Yeah so was I when I first heard about NOS - it goes against all the theory I've taught myself. How can sending staircases to the amp do anything decent to the sound?😱
Well it is rather a paradigm shift - still getting my head around the implications even now...
Well I'll pass on the open loop (at least for now) and the toobs (too limited vis a vis portability) but yeah, I wanna hear what the other enhancements bring to the party. Ever played with AD603 as output stage?
Hi,
Peak levels should be, for well recorded CD's, with average levels around 14dB below 0dBfs for acoustic music, this is called crest factor.
Normally the signal spends much of the time below -10dBfs, unless you have modern, compressed to near DC pop/rock..
It is not quite like that, have a look at the codes vs signal levels in some of the older DAC Datasheets.
No, we need all of them.
Not quite.
For acoustical recordings however we need to note that most recording mikes have around 22dBA to 26dBA self noise. Meaning the noise from the microphone is equivalent to around this SPL.
Peak SPL's from instruments can be of course very high, however minimalist acoustic recordings rarely have more than 100dB SPL average SPL with maybe 14 to 20dB peaks.
So one may argue that whatever can be recorded can be represented in 16 Bit Audio, if used optimally.
This does not mean having more bits during the recording is a bad thing. One can leave more footroom and headroom and later correctly scale the recording in terms of level... Much depends on post production.
And releasing recordings at more than 44.1KHz sample rate is probably "a good thing", given how large the subjective quality step from 44.1 to 48 is. Actually, C37 Guru Dieter Ennemoser suggests that the problems created by 48KHz sample rate are "C37" compatible and hence masked better by the human hearing than those from 44.1KHz... Not sure I believe his explanation fully though.
Ciao T
Peak levels should be, for well recorded CD's, with average levels around 14dB below 0dBfs for acoustic music, this is called crest factor.
Normally the signal spends much of the time below -10dBfs, unless you have modern, compressed to near DC pop/rock..
It is not quite like that, have a look at the codes vs signal levels in some of the older DAC Datasheets.
No, we need all of them.
Not quite.
For acoustical recordings however we need to note that most recording mikes have around 22dBA to 26dBA self noise. Meaning the noise from the microphone is equivalent to around this SPL.
Peak SPL's from instruments can be of course very high, however minimalist acoustic recordings rarely have more than 100dB SPL average SPL with maybe 14 to 20dB peaks.
So one may argue that whatever can be recorded can be represented in 16 Bit Audio, if used optimally.
This does not mean having more bits during the recording is a bad thing. One can leave more footroom and headroom and later correctly scale the recording in terms of level... Much depends on post production.
And releasing recordings at more than 44.1KHz sample rate is probably "a good thing", given how large the subjective quality step from 44.1 to 48 is. Actually, C37 Guru Dieter Ennemoser suggests that the problems created by 48KHz sample rate are "C37" compatible and hence masked better by the human hearing than those from 44.1KHz... Not sure I believe his explanation fully though.
Ciao T
Hi,
The SAA7220 does more than that, the offset and attenuation are more incidental...
Not if recorded with optimum levels and modern recording microphones...
Ciao T
The SAA7220 does more than that, the offset and attenuation are more incidental...
Not if recorded with optimum levels and modern recording microphones...
Ciao T
Hi,
Well, the problem is that we often mistake apparency for actuality.
The 'scope shows us a staircase and we go "😱"
Well, I'll give two to think about:
1) Non-Os creates an ultrasonic image that approximates quite closely the ultrasonic spectrum of music, with appropriate levels that track the music closely.
It has been demonstrated repeatedly that though inaudible as such with steady state tones (and with very poor pitch definition for anything else) ultrasonic content is "percieved", with the mechanisms and implications debated.
However this in the end, Non-OS provides a re-synthesised ultrasonic content that closely resembles what would be expected with music in reality.
The same is largely true of the "primitive unless detecting a sinewave" digital filters by Wadia (Digimaster) and Pioneer (Legato Link) and of the ultra-primitive digital filters like the early Luxman "Fluency" filter.
