Phew...did I get away with my previous post to you..I was afraid I was "gonna get it". But all I got was a reasoned/balanced view!
You are right ofcourse. Just to continue a little..there has to be said that there is a huge difference between audio engineers and electronics engineers. I remember when the co-ax vs optical thing was pretty much cut and dried in audio magazine and audiophile circles that toslink optical was generally worse than co-ax. The AT&T that Wadia used I think was seen as better than both co-ax and toslink optical. I told this one guy that on his last year of study (electronics)...and the way he looked at me when I told him that co-ax generally was found to be better...just said..."you don't know what you are talking about do you!"..But a chat to an engineer can be very useful because when he asks "why is that" you have to put things straight in your head..
Sierra Proto Express (http://www.sierraprotoexpress.com) will run "no touch" 4-layer PCBs for $51/pop, minimum two, with a $20 setup. They're the cheapest 4 layer place i've found, and i'm very satisfied with their work.foo said:
I don't think you'll need anything higher than 4 layers for a DAC...
Sounds like your buddy is the one that doesn't know what they are talking about
Unless the coax is a really bad design, like the stock Philips output designed to minimise EMI, it is better.
TOSLINK has too many problems to even think about using it.
As for the ATT fiber...........
Yeah, it could be very good, but it wasn't. All of the implementations, especially Wadia's, were flawed. So there.
You can tell your buddy that comes from an EE who has worked in both telecom and audio probably longer than he has been alive.
Kids.................youth is wasted on the young.
Jocko
Unless the coax is a really bad design, like the stock Philips output designed to minimise EMI, it is better.
TOSLINK has too many problems to even think about using it.
As for the ATT fiber...........
Yeah, it could be very good, but it wasn't. All of the implementations, especially Wadia's, were flawed. So there.
You can tell your buddy that comes from an EE who has worked in both telecom and audio probably longer than he has been alive.
Kids.................youth is wasted on the young.
Jocko
"What's in it? who cares! i mean, it's a f**ing Levinson!"
Only works for the "big boys". The rags will gush about what is inside, because of who makes it, and it will become the de facto standard that all other brands will be judged by. If you don't use the same stuff that Levinson uses, your product will be judged to be inferior. The rags will have the average audiophile convinced that if Levinson et al use it, then it has to be good.
They wouldn't use it if it wasn't? Would they? No, it has to be good. Doesn't it???
So, some upstart comes along (happens all the time), and they have better sense than Levinson. See how many units they sell with their superior knowledge and engineering.
Don't belive me???? Try it sometime, and see if you end up spending more time that I have explaining why something designed differently from Levinson can be good.
You are sort of right.......they don't actually care what it is, as long as Levinson has it, it has to be good. Period. Anything that doesn't must be dreck and chintz.
Jocko
Only works for the "big boys". The rags will gush about what is inside, because of who makes it, and it will become the de facto standard that all other brands will be judged by. If you don't use the same stuff that Levinson uses, your product will be judged to be inferior. The rags will have the average audiophile convinced that if Levinson et al use it, then it has to be good.
They wouldn't use it if it wasn't? Would they? No, it has to be good. Doesn't it???
So, some upstart comes along (happens all the time), and they have better sense than Levinson. See how many units they sell with their superior knowledge and engineering.
Don't belive me???? Try it sometime, and see if you end up spending more time that I have explaining why something designed differently from Levinson can be good.
You are sort of right.......they don't actually care what it is, as long as Levinson has it, it has to be good. Period. Anything that doesn't must be dreck and chintz.
Jocko
gmarsh said:
but these DIR's still are not good enough !!!!!!!!!!!!
Then mankind invented SRCs. These aren't good enough either
And all because no-one can design proper PLL circuits.... (I am not stating the TentLabs PLL is state of the art, we could do a better job but sales price would be affected then)
My advise: The RAM buffer concept is fine, but its' succes depends on the implementation. I never heard an implementation that was totally insensitive for the quality of the drive feeding it.
On chip and PCB crosstalk still spoil the " isolation" required
succes with the project, you shouldn't be discouraged: doing a layout in 10 days will bring you a PCB that will work but that is different from something that meets specs and sounds OK in the ears of those wanting to buy it.
foo said:
I am very interested since I thought to do something similar with my friend. There are several things that we think is important
a) bnc, optic and USB inputs.
b) never say "don't worry about technical detail", I would like that we have some influence on the shematics, or at least to discuss it and finally find the best solution
c) no DSP, microcontrolers and stuff unless you plan to make it "opensource". If you have a microcontroller you have a complete control over the project which is not idea, at least as I understande DIY. DIP switches are fine and one could play more easy. There is SRC4192 that works in parallel mode.
d) mono output and choice on slow and fast filter is preferable
e) we would be at the end interested first in PCB, than if your prices are correct maybe even with a PCB with soldered SMD components.
f) if it is possible PSU and whole circuit on one PCB, I hato doing cabling
c) we are prepared to pay up to 200$ with most importand SMD components (DAC is around 50$ for the pair, other digital stuff 50$ for the pair and that leavs 100$ for the PCB which should be plenty if you have a bigger volume)
As someone already noted, in the predesign considerations that I have done I do not see a need for a four layer PCB. Since I am not a great PCB designer maybe someone could explain the advantages of paying more
Re: Sounds like your buddy is the one that doesn't know what they are talking about
Explain the 'problems' in more detail.
