pcm56 questions

BeerCan

Member
2005-06-14 8:18 pm
US
Hi,
In my reasearch to build a pcm56 non os dac I came accross the folowing circuit PCM56 NONOS The circuit looked simple enough for me to study.
My question has to do with the inverter (74hc04). Is this necessary? I guess I really don't understand why it is there. I know this pin on the dac has to do with the timing, but I am afraid I am lost. Could someone point me to a primer? I am really trying to understand this circuit and do not wish to just "slap parts" together.

Thanks
 

rfbrw

Member
2001-10-26 11:51 pm
.
The PCM56 is a mono dac and as such left and right data is expected to be in parallel with a falling edge at LE triggering conversion.
The CS8412 is a digital audio input receiver and the data is output in serial form with left or right data signified by FSYNC being high or low, depending on the chosen format.
Without the inverter you would have either left channel only or right channel only data being converted.
 

BeerCan

Member
2005-06-14 8:18 pm
US
I see now I think. I read that in the data sheet but I did not get it. Now I see ( I think) One pcm56 needs to have its le set high and one set low otherwise I will just get single channel output. Basically data with the most significant bit first is RC and data with most significant bit last is LC.

Now as I see it wont that inverter introduce a delay in that one side? I realize it would be a small delay and probably academic, but it will be there right? If the delay is significant how would one compensate for it? I guess the obvious answer is to introduce a delay to the other dac.

Thanks again
 

rfbrw

Member
2001-10-26 11:51 pm
.
BeerCan said:
I see now I think. I read that in the data sheet but I did not get it. Now I see ( I think) One pcm56 needs to have its le set high and one set low otherwise I will just get single channel output. Basically data with the most significant bit first is RC and data with most significant bit last is LC.

No. Study the CS8412 and PCM56 datasheets.


Now as I see it wont that inverter introduce a delay in that one side? I realize it would be a small delay and probably academic, but it will be there right? If the delay is significant how would one compensate for it? I guess the obvious answer is to introduce a delay to the other dac.

Thanks again

The use of an inverter introduces a 22.7us delay but it doesn't seem to bother most and even if it does it can be removed with varying degrees of complexity
 
is not good circuit because have phase-shifting between channels

Perhaps you could explain why? Looks a lot like the usual issue of a single sample delay between channels (an argument that was done to death in the 80's, 8mm of air is the clue here) but that is not the same thing as a phase shift. Of course the PCM56 is a mid 80's design.

The reset of the designs are astounding. I am not sure I have ever seen such heroic efforts to cure a problem of one's own making.
 

ALP

Member
2004-06-07 4:32 pm
RU
Francis_Vaughan said:


Perhaps you could explain why?


"Orange book" standart. DAC`s both channels will be latched simultaneously.


Looks a lot like the usual issue of a single sample delay between channels (an argument that was done to death in the 80's, 8mm of air is the clue here) but that is not the same thing as a phase shift.[/QUOTE]


please, playback via http://www.y-min.or.jp/~nob/Audio/DAC/DAC03.html (or AudioNote x.1 DAC series) white (pink) noise mono phonogramm (you can find it on many test-CD`s) and you hear interesting effect :)

AP
 
"Orange book" standart. DAC`s both channels will be latched simultaneously.

Which seems to be saying what I thought. You get a one sample delay (actually to be pedantic the latching time is not the problem, but many DACs make this simultaneous with the output - which is what is in question.)

You could fix it with a simple 16 sample shift register. However it is NOT a phase shift. It is a constant delay of one sample period. This has exactly the same effect as an additional 8mm of air.

The problem with ascribing an "interesting effect" to this design is that you have not eliminated all the other vagaries of the design, and anything from Audio Note is going to be replete with those. You need to explain what a constant one sample delay in one channel will do.

The first generation CD players all had exactly the same delay - however as a result of using a single DAC and a pair of sample hold stages. Have a read of the PCM56 data sheet, it provides the schematic in all its gory glory. This goes all the way back to the CDP101.
 

ThorstenL

Previously: Kuei Yang Wang
Re: My 2x (!) DAC on PCM56 w\out digital filter

Konnichichiwa,

ALP said:
http://www.y-min.or.jp/~nob/Audio/DAC/DAC03.html - is not good circuit because have phase-shifting between channels :(

Hmm, it IS true that the DAC as shown above has a timeshift between the two channels of 22.7uS. However, is this really THAT BAD? In 22.7uS sound travels around 8mm.

Once you have made sure to position accuraltly equal to within 8mm's and you have found a way to lock your head in place with a similar accuract, it may be time to worry about the timing difference between channels, untill then I find they make no perciptible difference.

What does make one however is the "anti-sinc" filter I have been suggesting for a good while.

Sayonara
 

ALP

Member
2004-06-07 4:32 pm
RU
Hi!

However, is this really THAT BAD? In 22.7uS sound travels around 8mm

It`s BIG problem, because phase shift is NOT constant:

100 Hz - 0.4 degree
1kHz - 4 degree
10 kHz - 40 degree
20 kHz - 80 degree

This equvivalent - membrane of your tweeter\midrange of right channel "oscillated" about 8mm.

And you hear this. Try to play mono wideband noise... :)
You hear - sound over 5-6 kHz not localized in center - come with right channel.

This experiment I spend w\out any filters\amps\etc - just headphones directly connected to Iout of DACs.