2) Douglas Rife has argued that the ultrasonic images from ASRC (and by implication also non-os) provide a signal-level dependent "dither" that improves linearity.
I do not claim that either is the sole, or correct explanations, just tome theories to boil your (collective) noodle(s), they do kind fit the observations and are well shaved with old Bills razor, but no more than that...
Toobs I understand, but do try open-loop. The OPA860/861 diamond transistors look tailor-made for open loop IV and output drivers on DAC's...
No, had a look, looks interesting...
Ciao T
Well, the problem is that we often mistake apparency for actuality.
The 'scope shows us a staircase and we go "😱"
Well, I'll give two to think about:
1) Non-Os creates an ultrasonic image that approximates quite closely the ultrasonic spectrum of music, with appropriate levels that track the music closely.
It has been demonstrated repeatedly that though inaudible as such with steady state tones (and with very poor pitch definition for anything else) ultrasonic content is "percieved", with the mechanisms and implications debated.
However this in the end, Non-OS provides a re-synthesised ultrasonic content that closely resembles what would be expected with music in reality.
The same is largely true of the "primitive unless detecting a sinewave" digital filters by Wadia (Digimaster) and Pioneer (Legato Link) and of the ultra-primitive digital filters like the early Luxman "Fluency" filter.
2) Douglas Rife has argued that the ultrasonic images from ASRC (and by implication also non-os) provide a signal-level dependent "dither" that improves linearity.
I do not claim that either is the sole, or correct explanations, just tome theories to boil your (collective) noodle(s), they do kind fit the observations and are well shaved with old Bills razor, but no more than that...
Toobs I understand, but do try open-loop. The OPA860/861 diamond transistors look tailor-made for open loop IV and output drivers on DAC's...
No, had a look, looks interesting...
Ciao T
I start my holiday tommorow, so I will finaly have some spare time, I've printed this whole thread out so I can digest the relevant points.
Things are becoming clearer now😀
I'm just adding the device to my CAD library (Cadstar), to give you an idea how far divorced I have become from my past in electronics is that our current library only had two through hole components, both connectors! I was thinking about this the other day, and realised I was becoming more and more distanced mentaly from the good old days of electronics and layout. Most (all) designs these days involve multiple BGA's (FPGAs, CPLDs DDR memory), SMPS's and usually 10-20 layers. So playing with this is device is a refreshing change, and a good re-education.
Thanks
Things are becoming clearer now😀
I'm just adding the device to my CAD library (Cadstar), to give you an idea how far divorced I have become from my past in electronics is that our current library only had two through hole components, both connectors! I was thinking about this the other day, and realised I was becoming more and more distanced mentaly from the good old days of electronics and layout. Most (all) designs these days involve multiple BGA's (FPGAs, CPLDs DDR memory), SMPS's and usually 10-20 layers. So playing with this is device is a refreshing change, and a good re-education.
Thanks
While enjoying your holidays, by all means... Do try some of the suggestions on power supply and decoupling. I will be doing the same in a couple of weeks.
The discussion has moved on a bit towards output stages, as so many discussions on dacs do. I feel the output (gain) stage discussion is a matter of personal taste, while the psu/decoupling/layout discussion is about optimizing the correct implementation of this chip to unleash its potential.
The discussion has moved on a bit towards output stages, as so many discussions on dacs do. I feel the output (gain) stage discussion is a matter of personal taste, while the psu/decoupling/layout discussion is about optimizing the correct implementation of this chip to unleash its potential.
It's an important factor of the overall quality of the sound, no doubt. But so many discussions have already centered around this topic, and there is no truth about what sounds good. Some say opamps, some discrete, tubes, feedback or non feedback, all kinds of topologies... And all thesr discussions end with: "i'll be using [output stage name here] because that's what sounds best to my ears in my system".
The discussion of power supply/decoupling/and board layout is less subjective, and surely a less explored topic, which is quite specific to this chip. Hence my interest to further digest this topic.
