Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Blaine78]

Since the .178 update,
J river has gone back to low volume output on JRSS 2ch downmix, whilst watching 5.1/6ch movies
I've been using it for couple of months with no problem, sound is now low i have to turn the hifi up-to near full to get good volume. i don't like using the 'normalize volume' as it changes volume too much after loud passages happen.
the volume seemed fine before this update

[mojave]

As Matt mentioned in this thread, it is always a balance between volume and clipping. The clipping was happening in the bass and mostly from 50Hz and down using the previous 2 channel downmix. If you have clip protection turned on in the DSP and didn't notice a problem before, you can use the preamp in the Equalizer DSP to turn up the gain. Mute the volume and change from "Clip protection" to "Flat line overflows" and monitor the peak levels while playing bass heavy 5.1 content. If the peak levels never go over 100%, then your gain increase is OK.

To give yourself more headroom, you can use Parametric EQ to add a a high pass filter at the -3dB point of your speakers (or even lower). This will filter out the lower bass that is causing clipping. If you don't want two EQ DSP's activated, you can also perform the gain increase in Parametric EQ by adding an "Adjust the volume" filter and setting it for the Left and Right speakers. 

[Matt]

Here's another snippet of a discussion about this issue:

2.0 downmixing was discussed here: http://yabb.jriver.com/interact/index.php?topic=64702.
The mix was then set to be a bit louder than a mathematically safe mix would be.

Do the recent problem reports mean that the redefined mix was too loud after all or did we have a new bug that actually more or less disabled the level normalization?

The bug was that normalization was not getting applied due to testing code that didn't get removed.

Now that it's fixed, we're normalizing in such a way that clipping is possible but rare.  An example normalization spectrum might look like this:

-12 dB: adjustment required to mathematically eliminate any chance of clipping (when channels are coherent)
-6 dB: adjustment required to preserve volume assuming incoherent channels
0 dB: the bug level, resulting in clipping; interestingly this seems to be the level that DirectSound mixing uses which confuses me

So now we're back around -6 dB on the spectrum above.  Obviously the actual value varies based on the level and number of channels used to build each channel.

The constant used to slide between the extremes was chosen by doing real-world tests.  Feedback is welcome.

[mojave]

Audioholics has an article called 0dBFS & Bass Management of DVD / Blu-ray Players which explains the issue (and includes links to other articles about the issue, too). They showed that the Oppo BDP-83SE was sending more that 0dBFS for both 5.1 and 7.1 with bass management on. I've often wondered how many receivers get it wrong. I emailed someone involved with Dolby about a year ago and was told that, at that time, receivers were still using the 5.1 bass management calculations for 7.1 bass management. This can result in clipping when using 7.1 and bass management.

[Blaine78]

Just a personal thing i have of increasing (normalizing) volume on top of another volume that has been decreased. just adds more processing in the audio chain, even though it is 64bit processing.
Guess i'm saying i've never noticed any overload of sound watching many dvds and blu rays.
maybe if the volume is midway between what it is now and how it was, would be perfect too.

[mojave]

Just as an experiment I used Parametric EQ to decrease the L/R volume by 75 dB. I then moved the L channel to User 1 and the right channel to User 2. I moved them back from User 1 to Left and User 2 to Right and added gain of 75 dB. While playing back music and turning the PEQ on or off it is completely transparent.

I also lowered the L/R volume by 50 db in PEQ and then increased my system volume all the way up. It still sounded exactly the same.

[Matt]

I'm not sure everyone appreciates how much precision 64-bit audio offers.

Below is a test that applies 100 million random volume changes (huge changes from -100 to 100 dB), and then applies the 100 million volume changes again in the opposite direction.

It is bit-perfect at 32-bit, which is the highest hardware output bitdepth I have seen (most high-end hardware is 24-bit).

In other words, 200 million huge volume changes still provides bit-perfect output in all real world conditions.

This also means one volume change or a series of 100 million volume changes that add up to the same net result is bit-identical.

