How Loud Do You Need?

[fas42]

Main thing is that the system can sustain any volume I choose - should be able to go to a level where it is impossible to converse with someone unless they're standing right next to me, when playing high density pop or rock. Note, this is just the volume that matches getting realistic sound from recordings of acoustic instruments with lots of light and shade. Overall, the actual volume used will depend on the mood, and situation.

 

[RayDunzl]

Main thing is that the system can sustain any volume I choose

Do you have one?

 

[oivavoi]

I don't understand the need for a system that can play at any volume. I never play louder than short peaks of 100 db. Why do I need a system that can go beyond 105? To do that would require me to invest in drivers/speakers that can take a lot of power (or have very high sensitivity), and corresponding amps. In short, fewer speakers to choose from, and probably more expensive.

I don't know if there's something to it or not, but Alan Shaw claims that his Harbeth speakers are made/tuned for being played at moderate volumes. He thinks that speakers can be designed and optimized for different listening levels. I'm not sure if I understand his logic, but Alan Shaw is a pretty smart guy. (and lots of people seem to love the Harbeths, so he can't be doing everything wrong)

 

[RayDunzl]

TV is playing:

63.5dB Lzeq / 80.8dBz peak for the last hour or so...

 

[Blumlein 88]

All you guys seem to be implying we can make do with less than speakers which reach 144 db cleanly. Don't we need that for our 24 bit recordings?

Besides, if you have never listened to such speakers how can you be sure?

 

[RayDunzl]

All you guys seem to be implying we can make do with less than speakers which reach 144 db cleanly. Don't we need that for our 24 bit recordings?

I gave up on that thought when I found the extra 8 bits (in 24) led up to the first of 16 bits instead of being added at the other end where they would make a difference.

I thought "Wow, we can record everything except a Saturn V at their natural levels" and adjust our amplified sound pressure levels in-room as we saw fit, never (in the normal case) running out of headroom in the recording.

But noooooooooo.

 

[fas42]

Do you have one?

Not at the moment. In the past, yes - the current NAD is still locked on a fixed volume setting, hasn't been touched for quite some time - but if I put a modern, squeezed to max level all the time, pop number on this setup now the density of sound makes it hard to talk.

Why you need a system that handles high volumes well is so that crescendos are handled cleanly - an orchestral recording is put on with a soft start, which builds to a "noisy" climax - the latter should be handled effortlessly, just like listening to a live version in a concert hall works. A lot of power is not needed, just a system that can hit high SPLs cleanly - I've done this regularly with very ordinary speakers, just a bit of extra overall refinement required.

No-one needs 144dB peak - just around 110 is plenty - a rock recording peaking at this will get your ears ringing in no time ...

 

[RayDunzl]

I'd like to see the Kii Three 1khz test tone results at a similar volume.

If the plot below is any indication, I would think they don't have the same "problem", certainly not at 1kHz.

After looking at my ('308) distortion again, I thought "Hmm, maybe I was clipping or some other unfair condition), but, watching the harmonics rise out of the noise floor with 1dB increases in tone level, I decided that wasn't (necessarily) the case.

 

[DonH56]

Do you have one?

One what?

 

[oivavoi]

All you guys seem to be implying we can make do with less than speakers which reach 144 db cleanly. Don't we need that for our 24 bit recordings?

Besides, if you have never listened to such speakers how can you be sure?

Given that I've never heard a system that can play 144 db cleanly, it would be indeed be terribly arrogant of me to have an opinion on the matter.

 

[DonH56]

The ~6N rule is for SNR; for individual tones it is ~9N so you really need 216 dB signal to peak distortion or noise tone.

And, good luck with that...

 

[Sal1950]

Back in my "rock concert in the living room" days I was driving my Klispch La Scala's with a Phase Linear 700B. That rig could go a bit loud.

 

[Blumlein 88]

The ~6N rule is for SNR; for individual tones it is ~9N so you really need 216 dB signal to peak distortion or noise tone.

And, good luck with that...

I am not following you here Don, could you elaborate?

 

[DonH56]

For a sampled system with discrete levels, e.g. an ADC or a DAC, theory predicts (and measurements confirm) an SNR of about 6N+1.8 dB for an N-bit converter. This is assuming a sinusoidal input at full-scale (not clipped) and noise from quantization noise only. The derivation is not too bad but I'd have to look it up (I have it done several ways but my references -- and notes -- are not with me at the moment). SNR takes the single signal tone and compares it to all the noise; think of an FFT plot with a signal tone and compare that to the sum (actually root-sum-square) of the all the noise floor components at the bottom (the "grass"). The answer would be different if the input were not a sine wave, BTW.

Now, remember the SNR includes all the noise, all the "grass" at the bottom of the plot. Clearly, hopefully, each individual strand of grass must be much lower than 6N, or the resulting RSS value would be much larger than that. It turns out, again theoretically but empirically as well, that each individual strand (noise frequency bin in the FFT), is actually about 9N down. This derivation is not so straightforward, involving Bessel functions and other high-level math that makes my head hurt. Yes, I've done it, but it is more painful for a hairy-knuckled pea-brained engineer like myself.

HTH - Don

Article on sampling: http://www.whatsbestforum.com/showthread.php?1209-Sampling-101

 

[RayDunzl]

REW has (in addition to all else) an SPL Logger function.

