ZP80 output specifications


Userlevel 2
Hi,

Does anyone know whether the digital output from the ZP80 is 16 or 24 bit? The main reason I want to know is I'm interested whether the digital volume attenuation would have any noticable effect on the sound quality, can anyone help?

Thanks...

Paul

15 replies

Userlevel 2
I've been wondering about this myself and spend many hours this weekend measuring and computing in various ways to determine if ZP80 digital output and volume control are 24-bit. The answer is, yes they are! This means that you can attenuate the volume down to as low as -48dB (the last mark on volume control) with no increase in distortion. Nada. It isn't bad below that either. I'd say if you plug into 24 bit DAC as I do, you can use the SONOS volume control with impunity. Because it's all 24-bit, here is no "information loss" for 16-bit music down to the last volume mark on the controller. I may post details later.
Userlevel 2
SONOS Volume control is 24bit, I think

There's been a lot of talk on this forum about how when you use the
SONOS "Volume Control" you loose "information", particularly if you
are using the digital outputs on a ZP-80.

I did two experiments which show that there is no "information loss" to
16-bit program material until the volume is lowered to below -48dB.
But I'll start by asking two questions, which it seems like my
experiment answers:

1) Does the SONOS volume control use 24-bit digital signal processing?
2) Are the digital outputs fully 24-bit, as allowed by SPDIF
format.

Those who know better than me, please correct me if I am wrong. For
others, I warn you, this does get a bit technical.

Ordinary 16-bit audio is two bytes of data for each sound sample.
24-bit audio is three bytes of data. When digital data is sent down a
digital SPDIF connection (or the optical variety, which is the same
format except for the light carrier) there are actually 24 bits (three
8-bit bytes) of data space allowed for each sound sample. That was
one good idea incorporated into the original "Sony/Phillips Digital
Interface". Not all devices that output SPDIF output the full 24
bits, nor do all devices that input SPDIF read them all. But it
doesn't cause any incompatibility. Those devices that only output
16-bits simply leave the lower byte as all zeros, and those that only
input 16-bits simply ignore the lower byte.

Now, if you are performing volume adjustment in the digital domain
you can choose to truncate the information at the 16-bit boundary, or
you can start shifting it over into the lowest byte. If you do the
latter, you retain 16-bits of "information" until you have reduced
the volume enough so that nothing can appear in the highest byte
anymore, so everything is now in the two lowest bytes. Each byte has
8-bits, which corresponds to a dynamic range of 48dB. So you only
begin losing information to less than 16 bits at a volume level below
-48dB. Now you can see the relevance of the two questions above.

To test, I played a song through SONOS, and recorded the SPDIF output
of my ZP80 on a 24-bit Alesis Masterlink, through it's SPDIF input.
In this application the Masterlink is simply recording the digital
data on its harddrive. Of course, the Masterlink is a full 24-bit
unit and reads the whole 24-bit SPDIF signal. I didn't do any DSP on
the masterlink to change anything. Then I wrote the data out as
24-bit AIFF files on a CDR (what Masterlink calls CD24 format). I
then read those files into a Macintosh, where I hacked a C program so
it would print out the 3 bytes of data for each sample in hex.

As expected, when the ZP80 was set at "maximum volume" I only saw the
16 bit data in the two top bytes, and the lowest byte was all zeros.
The ZP80 was simply copying the 16-bit song file (an AIFF format file)
to its output with no dithering or other processing. But when the
ZP80 was set to -1dB (the first volume down setting), you could see
that the numbers in the top bytes were getting smaller, but now bits
were appearing in the lowest byte. This effect was magnified even
further at a volume level of -11dB (which is where the first "bar"
appears on the Sonos Controller).

But were the bits in the lowest byte the correct bits? The short
answer is yes. I may get to that in my next post with some sample
data.

I also confirmed these results by doing distortion measurements on the
output of my DAC, which conveniently has digital gain so I could keep
the output level constant down to -30dB. Distortion levels were
unchanged down to that point, which is the lowest I can measure
with the best accuracy. In fact, they showed very little increase
all the way down to -48dB, the last marker on the volume control,
and even remained below levels considered "audible on music" down
to the very last volume tick.

So if you feed your SONOS digital signal to a 24-bit DAC or processor, there is no "information loss" using the SONOS volume control as your main volume control or "digital preamp".
So the question is then: How common are 24-bit DACs? I have a DPA little bit three, I can find nothing about it on-line.....
Userlevel 2
Pretty much all DAC's are 24-bit now. That's what chipmakers make now. 16-bit DAC's are "obsolete." However, you can still find such DAC's in "retro" type gear which use collectible historic chips from before the mid 1990's.

