I guess you mean active powered speakers (studio monitors including filter (dsp or analog) and amp)?
Sorry but active can mean so much things that sometimes it's confusing... Just to be sure i understand your point Youknowyou. 🙂
On mine i targeted 105dbspl peak at listening position. I'm probably lower as i did some modification to original filter and used an LT transform for med so probably lost 6db in the process. But i never listen at these level! The gain surplus is there for headroom and peace of mind... I usually consider past 90dbspl listening level to high, especially when working, i tend to listen at lower levels than for entertainment. But for longer periods of time.
Last edited:
this is a 2002 comment from Thorsten loesch, I wonder how still relevent or irrelevent it is today. from this explanation, it seems bit reduction is just one of the problem digital attenuation invariably creates.
<<Okay, let's do this slowly for those who are obviously technically excessively challenged.
A CD signal encodes music as 65536 discrete voltage values 44100 times per second. Of those 65536 Values halve cover negative voltage values and the other positive. The Absolute theoretically possible peak-peak dynamic range is 96db, once this is normalised back to "analogue" stahndards of measurements the lowest level encodable signal is a sinewave at -90dbfs peak or -87dbfs RMS. Below that there are no bits, so no signal (ignoring for a moment dither issues which give an apparent increase in dynamic range).
Anyway, modern DAC's often boast supposedly "more" bits than the classic 16 Bits required for CD. The veracity of such claims can be established by looking at the THD/N ratio, in order to give a real 18db resolution it should be > -99db (RMS Measurement), in order to give a real 20db resolution it should be > -111db (RMS Measurement), in order to give a real 22db resolution it should be > -123db (RMS Measurement).
Now, if we digitally reduce the the signal level in the oversampling Filter or DSP processor (the way it is generally done for "digital volume controls) it means we alter the Data in such a way that the all levels are shifted down. If for example we attenuate the signal by 6db it means that Data that was previously occupying the MSB (most significant bit) is now assigned to the Bit below that. It also means that the top 6db of the dynamic range of the Converters is effectively "thrown away".
Obviously for intermediate attenuation levels the reduction is corosponding, but taking a "normal" CD-Player or DAC of today, you rarely find anything offering -99dbfs THD&N for full scale and even less often anything offering -111dbfs THD&N for full scale WITH signals having a greater wordlength than 16 Bit (16-Bit words are ALLWAYS limited to the 87db RMS or 96db Peak-Peak Dynamic range, no matter what).
So, if the attenuation reaches 18db for our postulated -99dbfs THD&N DAC with digital Volume control we have effectiffly "dropped a bit" from our resolution, as now our dynamic range has been reduced by 6db. So the output signal will now only the equivalent of 15 Bit resolution.
Of course, non of this takes into account jitter, the various transient distortion phenomenae in digital filters, the requirement of actually having to fully decode the increased wordlength of oversampled data corretcly to avoid the loss of information and rounding errors in the DSP processing for the digital Volume control etc. These ignored factors make the subject of "dropping bits" actually trivial, compared to the other issues and lead to the fact that competently implemented digital volume controls that offer good sound are rare, to say the least.
Even the arguably excellent WADIA Processors are reconed to sound better with an external analogue volume control of suitably high quality.>>
<<Okay, let's do this slowly for those who are obviously technically excessively challenged.
A CD signal encodes music as 65536 discrete voltage values 44100 times per second. Of those 65536 Values halve cover negative voltage values and the other positive. The Absolute theoretically possible peak-peak dynamic range is 96db, once this is normalised back to "analogue" stahndards of measurements the lowest level encodable signal is a sinewave at -90dbfs peak or -87dbfs RMS. Below that there are no bits, so no signal (ignoring for a moment dither issues which give an apparent increase in dynamic range).
Anyway, modern DAC's often boast supposedly "more" bits than the classic 16 Bits required for CD. The veracity of such claims can be established by looking at the THD/N ratio, in order to give a real 18db resolution it should be > -99db (RMS Measurement), in order to give a real 20db resolution it should be > -111db (RMS Measurement), in order to give a real 22db resolution it should be > -123db (RMS Measurement).
Now, if we digitally reduce the the signal level in the oversampling Filter or DSP processor (the way it is generally done for "digital volume controls) it means we alter the Data in such a way that the all levels are shifted down. If for example we attenuate the signal by 6db it means that Data that was previously occupying the MSB (most significant bit) is now assigned to the Bit below that. It also means that the top 6db of the dynamic range of the Converters is effectively "thrown away".