Right now, the DIR+SRC implementation is the way to go IMO... The clock that comes out of a CS8416 can't even be described (look at it on a spectrum analyzer, you'll laugh) and the clock that comes out of a TI DIR chip is considerably better (phase noise isn't as awful, far less spurious).
Though people cry foul with ASRC chips, fact is the good ones (AD1896) eliminate a substantial amount of jitter in their rate estimation loop and allow a high quality crystal to be used to clock the output DAC. Low frequency jitter (in the Hz) still gets encoded in the data which offends some people, and the effect of digital low pass filtering + dithering on the audio does bring about all sorts of issues, some measurable and some subjective. Essentially you have a choice of correct data with a bad clock or modified with a good clock... between mathematical reasoning, knowing what the CS8416's clock output looks like and doing listening testing with my jitter machine*, the ASRC approach wins me over.
As far as PLL's go, using a naked 1:1 PLL to clean up the output of a DIR chip makes sense. But to get the same jitter rejection as a good ASRC chip requires one hell of a good PLL design with a crossover frequency of a few Hz. In a world of 1% resistors and 5% capacitors with leakage currents that vary constantly throughout the same batch, building a such a PLL using regular DPLL design which good characteristics (locks in a reasonable amount of time, doesn't ring on transients or slip) is an excercise in maintaining your sanity.
And I speak this from experience - you think you audio clock guys have it hard, I design PLL's for high end FM radio broadcast equipment. My phase noise gets demodulated!
If I were to build the ultimate "jitter buster", I'd go with digital control and build a FIFO + adaptive DLL system using something like a Spartan-XL FPGA + configurator, lots-of-bits ladder DAC, filter and a high quality VCXO. Won't be cheap though.
*jitter machine = I modified a $20 Norcent DVD player to have an external master clock input. When driven with jittery clocks from various test equipment, the same garbage shows up on the SPDIF output and you can get all sorts of audible effects in different receiving equipment. Highly recommended, they're a cheap hack and they're a fun learning experience.
Jocko Homo said:
Explain the 'problems' in more detail.
Guido Tent said:
Right now, the DIR+SRC implementation is the way to go IMO... The clock that comes out of a CS8416 can't even be described (look at it on a spectrum analyzer, you'll laugh) and the clock that comes out of a TI DIR chip is considerably better (phase noise isn't as awful, far less spurious).
Though people cry foul with ASRC chips, fact is the good ones (AD1896) eliminate a substantial amount of jitter in their rate estimation loop and allow a high quality crystal to be used to clock the output DAC. Low frequency jitter (in the Hz) still gets encoded in the data which offends some people, and the effect of digital low pass filtering + dithering on the audio does bring about all sorts of issues, some measurable and some subjective. Essentially you have a choice of correct data with a bad clock or modified with a good clock... between mathematical reasoning, knowing what the CS8416's clock output looks like and doing listening testing with my jitter machine*, the ASRC approach wins me over.
As far as PLL's go, using a naked 1:1 PLL to clean up the output of a DIR chip makes sense. But to get the same jitter rejection as a good ASRC chip requires one hell of a good PLL design with a crossover frequency of a few Hz. In a world of 1% resistors and 5% capacitors with leakage currents that vary constantly throughout the same batch, building a such a PLL using regular DPLL design which good characteristics (locks in a reasonable amount of time, doesn't ring on transients or slip) is an excercise in maintaining your sanity.
And I speak this from experience - you think you audio clock guys have it hard, I design PLL's for high end FM radio broadcast equipment. My phase noise gets demodulated!
If I were to build the ultimate "jitter buster", I'd go with digital control and build a FIFO + adaptive DLL system using something like a Spartan-XL FPGA + configurator, lots-of-bits ladder DAC, filter and a high quality VCXO. Won't be cheap though.
*jitter machine = I modified a $20 Norcent DVD player to have an external master clock input. When driven with jittery clocks from various test equipment, the same garbage shows up on the SPDIF output and you can get all sorts of audible effects in different receiving equipment. Highly recommended, they're a cheap hack and they're a fun learning experience.
Could you elaborate? Is there anything I can do to improve that?
The marketing department says something different..
http://www.cirrus.com/en/products/pro/detail/P1005.html
Gmarsh,
Do You think of something like this?
http://www.lavryengineering.com/white_papers/jitter.pdf
Do you think with a micro it would be still expensive? Ok, the high quality ladder dac should still be there.
