logo

Google Search  |  Meridian Systems  |  Meridian Streaming  |  Restaurant  |  WiKi  |  Duncan's Meridian Info  |  Board Rules  |  Restaurant Rules
Previous Thread
Next Thread
Print Thread
Rating: 1
Page 4 of 23 1 2 3 4 5 6 … 22 23
Joined: May 2000
Posts: 8,364
Likes: 14
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: May 2000
Posts: 8,364
Likes: 14
Originally Posted by dr.bill
I clearly heard a difference in your tweak. As much so as putting the HD621 before the g68. I may also make various arguments that if your DSPs are not 8k's then you are not hearing the full range of what DSP could really do anyhow (and/or age/etc.). I guess this line of thought is not welcome so I'll stop and let's get back to regular 3db tweak programming smile

Brujoisms.

Not having 8k's may be entirely be true to limit audio differences I can hear in my system.

No idea how dune compares to Transporter and no idea if that is the reason for my results. But Transporter was designed by a genius with a lot of attention to jitter.

If I suddenly had to cut back on things the top two things that would remain are.

Transporter and DSP-6000's.


- Fiddler (IMHO)

If you think listening to music is fun, try learning an instrument.
HT: G61RSL, 4x5K, 5.5KHC, 2xSC6000, Transporter, HD621, Philips BDP7501 4K, Tivo, 65" LG OLED 4K C6
Computer: Foobar2000, 5K, Mojo, HiFiMan HE1000 V2
Joined: May 2000
Posts: 8,364
Likes: 14
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: May 2000
Posts: 8,364
Likes: 14
I just did another quick test to confirm something using TASCAM.

The DSP Presets adjustments for the rears is indeed done digitally. It has 30dB !! of "DIGITAL" range. That's 5bits of resolution. And even if a system is setup normally (calibrated at 0dB) I believe you can digitally clip them if you go over +3dB and you lose resolution as you go down from there. Not that anyone would be crazy enough use that much of the range.

And you add any DSP Speaker calibration offset below +3dB on top of that !!

I was kinda hoping Meridian might "Tell" DSP Speakers to lower themselves (when adjusting with Presets). But that does not appear to be the case.


- Fiddler (IMHO)

If you think listening to music is fun, try learning an instrument.
HT: G61RSL, 4x5K, 5.5KHC, 2xSC6000, Transporter, HD621, Philips BDP7501 4K, Tivo, 65" LG OLED 4K C6
Computer: Foobar2000, 5K, Mojo, HiFiMan HE1000 V2
Joined: Dec 2000
Posts: 866
Likes: 2
Pan-dimensional being
Pan-dimensional being
Offline
Joined: Dec 2000
Posts: 866
Likes: 2
Originally Posted by George Mills
I've been thinking hard about this and I think I might understand what we are dealing with.

<snip>

When you adjust volume digitally you are shifting down the MOST significant bits !! And pushing the least significant ones out (the ones Meridian added and is tossing as you lower volume digitally).

<snip rest of very informative post #115800>

We have come a long way from our initial (jumped to) conclusion: that there might be a bug in the G68 firmware. This conclusion was based on the report that the raise in quality was not gradually with the raise in level, but only worked on exactly +3dB.

With George's new 0.5dB difference test it became clear the effect might be gradually after all. This also means that you might indeed say that digital volume control comes at a cost (which George suggested earlier).

Your above post explores that further to explain where the cost factor lies: with the DAC. It got me thinking, and maybe the following might explain it a bit more. Most current DAC's are the delta/sigma type: this is a sort cross between 1-bit and multi-bit. Mostly they are around 4 to 6 bits and use oversampling to compensate for the lack of resolution (like SACD uses 64 times oversampling for its 1 bit scheme). So if you lower your volume with 1 bit (6dB), you actually reduce the resolution of the DAC to only 3 to 5 bits. It is clear that the effect of digital volume can heavily depend on the DAC used in the system. For instance there may also be a big difference between Mk1 DSP speakers (5K, 5K5 and 6K) and Mk2 and current DSP speakers. Did anybody try the +3dB mod with Mk1 DSP speakers? It's just a thought.

