensen transformers will require more or less time than the DS’s transformers. By design I want the analog to be insensitive to heat and run cool. Those will make burnin take a little longer. On the other hand some of the components run a little hotter and further I have some 1W resistors around the board that I can turn on or off with software that can warm up critical parts of the board. Perhaps we’ll put a burn-in option on the UI :slight_smile: (Probably not really.)

There also are explicit changes to keep the analog box in the steady state while the digital box is off-line (powered off, being upgraded, etc.)

1 Like

What kind of signal is sent from the digital box to the analog one over optical? Is it some proprietary signal format you’ve created? Is it some sort of transport format or is the signal already analog in nature at this point?

It’s a highspeed serial (but easy to decode) signal. It has multiple channels for DSD and a few more for control of things like the grounding selection relays, attenuators, mute, etc. By keeping it simple I can do a self aligning serial to parallel conversion synchronously with the system clock and reclockers to keep the noise and potential clock beating down.

Thanks Ted. I appreciate you taking the time to answer questions.

DSD blurs the line between digital and analog. It’s cool though… there are these chopped fragments of left channel and right channel DSD audio interspersed with some other control bits coming down the fibre. The DSD data will have to be transmitted faster than playback speed, then buffered, aligned and played back at the correct speed.

But it’s not invalid to think about those DSD fragments being effectively analog bursts of sound in the form of light pulses. If you directed one of them into a suitable light-dependent-resistor in series with a speaker and an applied voltage, you’d get a brief moment of sped-up audio.

@tedsmith – do you have a figure on what the total delay across the optical fibre will be due to buffering at both the transmitting and receiving ends? I’m hoping it won’t be noticeable for AV contexts.

The DSD isn’t buffered in the analog box. It shows up just in time. The input buffering is in the digital box, that lets the analog box be dumber and not have multiple unrelated clocks (which is always evil.) The analog box measures the phase on the return signal and uses a phase aligner to advance or retard upgoing clock till the downcoming bits are exactly aligned with the master clock. (Note that it’s aligning a copy of the master clock, not the master clock itself.) The digital box runs a PLL on the clock from the analog box and does everything synchronous to that (just like the DS’s FPGA does.)

There shouldn’t be a hardware difference in the input to output delay between the TSS and the DS (i.e. it will be less than a DSD256 sample time.)

Hmm, so your multiplexing is per-sample? That’s wild, I love it. Everything I wrote earlier is wrong for the TSS context but otherwise I stand by it :smiley:

I am deliriously happy with my DSD Sr., and to be quite frank, I never thought I would consider shelling out the cash for the TSS. Not to say it is not worthy of the price tag but as we get to that 95 percentile and have to spend ungodly to leach out that last few percentages of SQ nectar, it becomes more of a gut check or moral experiment than anything else. The more I read Ted’s posts in this thread, the more I realize that there is pretty much a zero change the TSS will not be in my stable.

1 Like

Are the output devices the same as those used in the DS?

I’m not sure if you are talking kinds of connections, transformers, digital switches, or what.

There will still be a “parallel” unbalance and balanced connectors. There will be some configuration for using the RCA balanced as well.

The transformers will be the nice Jensen’s.

The digital switches will be higher current lower noise versions.

Thanks, Ted. It was the digital switches I was curious about. Are we privy to the part number?