Interesting approach.
Ciao, George
Do You think of something like this?
http://www.lavryengineering.com/white_papers/jitter.pdf
Do you think with a micro it would be still expensive? Ok, the high quality ladder dac should still be there.
Interesting approach.
Ciao, George
Almost exactly what I had in mind... though I planned on going all-out with the design and using full adaptive PI control with quality of lock indication and the whole works, not just an up-down counter.Joseph K said:
It's possible to use a bunch of discrete parts (SRAM chip or FIFO chip surrounded by 74x logic) along with a micro, but I think the smallest/most "elegant" solution will be a single FPGA. Maybe even a big CPLD, though the amount of memory available inside those is considerably less. I wish you could buy a dsPIC microcontroller with two codec interfaces on it, it'd be perfect for the task.
Well, for starters, the rise and fall times are nowhere near symetrical. It is a multimode transmission, so there is no way that the pulse will arrive without any dispersion. I have never taken that concept serious enough to examine if it can work properly on a 1 m. piece of plastic fiber. I know for a fact that the ATT will not.
All optical transmisson schemes are designed with (what us RF types would call) a certain path loss. When you have a link that it "too hot", the laser has problems. In the case of ATT, anything under 1 km. is "too hot". Don't know what the optimum link is for TOSLINK, but I doubt it is 1 m. Sorry, but I no longer have access to all of the optical spectrum analysers and such it would take to give a definitive answer. But is TOSLINK, which is multimode. Hardly worth the effort.
Jocko
All optical transmisson schemes are designed with (what us RF types would call) a certain path loss. When you have a link that it "too hot", the laser has problems. In the case of ATT, anything under 1 km. is "too hot". Don't know what the optimum link is for TOSLINK, but I doubt it is 1 m. Sorry, but I no longer have access to all of the optical spectrum analysers and such it would take to give a definitive answer. But is TOSLINK, which is multimode. Hardly worth the effort.
Jocko
How did I miss this one.........
Well, I would worry. Everyone who has heard of Ohm' Law thinks that they can design audio gear.
OK, so it is easier than building a Mars lander. Still doesn't mean a handful of computer weenies can make a go of it.
Just like when they say "It isn't about the money", it always is about the money. So..................don't worry about the technical details! The details will design themselves.
Yeah, maybe we should encourage them. I need a cheap laugh.
Jocko
Well, I would worry. Everyone who has heard of Ohm' Law thinks that they can design audio gear.
OK, so it is easier than building a Mars lander. Still doesn't mean a handful of computer weenies can make a go of it.
Just like when they say "It isn't about the money", it always is about the money. So..................don't worry about the technical details! The details will design themselves.
Yeah, maybe we should encourage them. I need a cheap laugh.
Jocko
hm, if you take a look at the innards of such a product, I would bet that it's not an up-down counter what they use..
a nice dac from inside
But in that article Lawry did the homework [at least for me] to estimate the necessary bit depth of the FIFO, the ~ optimal resolution of the dac etc. For me the nice fact is that it's not neccessarily deep, neither of them.
With a micro [or dspic] one could apply different filter algorithms, and even changing them on the fly. When using the FIFO pointer / VCO pll approach. But it would also be possible to simply MEASURE input recovered data clock / local osc. frequency, adjust them to the closest possible, and start the pll algho with those initial parameters. Bye bye pull-in time..
Then, with this approach, as Lawry has pointed out as well, you can apply optical /magnetic decoupling at a [few] serial data points, and so separate totally the noisy digital recovery section from the actual analog D/A converter.
magnetic couplers
Ciao, George
a nice dac from inside
But in that article Lawry did the homework [at least for me] to estimate the necessary bit depth of the FIFO, the ~ optimal resolution of the dac etc. For me the nice fact is that it's not neccessarily deep, neither of them.
With a micro [or dspic] one could apply different filter algorithms, and even changing them on the fly. When using the FIFO pointer / VCO pll approach. But it would also be possible to simply MEASURE input recovered data clock / local osc. frequency, adjust them to the closest possible, and start the pll algho with those initial parameters. Bye bye pull-in time..
Then, with this approach, as Lawry has pointed out as well, you can apply optical /magnetic decoupling at a [few] serial data points, and so separate totally the noisy digital recovery section from the actual analog D/A converter.
magnetic couplers
Ciao, George
Jocko Homo said:
Good explanation... I never thought of multimode operation, and it can certainly cause jitter. How much of an effect it actually makes is still up in the air.
Another thing I've thought about is reflections... a badly terminated coax cable can reflect, causing "echos" on the cable which is probably one of the worst things you can feed a clock recovery circuit. How the heck do you get an "impedance" right on an optical cable so the far end isn't shiny at all?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.