Obviously for intermediate attenuation levels the reduction is corosponding, but taking a "normal" CD-Player or DAC of today, you rarely find anything offering -99dbfs THD&N for full scale and even less often anything offering -111dbfs THD&N for full scale WITH signals having a greater wordlength than 16 Bit (16-Bit words are ALLWAYS limited to the 87db RMS or 96db Peak-Peak Dynamic range, no matter what).
So, if the attenuation reaches 18db for our postulated -99dbfs THD&N DAC with digital Volume control we have effectiffly "dropped a bit" from our resolution, as now our dynamic range has been reduced by 6db. So the output signal will now only the equivalent of 15 Bit resolution.
Of course, non of this takes into account jitter, the various transient distortion phenomenae in digital filters, the requirement of actually having to fully decode the increased wordlength of oversampled data corretcly to avoid the loss of information and rounding errors in the DSP processing for the digital Volume control etc. These ignored factors make the subject of "dropping bits" actually trivial, compared to the other issues and lead to the fact that competently implemented digital volume controls that offer good sound are rare, to say the least.
Even the arguably excellent WADIA Processors are reconed to sound better with an external analogue volume control of suitably high quality.>>
That's not how I read that piece. It's actually mostly in agreement to what we have discussed so far, with a lot of other points brought up here in this thread. My own DAC has been tested by third parties and should be good up to 20 bit of resolution. I'm sure many more are available with higher resolution than true 16 bit today.
The last part makes it clear that Thorsten Loesch was not a fan of digital in 2002. Many today are not a fan of digital. Should we all start to doubt digital because of that?
The part about rounding errors etc. is why in JRiver 64 bit space is used internally.
Care to explain a bit more about your perceived degradation when using internal volume in JRiver? What driver, which DAC, when you did the test, did you control the knobs yourself?
I know I don't/cannot trust myself that way. I remember making a quick cut in EQ and go back to listening, yeah, sounds much better. After a while I go back behind the monitor only to find out the check mark was still off on that part of the DSP engine. So in my head it had worked wonders, cutting like I did. But I actually hadn't done anything. True story, proved to me the power of my mind is stronger than I'd like it to be.
So for me, I believe you convinced yourself, no doubt there, but you need to bring more to the table to convince me too. Not because I don't trust or believe you, just because you're human like I am. 😉
No foul intentions here. Just curiosity.
I'm certain Thorsten's mathematical errors and misconceptions remain as inaccurate now as they were in 2002. Martin Mallinson may be a more helpful introduction for those new to the topic.
Couple of watts with the beasties you had!
Here is a chance for many to get confused. FS is a peak level. I certainly do if Mr beer has come to visit! Doesn't help that there are two definitions of 0dBFS. But hey what's 3dB amongst friends.
Bill i don't agree about that. DBFS mean db full scale: this is the actual digital limit of your converter, it as nothing to do about peak value this is a defined value (as spl is a defined value : 94dbspl:1Pa). When you reach o dbfs this is the end of the digital scale. Often this is bad understood because in analog we think the other way around: you start with noise floor and then you go upscale, in digital this is the opposite: you start at a maximum and the dynamic range expand downward. In practice AES define dbfs from an rms value.
When Pano say -14db fs this mean what it mean: 14db crest factor allowed before digital clipping.
As you noted this is not a 'standard' value choosen by Pano. Maybe he never listen to classical non processed music?... but does that exist anymore?! 🙂
If you want the AES definition:
https://www.ak.tu-berlin.de/fileadmin/a0135/Unterrichtsmaterial/KT-Labor_WS0809/1_ADDA/aes17.pdf
Twest820: thank you this is interesting (i did stop before jitter). What i find weird is that in 2011 he still talk about 16 bit dacs... One thing Thorsten loesch is right about is that every digital treatment increase wordlength of original digital signal. This may be an issue if not dithered properly. This is one of the reason state of the art digital processor are such pricey item.
Weiss audio eq and compressor are great example. Theyr converters are not bad too... Swiss Technology! But do not look at retails prices... last time i did i nearly had an heart attack. 😀
Last edited:
Something just hit me in the face after watching the video on link: i should use 650 z input amplifier and same pad value for 135db snr!
WTF! I'll have to ripp off the white cathode output stage of a La2a Teletronix compressor (circa 1968) to drive the amp as a buffer between my all digital chain!