The discussion of power supply/decoupling/and board layout is less subjective, and surely a less explored topic, which is quite specific to this chip. Hence my interest to further digest this topic.
Hi Studiostevus,
that is what I am going to focus on, I am basicly doing it as an interlectual exercise (the complex supplies intrigue me), and will concentrate only on the decoupling and planes around the TDA. I will post the layouts so we can discuss the various aspects of the different layouts and thier pros and cons.
I am also in the final stages of finishing some open baffle speakers, so will be playing with them as well.
that is what I am going to focus on, I am basicly doing it as an interlectual exercise (the complex supplies intrigue me), and will concentrate only on the decoupling and planes around the TDA. I will post the layouts so we can discuss the various aspects of the different layouts and thier pros and cons.
I am also in the final stages of finishing some open baffle speakers, so will be playing with them as well.
Hi,
Kids, you are both wrong...
With the TDA1541 ALL aspects must be "just so" or you loose some performance...
The I/V stage is important to ensure linearity from the DAC (even if we afterwards mess up all this nice linearity with open loop tube circuits).
The DAC's output is quite limited in the voltage it can swing for best linearity and the output spectrun IMNSHO is not nice when this exceeeded and the extra 2mA offset do not help the situation.
Some active I/V stages can have rather large source/emitter impedances that are worse than resistor based I/V...
The complex supplies and the fact that one of the supplies is technically speaking analogue ground for closing the analogue current loops does not help.
There are many other areas in the TDA1541's design where the technology of the age it is from and the innovations in it's design make for "interesting" behaviour and requirements.
The fun part is that is that a very nearly "generic" design with just CS8412, TDA1541 and a simple passive I/V stage and a tube gainstage can already sound good enough to outclass many modern DAC's (like for example the Zanden DAC), but the potential attainable by optimising the circuit design and layout is so much greater...
And let's not forget the way DEM reclocking frequencies and layouts can interact as well...
Ciao T
Kids, you are both wrong...
With the TDA1541 ALL aspects must be "just so" or you loose some performance...
The I/V stage is important to ensure linearity from the DAC (even if we afterwards mess up all this nice linearity with open loop tube circuits).
The DAC's output is quite limited in the voltage it can swing for best linearity and the output spectrun IMNSHO is not nice when this exceeeded and the extra 2mA offset do not help the situation.
Some active I/V stages can have rather large source/emitter impedances that are worse than resistor based I/V...
The complex supplies and the fact that one of the supplies is technically speaking analogue ground for closing the analogue current loops does not help.
There are many other areas in the TDA1541's design where the technology of the age it is from and the innovations in it's design make for "interesting" behaviour and requirements.
The fun part is that is that a very nearly "generic" design with just CS8412, TDA1541 and a simple passive I/V stage and a tube gainstage can already sound good enough to outclass many modern DAC's (like for example the Zanden DAC), but the potential attainable by optimising the circuit design and layout is so much greater...
And let's not forget the way DEM reclocking frequencies and layouts can interact as well...
Ciao T
I am not a kid
Thorsten, I am not a kid being double your age. 😡
I do not agree with most if your ideas.
🙁
Show some respect man!
Thorsten, I am not a kid being double your age. 😡
I do not agree with most if your ideas.
🙁
Show some respect man!
Hi,
Really THAT ripe an old age... 🙂
Then don't use them.
The use of "kids" was meant to be humorous, sorry if it offended...
Ciao T
Really THAT ripe an old age... 🙂
Then don't use them.
The use of "kids" was meant to be humorous, sorry if it offended...
Ciao T
Outclass the zanden dac? Quite a statement, i thought zanden had its stuff together
(or is this rebelism: its expensive, therefore must be a flawed design)
Ok, point taken: the output stage's interaction with the dac and impact on the chip's performance do matter. But the gain section/buffer after the primary i/v conversion less so. (again, i dont mean they are unimportant, but they are very subjective, and a multitude of solutions can render satisfying results, e.g thorsten uses a different stage than the amr design, although both are appreciated by their owners)
Anyhow, i am looking forward to marce's intellectual exercise to optimize power supply and decoupling.