That`s right. Peter Quortrup don`t like me... :))))))))))


...seems you have an interesting website but most pictures do not show up and it is in russian...

I`am in migrate from another hosting after publicate Azur640 circuit... Await some time.


You could fix it with a simple 16 sample shift register. However it is NOT a phase shift. It is a constant delay of one sample period.

I use "break-clock" for loading data in left&right PCM56.

The first generation CD players all had exactly the same delay - however as a result of using a single DAC and a pair of sample hold stages. Have a read of the PCM56 data sheet, it provides the schematic in all its gory glory. This goes all the way back to the CDP101.

CD of first generation use sample&hold circuit and one mono-DAC.
This not my way. ;)

I`m soldering some copies of AN DAC`s (on the full copied AN PCB). Don`t like...

AP


AP
 

ThorstenL

Previously: Kuei Yang Wang
Konnichiwa,

ALP said:
It`s BIG problem, because phase shift is NOT constant:

100 Hz - 0.4 degree
1kHz - 4 degree
10 kHz - 40 degree
20 kHz - 80 degree

This equvivalent - membrane of your tweeter\midrange of right channel "oscillated" about 8mm.

Nope, it is equivalent to displacing the whole speaker by 8mm, or your head, no oscillation.

ALP said:
And you hear this. Try to play mono wideband noise... :)
You hear - sound over 5-6 kHz not localized in center - come with right channel.

I used the "stopped clock" operation with PCM63. I could not hear any problems with imaging, but the extra logic certainly made the music worse.

ALP said:
This experiment I spend w\out any filters\amps\etc - just headphones directly connected to Iout of DACs.

With headphones (no distance to sound source) this may indeed become audible, however the source may be elsewhere still (quite easily ones own hearing is imbalanced if over 30....).

Sayonara
 
This equvivalent - membrane of your tweeter\midrange of right channel "oscillated" about 8mm.

Again, to join the chorus. No it isn't. It is hard to make this more plain. It is a simple
time delay.

Sure you can tranform this into a change of phase delay at a given frequency, but that is simply moving the delay to a different representation space. A cute mathematical reality, but otherwise meaningless. Further, the ear is insensitive to this, it is actually unable to resolve such a small time change. So even in headphones it won't be audible.

My point about the CDP101 and others is that exactly the same delay is apparent. It comes about through essentially the same mechanism. Further, it would be very interesting to see which other modern stereo DACs actually exhibit the same issue, hidden within their inner workings.

There are so many other issues in making a good DAC, but this one is totally insignificant. Compromising any part of the design in order to correct for it is a bad compromise. The total lack of PSRR in all the logic gates in the way, and thus the propensity to introduce significant signal correlated jitter into the clock will cause a vastly more important degradation in sound.
 

ALP

Member
2004-06-07 4:32 pm
RU
Hi!

A different quantity of the counting is used for the transfer of different frequencies in standard CD- audio: for 1 kHz - their 44, for 10 kHz - only 4, 20 kHz - hardly it is more than 2. Therefore the delay of one of counting of 44 is imperceptible during the construction of analog signal in 1 kHz, and here 1 of 4 on 10 kHz - it is essential. If we reproduce the signal of one frequency, then phase shift would be actually imperceptible. But real sound signal has wide spectrum. I note that 8 mm - distance _ considerably _ exceeding the mobility of the membrane of tweeter.
I is too lazy in order to conduct prolonged controversy about this fact. Gather on model pay one and different versions and compare. This is question one in the evening.

AP
 
hi,AP

the delay-time and the phase-shift is the same thing.....I can not say more about it in English.....but I think that you are on the wrong way, just like Zanden' s theory which saying that: the more worse,the more phase-shift, and 180 degree phase-shift cause the serious problem....oh,my god!

basing on my NOS-DAC experience,the best sound is from the steep LPF which has the most worse group-delay ;)

http://www.diyaudio.com/forums/showthread.php?s=&postid=672387&highlight=#post672387

X.G.
 
I is too lazy in order to conduct prolonged controversy about this fact.

Clearly.

Simply put, you have got it totally wrong. You are double compensating. The delay of a single sample period is the same period of time no matter what the frequency.

Simple question. Derive the frequency vs phase charateristics of an 8mm air delay. Is it the same as the delay in the digital domain as a one clock delay? If not, why not? If it is, does misplacing one speaker by 8mm cause the effects you claim? If not, why not.

You might like to try to dual of this problem. Derive the time delay at each frequency to effect a 180 degree phase shift. You might notice that you have a frequency dependent delay. This clearly must sound utterly dreadful. So why does it not?
 

ALP

Member
2004-06-07 4:32 pm
RU
Hi!

Are you read my posts correctly?
My DAC`s _DON`T HAVE_ "interesting" effects (phase-shift).

Nobo`s DAC (http://www.y-min.or.jp/~nob/Audio/DAC/DAC03.html), AN DAC`s series x.1 - _HAVE_.

I`am directly compare this - http://www.zeuslab.narod.ru/images/eXTigy/Cx2_2.jpg and this http://www.y-min.or.jp/~nob/Audio/DAC/DAC03.html.

http://www.y-min.or.jp/~nob/Audio/DAC/DAC03.html - this more simple way than my DAC. If this was valid solution, then I not would begin to make more hard solution.

Make both solutions and THAN describe differences. Don`t look at simple ways.

See orange book. Enough.

AP