I was trying to skip capacitors from signal path and now i shoud bring them back plus tube and transformer including nfb?!!! Will have to use shunt regulator to keep snr low enough!!!! 🙂
WTF! I'll have to ripp off the white cathode output stage of a La2a Teletronix compressor (circa 1968) to drive the amp as a buffer between my all digital chain!
I was trying to skip capacitors from signal path and now i shoud bring them back plus tube and transformer including nfb?!!! Will have to use shunt regulator to keep snr low enough!!!! 🙂
NO. They define from a sine wave whose PEAK is FS. There is another definition for a square wave with a peak of FS. It's in the link you posted. It is also for a single frequency. My miniDSP SE outputs are 2V RMS. So an AES 0dBFS sine wave will show as a 2V RMS wave on my scope? Peak will be 1.414 times that?
I find your use of crest factor slightly confusing to let's see if we can get on the same page. A sine wave has a crest factor of 3dB. A square wave 0dB, normal music 12-14dB. So if you are playing music that is hitting FS, then the AVERAGE level is -14dB. Is that what you meant?
Pink noise, lets say has a 10dB Crest factor (it varies). The SPL meter measures RMS level and unless you are lucky not true RMS. But with a good meter pink noise at 85dB SPL will have peaks 10dB above that? Am I still on your page?
So Pano said '-14dBFS RMS pink noise' As AES defines FS as a peak not an RMS value what does it mean? For those of us who think in analog we could be 3dB out one way or the other! Everything is RMS except the hard end stop in digital!
It should show at 2Vrms on your scope yes. Peak it should be aproximately 2.7v rms.
Why do you mind about peak? This is just another way to quote the same value (without integration).
Yes it is. Crest factor is the allowed dynamic range above rms level.
Pink noise crest factor is 6 db. You can check this as 10db short time yes, but once integrate over sufficient period this is 6db. It's because of it is convenient to use (and because the filtering of high frequency is close to what our brain treat audio message).
Spl meter is defined ballistic (as vu meters are). You should have the choice between differents curve: A and C usually. C integrate the low end and should be used to calibrate your system. When you read 85dbspl on your meter using C curve, you are calibrated on rms value (6db crest factor is at work but integrated into the reading).
I repeat, aes define from an rms sine from which peak value blink an overload led. You should'nt worry about peak levels. We talk about rms value.
This peak thing is BS! It don't mean many things audio related : our brain is sensitive to rms value (it's why vu meter ballistic is usefull: 300ms in/300ms out), not to peak value (3ms in (but in fact it is variable between analog and digital one.../1 sec out). In fine peak is here just to check for the gear behavior. Once you have a system which is calibrated using rms value everything should be fine. Forget about that PPM peakmeter: you are a human not a converter! 😉
The one which put one on the big analog desk should be burned till the end of times!!! 😛
Last edited:
I think you put an extraneous rms in there, marked in bold! I don't mind about peak, but it is used in several ways.
. I found references that said it's adjustable on good pink noise sources!
We shouldn't worry, but we should be aware. For example THX specs are set for 85dB ref, 105dB peak. You cannot measure that peak as SPL meters don't work that way. But if someone says ' in row 7 a symphony orchestra peaks at 105dB on my SPL meter' and you want to be able to reproduce that in your living room (lucky batsrad), that isn't peak the way THX mean it and in fact you need a peak of up to 120dB to get the 105dB SPL.
Or to put it another way, my speakers were factory spec'd at 103dB SPL max at 4m. Horror, they are not THX compliant....but hang on 🙂
If you don't measure stuff every day it's easy to trip over bad use of peak.
It's because i'm in love with it! 🙂 No in fact it's because i have a scale of dbu: volt value in rms... and when i want to know a peak value i just read the value 3db up in the scale! 🙂
Maybe it should have an use. But for audio i don't see it, but i do not know everything. Maybe someone more techy than me could explain the use of it.
Well as said before spl meter do integrate so when you mesure 105db during live orchestra this is 105db integrated... What does it mean?... well as long as the violin (or the whole symphonic orchestra) won't produce pink noise it is difficult to know. 🙂
That said i'll be off topic a bit about the myth to be able to have an orchestra in your living room: THIS IS A MYTH!!! Not because of limits of your loudspeakers but because of acoustic of room! If you want to clearly reproduce the sound capted by the mic couple located above the head of the director, you should have a room with early reflection GREATER than the one present in room during recording. Given you probably don't live in an arena chance are poor that it is possible. An other way to have that is to have an RFZ. But this is not a living room anymore if you manage to do that at home. Calls it a home studio, or a studio home. Not sure about the waf! 🙂
Because of distance, etc,etc... But you know most studio monitors aren't able to produce taht kind of spl. At least not anymore (it was the case during 70's, 80's and 90's). Now it's nearfield 'of the day' (hype change every year or so about THE BRAND).