(or is this rebelism: its expensive, therefore must be a flawed design)
Ok, point taken: the output stage's interaction with the dac and impact on the chip's performance do matter. But the gain section/buffer after the primary i/v conversion less so. (again, i dont mean they are unimportant, but they are very subjective, and a multitude of solutions can render satisfying results, e.g thorsten uses a different stage than the amr design, although both are appreciated by their owners)
Anyhow, i am looking forward to marce's intellectual exercise to optimize power supply and decoupling.
Might not be the topic of this thread but im looking at ways to get around the 5,6MHz limit to be able to feed the dac 192k data and a few pointers to with route to take would be much appreciated.
reduced clockrate (and data ) by half is a possibility but a period of reduced activity of all the shift registerers before conversion would he advantageous and changing the mode and feeding both channels simultaneously makes the dac latch with wclk wich have to be divided down from the master clock.
Im not sure witch route to take and some input would be welcome
reduced clockrate (and data ) by half is a possibility but a period of reduced activity of all the shift registerers before conversion would he advantageous and changing the mode and feeding both channels simultaneously makes the dac latch with wclk wich have to be divided down from the master clock.
Im not sure witch route to take and some input would be welcome
Hi,
You misunderstand, I meant the Zanden DAC with it's antique TDA1541 and a fairly straightforward digital design (though complex analog filter) and simple analogue stage beats many modern DAC Designs.
I preferred not to make any claims regarding anything I had my hand in...
Yes, the key is to limit the voltage swing on the output to a sensible minimum.
They have more in common than different. AMR would use 6072A if we could get enough in the open market. A simple Hybrid Mu-Follower or a Gomes stage are both very clean, low output impedance etc. I would have used a Gomes stage as in the AMR designs, but it would not fit...
So do I.
I may learn a fair bit... My layout skills are mostly still 20 years ago, analog, mixed signal and radio. It will be interesting to see how someone approaches is who has this same background, but has kept up with the times, when they are a changing (while I choose to change my career instead)...
Ciao T
You misunderstand, I meant the Zanden DAC with it's antique TDA1541 and a fairly straightforward digital design (though complex analog filter) and simple analogue stage beats many modern DAC Designs.
I preferred not to make any claims regarding anything I had my hand in...
Yes, the key is to limit the voltage swing on the output to a sensible minimum.
They have more in common than different. AMR would use 6072A if we could get enough in the open market. A simple Hybrid Mu-Follower or a Gomes stage are both very clean, low output impedance etc. I would have used a Gomes stage as in the AMR designs, but it would not fit...
So do I.
I may learn a fair bit... My layout skills are mostly still 20 years ago, analog, mixed signal and radio. It will be interesting to see how someone approaches is who has this same background, but has kept up with the times, when they are a changing (while I choose to change my career instead)...
Ciao T
Hi,
No need to get around it. Any TDA1541A I tried workes at 6,144MHz, so just make sure to use 16 Bit I2S mode (BCK = 32 Times WCK).
Ciao T
No need to get around it. Any TDA1541A I tried workes at 6,144MHz, so just make sure to use 16 Bit I2S mode (BCK = 32 Times WCK).
Ciao T
My copy of the datasheet (Feb 91) says it maxes out at 6.4MHz, that's good enough for 192kHz. You can go up to 384kHz by adopting a different input format where left and right channel data enter on their own dedicated pins.
Thorsten, does the AMR using transformer after the R i/v and before tube? in that case I remembered one of your schematics if I'm not wrong...🙂 is it one of your 'wisdom' here ?
Hi,
I started with a long time ago. I removed transformers at that point long before Audio Note started doing so...
So, short and sweet, no.
Ciao T
I started with a long time ago. I removed transformers at that point long before Audio Note started doing so...
So, short and sweet, no.
Ciao T
- Status
- Not open for further replies.