Code: [Select]

// constants
const int nIterations = 100 * MILLION; // apply lots of volume changes
double dValue = 0.9323402123; // starting value (doesn't really matter what value we choose since it changes by huge random amounts)

// take a snapshot of the value at 32-bit (the highest used bitdepth of any hardware; most high-end hardware is 24-bit)
int nValue32Bit1 = CConvertFromFloatToInteger<32>::Convert(dValue);

// apply lots of random volume changes
JRDoubleArray aryDecibelChanges;
for (int i = 0; i < nIterations; i++)
{
double dDecibelsChange = (fabs(dValue) > 1.0) ? JRMath::GetRandomNumber(0.0f, -100.0f) : JRMath::GetRandomNumber(0.0f, 100.0f);
dValue *= CDecibels::GetMultiplierFromDecibels(dDecibelsChange);
aryDecibelChanges.Add(dDecibelsChange);
}

// apply inverse of volume changes
for (int i = 0; i < nIterations; i++)
{
double dDecibelsChange = -aryDecibelChanges[i];
dValue *= CDecibels::GetMultiplierFromDecibels(dDecibelsChange);
}

// take another snapshot of the value at 32-bit after all the volume changes
int nValue32Bit2 = CConvertFromFloatToInteger<32>::Convert(dValue);

// test for changes
int nDelta = nValue32Bit1 - nValue32Bit2;
ASSERT(nDelta == 0);


[Blaine78]

Just as an experiment I used Parametric EQ to decrease the L/R volume by 75 dB. I then moved the L channel to User 1 and the right channel to User 2. I moved them back from User 1 to Left and User 2 to Right and added gain of 75 dB. While playing back music and turning the PEQ on or off it is completely transparent.

I also lowered the L/R volume by 50 db in PEQ and then increased my system volume all the way up. It still sounded exactly the same.

it doesn't sound the same to me, it loses dynamics, maybe due to the lower output of the DAC to the pre amp. when you boost the volume (voltage) in the analogue domain, this is set at the same SPL by the way, things start to sound much punchier and alive. I don't want to get into the argument of what sound the same to one, and not to another. to me, reducing the volume at the source whether it digital or analogue, has an affect downstream. when i watch a DVD that i know well, the sound is now a tad thinner and less punch, plus the fact i have to increase the volume a lot and if forget to turn it down for 2 channel, nearly blow the speakers and upset neighbours). volume normalising is not my thing, i like it unchanged by automatic attenuation. i only put this point to you as there was nothing wrong with it before.

Don't mean to be out of line here, but changing the volume in the digital domain is not bit perfect. processing is being applied, the stream is being reconstructed. Is why i'll never use replygain, to my years it sounds tad flat.

[phusis]

it doesn't sound the same to me, it loses dynamics, maybe due to the lower output of the DAC to the pre amp. when you boost the volume (voltage) in the analogue domain, this is set at the same SPL by the way, things start to sound much punchier and alive. I don't want to get into the argument of what sound the same to one, and not to another. to me, reducing the volume at the source whether it digital or analogue, has an affect downstream. when i watch a DVD that i know well, the sound is now a tad thinner and less punch, plus the fact i have to increase the volume a lot and if forget to turn it down for 2 channel, nearly blow the speakers and upset neighbours). volume normalising is not my thing, i like it unchanged by automatic attenuation. i only put this point to you as there was nothing wrong with it before.

Don't mean to be out of line here, but changing the volume in the digital domain is not bit perfect. processing is being applied, the stream is being reconstructed. Is why i'll never use replygain, to my years it sounds tad flat.

I must back up Blaine78 on the issue of using Paramtric EQ, which does indeed affect audio quality. I know the numbers at play here are impressive(64-bit and all), but I rely on my ears(being a years-long highly (ear-)trained audiophile), and they clearly tell me otherwise - namely that EQ and other filtering processes later in the chain in many instances, even in high-bit domain and when in-effect, does change AQ to the worse. Relying on numbers only brings us so far - that is, mostly not in favor of or in any way sufficiently to account for audibility. The human ear/hearing is a much more sensitive and complex instrument than mere numbers would indicate, the incorrect correlation to which points to a limited approach in measurement and theorizing(which is no shame when looking at the complexity at hand), and not that we simply hear wrong. There's no need to state absolutely and in all cases that theories/numbers don't apply, but what's at play in this specific case does signify change to the worse, at least with parametric EQ as well as the annoyance of the too-long-in-effect volume leveling of clipping protection.

I'm sorry to have the effect of my initial inquiry on the reduction of output level lead to problems for others - here Blaine78 with his set-up. My speakers sport a very high efficiency, and so this puts me in a more favorable position. On the other hand, to address the highlighted part above, there were problems prior to the .178 build, that of bass clipping, and this has been accounted for. So what to do? Mustn't we strive to attain perfect playback without the use of after-filtering like Parametric EQ's and such? - I mean, no less would be expected of a high(est) quality playback application like J River, right? Of course, I do the easy part being no software technician and just pointing out where the problems arise, as do many many others in a much more elaborative fashion, but I hope you would at least try to trust and follow what very trained ears(with highly revealing set-ups) are reporting in cases like these.