Joined: May 2000
Posts: 8,364
Likes: 14
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: May 2000
Posts: 8,364
Likes: 14
Originally Posted by evdberg
We have come a long way from our initial (jumped to) conclusion: that there might be a bug in the G68 firmware. This conclusion was based on the report that the raise in quality was not gradually with the raise in level, but only worked on exactly +3dB.

With George's new 0.5dB difference test it became clear the effect might be gradually after all. This also means that you might indeed say that digital volume control comes at a cost (which George suggested earlier).

Your above post explores that further to explain where the cost factor lies: with the DAC. It got me thinking, and maybe the following might explain it a bit more. Most current DAC's are the delta/sigma type: this is a sort cross between 1-bit and multi-bit. Mostly they are around 4 to 6 bits and use oversampling to compensate for the lack of resolution (like SACD uses 64 times oversampling for its 1 bit scheme). So if you lower your volume with 1 bit (6dB), you actually reduce the resolution of the DAC to only 3 to 5 bits. It is clear that the effect of digital volume can heavily depend on the DAC used in the system. For instance there may also be a big difference between Mk1 DSP speakers (5K, 5K5 and 6K) and Mk2 and current DSP speakers. Did anybody try the +3dB mod with Mk1 DSP speakers? It's just a thought.

Quote from Wiki "I don't think it gradually gets better as you approach +3.0dB. But I could be wrong on that." I always try to cover my a$$ then I followed it up with a test to confirm it.

I've thought of a dozen more tests I could do fully understand things. The rears one I just did was one of them. While my wife was out on a walk smile

But I'm confident +3dB is ideal whether it was gradual or a step. But I could be wrong on that too.

Your right that it looks less like a bug. But more like a design weakness (IMHO) in not taking full advantage of the DSP Speaker DAC. And it would be very easy to fix (or should I say improve the implementation, and to be honest I'm amazed they didn't).

Don't some Delta sigma's also come in 1-Bit? I'll have to have my local EE genius explain DAC's again to me smile

I'm glad your reading my posts in detail and understand the implications of the "progress" and I think we are in complete agreement.


- Fiddler (IMHO)

If you think listening to music is fun, try learning an instrument.
HT: G61RSL, 4x5K, 5.5KHC, 2xSC6000, Transporter, HD621, Philips BDP7501 4K, Tivo, 65" LG OLED 4K C6
Computer: Foobar2000, 5K, Mojo, HiFiMan HE1000 V2
Joined: Nov 2000
Posts: 2,766
Likes: 6
Robert.W
Great Green Arkleseizure
Robert.W
Great Green Arkleseizure
Joined: Nov 2000
Posts: 2,766
Likes: 6
George,

from your post:

"When you convert 16bit to 24bit you can't just "stretch" 16 bits into 24 or you'd push a bunch of significant resolution down to levels you won't hear.
16 bit and 24 but are on the SAME scale starting from the TOP bits. This is done for compatibility of 16, 20, 24 bit SPDIF. This is why low noise floor is critical to hear any improvement over 16bits. It's all dynamic range added to the botto
m."


Top bits a clear-its a practical reason for your D/A converter.
A digital full signal(all bits are set)have to correspond to a full analog singnal.
That means the output voltage at the D/A converter cant be higher with 24 bits as with 16 bits,
as then the following amplifier get overloaded.
And no bit set means (theoratically ) zero voltage from the D/A converter and zero voltage input to the following amplifier.
This is the minimum output from the D/A converter or the minimum input to the following input-amplifier.

The 16 bits A/D conversion from the record means that you have between the loudest signal is full range(all bits are set) and the zero bit(no bit set) 16bit steps or 65536 single steps.
When you want to convert this to 24bit the top bit is the same(all 24 bits are set) and the zero bit are the same (no of the 24bits are set).
And logically between this two conditions you have now 24 bit-steps or 16777216 single steps.

Thats the correct way to convert 16 bit words to 24bit words.
If you dont make that in that way,you dont have a linear signal
means full is full -half is half- zero is zero.

For example when only the MSB is set(all other are non set) in a 16bit word its the same analog output after the D/A converter as if you only have set the MSB in 24bit word-means half the full signal.
So booth(16 and 24 ) have the same span of the signal (full signal or nothing at all) BUT have different step size, means the 16 bit step is more coarse then a 24 bit step.