Here's listening to the radio. Some talk and (to me) obscure tunes- WMNF

SPL Data Logging
The Logger button opens the SPL Logger graph window. The record button in the top right corner of the SPL Logger graph starts or stops logging of SPL values. When logging is in progress the SPL meter on/off, calibrate, weighting, filter time constant and high pass filter buttons are disabled. The logger records the SPL (with the currently selected weighting and time constant), the minimum and maximum values, the unweighted and uncalibrated peak value, the equivalent sound level and the sound exposure level. A Save button above the SPL Logger graph allows the data recorded to be saved to a text file using the text delimiter set in the REW File menu, log files can be loaded using the Open button.

 

[Blumlein 88]

For a sampled system with discrete levels, e.g. an ADC or a DAC, theory predicts (and measurements confirm) an SNR of about 6N+1.8 dB for an N-bit converter. This is assuming a sinusoidal input at full-scale (not clipped) and noise from quantization noise only. The derivation is not too bad but I'd have to look it up (I have it done several ways but my references -- and notes -- are not with me at the moment). SNR takes the single signal tone and compares it to all the noise; think of an FFT plot with a signal tone and compare that to the sum (actually root-sum-square) of the all the noise floor components at the bottom (the "grass"). The answer would be different if the input were not a sine wave, BTW.

Now, remember the SNR includes all the noise, all the "grass" at the bottom of the plot. Clearly, hopefully, each individual strand of grass must be much lower than 6N, or the resulting RSS value would be much larger than that. It turns out, again theoretically but empirically as well, that each individual strand (noise frequency bin in the FFT), is actually about 9N down. This derivation is not so straightforward, involving Bessel functions and other high-level math that makes my head hurt. Yes, I've done it, but it is more painful for a hairy-knuckled pea-brained engineer like myself.

HTH - Don

Article on sampling: http://www.whatsbestforum.com/showthread.php?1209-Sampling-101

Thank you Don. I read your linked post which was nice. Still didn't quite get it. I then ran across this:

http://www.analog.com/media/en/training-seminars/tutorials/MT-003.pdf which I think got it across to me. The next one below helped too. I don't think Don needs this, but it might help someone else wondering about it like me.

http://www.analog.com/media/en/tech...namic-Range-in-Wideband-GSPS-ADCs-MS-2660.pdf

So basically if we looked at any one hertz and say we had a 1000 hz full scale tone, we would find any other one hertz point say 1500 hz would be at a level of 9N below full scale. We aren't RMS'ing all the other frequencies together. Keeping in mind your usually available FFT bins aren't one hertz exactly. And it ignores leakage of windowing in a real FFT. And one can see with larger FFTs that you do get close to this theoretical result. So have I mangled any of it too much?

I can't say I get why the number would be 9N, but accept that it is. And am I right this would only refer to SFDR in a perfect system?

 

[DonH56]

As I said the proof is a bit complicated. And yes all these equations apply to quantization noise only -- no other sources of noise or distortion, i.e. a "perfect" (but quantized) system. The bin size falls out when you run the math, but you have to use enough points to reach the penultimate level. I usually don't as you can see in the example figure attached (I am only around 130 dB SFDR) but hopefully it helps. For most all my simulations and tests I follow the IEEE Standard (1241, or see 1057 for the waveform digitizer standard that preceded it; most vendors use that or something similar) so I don't need to window the results. There is a simple equation to calculate the noise floor based on the number of FFT samples but I don't recall it off-hand (have not been doing data converters the past few years). That is not the same as the fundamental noise floor set by quantization.

 

[Blumlein 88]

As I said the proof is a bit complicated. And yes all these equations apply to quantization noise only -- no other sources of noise or distortion, i.e. a "perfect" (but quantized) system. The bin size falls out when you run the math, but you have to use enough points to reach the penultimate level. I usually don't as you can see in the example figure attached (I am only around 130 dB SFDR) but hopefully it helps. For most all my simulations and tests I follow the IEEE Standard (1241, or see 1057 for the waveform digitizer standard that preceded it; most vendors use that or something similar) so I don't need to window the results. There is a simple equation to calculate the noise floor based on the number of FFT samples but I don't recall it off-hand (have not been doing data converters the past few years). That is not the same as the fundamental noise floor set by quantization.

Yeah I understand the deal with FFT bins at least. 16K plots basically "see into" the RMS noise level 39 db and 64 K FFTs get us 45 db into it. So if I generate digitally a 441 hz sine wave at full scale the other bins in a 64K FFT show about 190 db down which is getting close to the 216 db from the 9N formula.

 

[DonH56]

The references I have are from a PhD dissertation (whilst I was at UCLA working on my master's and working full-time in an R&D department) and an internal paper. I don't have permission to publish either. This paper discusses the 9N SFDR derivation, however: http://essay.utwente.nl/61335/1/MSc_PI_Bicker.pdf It distinguishes differently when N is 4 or less, however, see reference [8] in that paper. I had Dr. Abidi for several classes at UCLA way back when so that may be the reference I had in mind; the other I found tonight does not have the difference around N=4 but is more complex to follow. With appropriate simplifying assumptions it may provide the same result; I am not sure (and am too tired to dig into it deeply tonight, sorry).

I can't recall if the derivation is in the IEEE Standard.

HTH - Don

 

[RayDunzl]

Go ahead and post it but redact the juicy parts.