Would there be ways for the users to manually select a variety of pre-set (source-)output levels in future builds when the issue of bass clipping has been solved, so to accommodate for the varying set-ups(with regard to sensitivity)?

[Matt]

JRSS downmixing always has a digital volume (called matrix normalization).

If you want to adjust that digital volume, you can use DSP Studio > Parametric Equalizer > Adjust volume.

[Blaine78]

thanks phusis for sharing. it's very hard to share these observations without looking like a loon
I completely understand the dilemma of downmixing a 5.1/7.1 stream to 2 ch, there will always be a trade off.
i'll just suck it up, and accept this is the way it has to be. to me it sounds better putting up with the lower volume, than add more processing on top of it. keep the stream as direct as possible is always the clear winner, even in digital.

[Alex B]

i'll just suck it up, and accept this is the way it has to be. to me it sounds better putting up with the lower volume, than add more processing on top of it. keep the stream as direct as possible is always the clear winner, even in digital.

MC just performs calculations to the numeric values. It doesn't matter if it first calculates a volume level reduction and gets the resulting values and after that makes new calculations and gets a new set of values when the precision of the internal DSP is hugely more accurate than the precision that is possible in the input and output formats.

I don't know if MC actually makes only one "real adjustment" to set the volume level after all factors are put into the equation or if it keeps intermediate resulting values in memory during the different stages. In any case that does not matter when both methods produce exactly the same resulting code after the floating point format is converted to the integer output format (as Matt explained earlier).

The downmix from 5.1 or 7.1 multichannel to 2.0 is a newly created audio stream. In this sense it is the original version and there is no exact reference volume level for it.

Ideally the 2.0 downmix should be audibly about as loud as the original multichannel sound track is when both are played through the same multichannel amp (without applying any processing on the amp), but that is not be possible if the individual channels in the multichannel stream are close to full scale (i.e. maximum possible level).

A mixer that causes audible clipping distortion with loud program material would be badly programmed and unacceptable.

The current compromise takes a small risk (it is louder than a theoretically safe mix should be), but with normal program material it should be fine.

plus the fact i have to increase the volume a lot and if forget to turn it down for 2 channel, nearly blow the speakers and upset neighbours).

This problem is not specific to 2.0. Those who have a multichannel system face it all the time. The movie soundtracks are often very dynamic and contain lots of relatively quiet dialogue. The heavily compressed victims of the so called loudness war (i.e. most of the current pop and rock music) have quite a bit louder average volume levels.

[Blaine78]

MC just performs calculations to the numeric values. It doesn't matter if it first calculates a volume level reduction and gets the resulting values and after that makes new calculations and gets a new set of values when the precision of the internal DSP is hugely more accurate than the precision that is possible in the input and output formats.

I don't know if MC actually makes only one "real adjustment" to set the volume level after all factors are put into the equation or if it keeps intermediate resulting values in memory during the different stages. In any case that does not matter when both methods produce exactly the same resulting code after the floating point format is converted to the integer output format (as Matt explained earlier).

The downmix from 5.1 or 7.1 multichannel to 2.0 is a newly created audio stream. In this sense it is the original version and there is no exact reference volume level for it.

Ideally the 2.0 downmix should be audibly about as loud as the original multichannel sound track is when both are played through the same multichannel amp (without applying any processing on the amp), but that is not be possible if the individual channels in the multichannel stream are close to full scale (i.e. maximum possible level).

A mixer that causes audible clipping distortion with loud program material would be badly programmed and unacceptable.

The current compromise takes a small risk (it is louder than a theoretically safe mix should be), but with normal program material it should be fine.

This problem is not specific to 2.0. Those who have a multichannel system face it all the time. The movie soundtracks are often very dynamic and contain lots of relatively quiet dialogue. The heavily compressed victims of the so called loudness war (i.e. most of the current pop and rock music) have quite a bit louder average volume levels.

Thanks for reply alex,
understand all above, i do a bit audio mastering, and music production and not unfamiliar with this territory. i guess what i meant by, adding processing on top of the low output of 2 ch, was the post normalising, which i leave off, to keep the stream as direct as possible.
As mentioned before, the low output is the trade off of downmixing to a 2 ch. Is a shame all blu rays and dvds don't come standard with a choice of 2.0 ch downmixed at the studio. always thankful for the handful that do.