Oh yes! Even for dedicated technician.
Lottery home theatre purchase will be all ATC. They DO manage the SPLs. Pick you model based on 'how quickly would sir like to go deaf' 🙂
I think we are on the same page now. Sorry for the digression, but felt it was important to be crystal clear on a couple of things.
They are nice monitors. Especially for acoutic music. But i find them boring with other genres! And i lived for 2 years with this one:
http://atcloudspeakers.co.uk/wp-content/uploads/2012/06/ATC-SCM110ASL-Pro.png
All i can say is that for my taste the soft dome medium is not something i like. I've heard one system based on that one which i though was exceptional, designed (as the whole studio) by the previous owner of Boxer :
http://www.floatingpointaudio.com/images/karism-productions_karim-succar.jpg
http://www.akadesign.co.uk/images/projects/recording_mixing/KAS-1-lg.jpg
http://www.akadesign.co.uk/images/projects/recording_mixing/KAS-3-lg.jpg
Monitor is 15" tad, ATC medium, can't remember the tweeter ref. Active dsp multiamp. 🙂
I prefer compression drivers now, maybe i'm going deaf. 😛
It's alright. Like i said this PPM thing is not clear even for some technician. And i wasn't clear about that in the previous post.
Last edited:
Lived there for 2 years
And there for 7 years.
But big island is mo bettah. 🙂
You are a surfer? Or windsurf? That's huge in France.
Funny, when I lived in Montmartre (Place des Abbesses) people were jealous. The grass is always greener somewhere else, I suppose.
🙂 Bodyboarder.
But i would like to see a giant swell hitting Jaws once in my life!
I plan a trip to Nazare in Portugal to see giant wave at prai do norte next year.
Most convenient place in europe to see this as you are protected at the top of the cliff!
Ah Montmartre. C'est un endroit sympa c'est vrai. Tres romantique...si on oublie les parisiens! 😉
But i would like to see a giant swell hitting Jaws once in my life!
I plan a trip to Nazare in Portugal to see giant wave at prai do norte next year.
Most convenient place in europe to see this as you are protected at the top of the cliff!
Yes we have nice spot. Cold water but some nice places.
Ah Montmartre. C'est un endroit sympa c'est vrai. Tres romantique...si on oublie les parisiens! 😉
Last edited:
It's not the Parisians, it's the drunken Swedish and German tourists who sang every night below my window. 😡
The Full Scale thing leads to a lot of confusion, I know, just look at the thread linked in my sig line.
0dB FS is the maximum level you can get with digital. It's a peak level. Any pure sine wave will have to be at least 3dB below that, average. If the peaks are touching Full Scale 0dB, then the RMS value of that sine will be -3dB below the maximum peak level.
However, as Bill points out, there is another 0dB reference level. Same sine wave, same level, but we call it 0dB sine, because that's as loud as a sine wave can ever get in the digital signal. That is a valid reference. I am used to FS (full scale) as a reference, so I try to be careful to include the FS when taking about levels. A normal, nicely dynamic CD will have an average level of -18dB FS. Or -15dB below loudest sine. It can be confusing, until you understand what is going on.
0dB FS is the maximum level you can get with digital. It's a peak level. Any pure sine wave will have to be at least 3dB below that, average. If the peaks are touching Full Scale 0dB, then the RMS value of that sine will be -3dB below the maximum peak level.
However, as Bill points out, there is another 0dB reference level. Same sine wave, same level, but we call it 0dB sine, because that's as loud as a sine wave can ever get in the digital signal. That is a valid reference. I am used to FS (full scale) as a reference, so I try to be careful to include the FS when taking about levels. A normal, nicely dynamic CD will have an average level of -18dB FS. Or -15dB below loudest sine. It can be confusing, until you understand what is going on.
Your situation is unusual, at least for home music listening. For mastering, maybe not.
You need a large range of volume settings.
When I was running an active system, I often wished for 2 or 3 gain structure settings. Soft, Medium and Loud. Baby Bear, Mama Bear and Papa Bear. I never did build such a thing, but it would have been useful. In day to day listening, 10dB was all that was needed so I never bothered with the 3 tier approach. Max 10dB attenuation left me with at least 110dB of S/N and no measurable distortion.
- Status
- Not open for further replies.