I'd be willing to be that ZP-80 itself has a 24-bit DAC, but I haven't measured it yet.

Note that when I say "24-bit" I mean that they read all 24-bits and attempt to recreate them as best they can. They actually only achieve 18-21 bit performance. Actual 24-bit performance is as yet unachievable. It represents the noise level in a 10 ohm resistor, -144dB, which is unachievable in analog equipment also.

I don't think recordings necessarily need 24-bits, but when you are doing things TO a 16-bit signal, such as lowering the level, raising it, or mixing 16-bit signals, it is useful to have the extra bits to be sure none of the 16-bit performance is lost. In fact, some DSP's have even more bits, such as 32, for this reason.

When lowering the volume with 24-bit digital signal processing, and when you get down below -48dB, you can hardly even hear the music anymore, let alone the distortion, which is still many times lower, so if you are no longer achieving 16-bit performance, it really doesn't matter. With 24-bits available, you can set your system up to have headroom above ordinary listening levels, and not lose your 16-bit performance at ordinary listening levels and even below that. Without the 24 bits, that "headroom" would be eating into your DAC performance.
Userlevel 2
This is huge.

Thanks.

I of course only get about 70% of what you've done, but appreciate 100%.

Is there a formula to explain (i) why -48db is the magic #, and (ii) how -48db translates to the leftmost volume "bar"?

Really looking forward to the next post on this with sample data. Find this fascinating. Also kind of wondering (in the sense that one wonders about the outcome of murder mysteries) how you're going to test if the ZP80 has a 24 bit DAC.

Now I just wish that someone could confirm the ZP80 could carry hi-res audio files... but I think that's unlikely as Sonos employees have already burst my bubble on this one. Though I'd be interested in your take on whether this is a technical no-go or just a commercial no-go...

Thanks again!

J
Userlevel 2
if the BOM posted somewhere else on the forums here is correct, the hardware should have this 24 bit DAC

http://www.cirrus.com/en/pubs/proDatasheet/CS42416_F1.pdf

24 bit flac support it could just be as simple as a software upgrade, (if the hardware assumption holds true) as they already feeding 20 or 24 bits worth of audio to the DAC

then again there are other things to factor in, like the other custom hardware bits...

and of course wireless bandwidth usage will go up a bit :^)
Userlevel 2

and of course wireless bandwidth usage will go up a bit :^)


Just out of curiosity - on average, how much bigger is a 20- or 24-bit FLAC file than a redbook FLAC file? Realize there may be no easy answer (eg, can imagine in my non-techie way that an individual piece of music would have lots of samples that look the same if recorded in 20-bit mode, but have micro-details that make them look quite different when recorded in 24-bit).

Very curious about this... would love to be able to use SONOS as my system going forward into the brave new world of hi-def audio (I never got on the HDCD or SACD or whatever format bandwagon with physical CDs). Multichannel a whole different beast again... but that's a different day.

J
Userlevel 2
Here is a sample of a 24-bit/48kHz flac file:

taping lineage:
Source: DPA4061's > Church Audio ST-9100 (~+10dB, no HPF) > M-Audio Microtrack 24/96 (1/4'' TRS, L, WAV 24bit/48kHz)

d1t03.flac = 79.7Mb (2,304 kbps)

Dithered down to 16bit/44.1kHz with dbPowerAmp reference edition:

d1t03.flac = 36.1Mb (706 kbps)

I think you would need to be wired for 24bit audio. I would very much like Sonos to add support for 24bit flac files and ReplayGain. Plus an automatic "Recently Added" category.
Userlevel 2
But were the bits in the lowest byte the correct bits? The short
answer is yes. I may get to that in my next post with some sample
data.


Would love to hear more about this, Charles... this seems to be a much better interpretation of the quality of the "digital pre-amp" than I had previously read on these boards (eg, that it would be accurate only to ~75% volume, or around the first "tick").

J
Userlevel 2
This is huge.
Thanks. I of course only get about 70% of what you've done, but appreciate 100%.

Is there a formula to explain (i) why -48db is the magic #, and (ii) how -48db translates to the leftmost volume "bar"?


-48dB is only the magic number because it happens to correspond to 8 bits, which is the difference between 24 and 16.

I personally believe that 16 bits are all that is needed in a good recording. Though that is a statement of opinion. Anyway, it is a defacto standard in marketing. Production has been using "24 bit" gear for years, typically 48khz/24 bit and now even 192Khz/24 bit, though pros typically go with 96Khz/24 bit. While production is done 24 bit or better, it is boiled down to 16bit with the signal maxed out pretty close to 0dB, so all the bits are being used fully. It's useful to have the extra bits in production even if no where else.

Whenever you are doing mixing or volume control it is useful to have extra bits. Since "24 bit" is the current professional standard, and "16 bit" is the current consumer standard, that's the important coversion. And the difference is 8 bits (24-16). Each bit is worth approximately 6dB, so 8 bits gives you approximately 48dB's (8*6). (Actually, 16 bits gives you close to 98.2dB, IIRC, so the number for 8 bits would be 49.1. But the 6dB/bit convention is very strong, and such numbers are often used as the "nominal" numbers, though they are just slightly off.

When you do volume control, you can't add precision, but you might lose some of the last bit of precision at non-even divisions. This is a tiny amount and is typically even ignored. If you stuck with even-only divisions you would be forced to attenuate in multiples of about 6dB, which would not be nice.

But starting from 16bit samples, if you are dividing within a 24-bit register to achieve volume reduction, you are still working with 16-bits or more until you have lost the topmost 8-bits.

And that happens to mean 48dB (nominal) of volume reduction in this case.

Sonos engineers appear to have lined the last red bar on the virtual volume control that appears on the display of the controller when adjusting the volume....with the magic 48dB point. Above that point, 16 bit data is not losing any precision. Below that point, it probably doesn't matter, because you can hardly hear it anyway. This may just be coincidence, or it may be clever design, I don't know.

Personally, I quibble with the rather large volume adjustments within each range. I would have preferred constant 0.5 dB which is pretty standard nowadays.

But anyway, I don't know when I'll post any data. Actually we could get into complicated arguments about that. It's not as simple as you might imagine for complicated reasons. But it's not conceptually hard, it's simply applying a division. So, when you divide 1 by 3, for example, you get 0.33333... etc., for as long as you have digits. So that effect occurs right away. Right away, volume reduction from 16 bits into 24 bits can (and usually does) fill up all the bits with a new pattern. Even a small volume reduction is allowing "more bits", but it doesn't actually have more information. But when the volume has dropped by 48dB, for sure the signal is limited to 16 bits, where it has no more than the original 98dB of S/N of what 16 bits are capable of ideally.

Good to hear about the DAC's. Probably all being "used" for the volume control and tone control (which is probabably also based on 24-bit or better DSP).

It would be nice to think fully handling 24bit program material would be merely a matter of programming. But anyway, not much equipment even lives up to anything like the 98dB potential of 16-bits. And now not much music gets below full range anyway, and it's all noise to begin with.

End for now.

Charles
Userlevel 2
Personally, I quibble with the rather large volume adjustments within each range. I would have preferred constant 0.5 dB which is pretty standard nowadays.


The Sonos UPnP volume APIs go from 0-100, and from what I remember the Sonos controller software moves in 5 unit steps.

So you can get better granularity...
Userlevel 2
Charles, did you ever get anywhere further with this project? A bunch of us were waiting with bated breath...

Thanks,
J
Userlevel 2
Great read and thank you for the work of all involved. I have used my Sonos to feed my DAC from the digital coax. I have been bypassing my pre amp to eliminate one item but there is significant noise when using the digital volume control below the "1/4" dash on the Sonos and when muted. Set as a line out into a pre amp disables the volume control but the sound is great. Even using the pre amp and using the sonos for some volume attenuation down to the "3/4" dash I cannot hear any of the artifacts. Anyone know why? Bit Chopping?
dcfis,
When you say "noise" do you mean "hiss"? Is the noise completely gone when you use the preamp, or just diminished? It sounds most likely to be an impedance mismatch. What is the input impedance of your amp? and the output impedance of your DAC? This may not be related to any "digital artifacts". If you are getting better results with your preamp, you may want to leave it in line.
Userlevel 2
Badge
Did anyone ever independently confirm that the digitial volume control is transaprent down to -48db?

Reply

    Cookie policy

    We use cookies to enhance and personalize your experience. If you accept you agree to our full cookie policy. Learn more about our cookies.

    Accept cookies Cookie settings