BUT booth are linear from zero to full signal !

So your statement:

"When Meridian says Resolution Enhancement (16 to 24bit) their may be now some new data in the low bits (from dithering and processing) but they are still VERY low level bits (most insignificant ones). The top 16 most significant should be more or less be the same."

couldnt be true.

What you describe here is leave the top 16 bits the same and only add artificial bits from dithering,processing in the lower 8 bit from a 24 bit word.
Sorry George, but that is complete unlogical for me as the signal then have on the top 16 coarse steps(the original 16bit from full signal to zero) and then you will add 8 bits more BELOW (the original 16bit) zero ?

Maybe I didnt catch your explaination, what do you mean?


Robert


800v3/218/271/8k.2/5k5s smile /5kC/DSP3200/Reviver/
TT-ELP/Graham Slee Revelation/Trinnov Altitude 32/OPPO 205
__Ayon/HB/Shunyata/Miltzow -- power cables,
__Ayon/Jungson/Miltzow --analog, Miltzow-- digital,
__Vortex/WMA ,
Joined: Sep 2007
Posts: 1,986
Likes: 55
Knows where his towel is
Knows where his towel is
Joined: Sep 2007
Posts: 1,986
Likes: 55
Originally Posted by robert.w
"When Meridian says Resolution Enhancement (16 to 24bit) their may be now some new data in the low bits (from dithering and processing) but they are still VERY low level bits (most insignificant ones). The top 16 most significant should be more or less be the same."

couldnt be true.

What you describe here is leave the top 16 bits the same and only add artificial bits from dithering,processing in the lower 8 bit from a 24 bit word.
Sorry George, but that is complete unlogical for me as the signal then have on the top 16 coarse steps(the original 16bit from full signal to zero) and then you will add 8 bits more BELOW (the original 16bit) zero ?

George is correct.

Resolution enhancement can be added in two ways;

1, Pad the least significant bits (LSB) with zeros (this is what SPDIF looks like anyway because it's always 24 bits)

2, Use DSP to re interpolate the LSBs, effectively smoothing the steps between them. (16 to 24 bit will add 256 extra steps)

Reducing the signal by 6dB is equivalent to reducing the word length by 1bit, BUT the reality is that the signal is still 24 bits and any further DSP is done with 24 bits of accuracy (or more depending on the DSP engine)

The problem here is trying to understand digital signals, bits and dBs. You can't take away 0.5 bits, but you can take away 0.5dB.

My thinking is that the +3dB and -3dB settings allow the original signal to be bit perfect in the MSB due to the -3dB processor offset, in effect making the signal 0dB or -6dB and -6dB is very easy maths, the whole word can be shifted 1bit towards the LSB.
For signals up to 23bits, no actual information would be lost, for 16bits there will be no loss at all.

Steve.

Joined: Feb 2002
Posts: 8,816
Likes: 87
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: Feb 2002
Posts: 8,816
Likes: 87
All very interesting to get into the bits and bytes theory. The empirical question is:-

Why does it sound different - at normalised volumes ?

We may all spend more time listening to establish whether it is a 'better different' or just 'different'. I'm going to reserve judgement for now as best I can.


Rgds,
Nick


Joined: May 2000
Posts: 8,364
Likes: 14
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: May 2000
Posts: 8,364
Likes: 14
Originally Posted by Steve Eyles
George is correct.

Resolution enhancement can be added in two ways;

1, Pad the least significant bits (LSB) with zeros (this is what SPDIF looks like anyway because it's always 24 bits)

2, Use DSP to re interpolate the LSBs, effectively smoothing the steps between them. (16 to 24 bit will add 256 extra steps)

Reducing the signal by 6dB is equivalent to reducing the word length by 1bit, BUT the reality is that the signal is still 24 bits and any further DSP is done with 24 bits of accuracy (or more depending on the DSP engine)

The problem here is trying to understand digital signals, bits and dBs. You can't take away 0.5 bits, but you can take away 0.5dB.

My thinking is that the +3dB and -3dB settings allow the original signal to be bit perfect in the MSB due to the -3dB processor offset, in effect making the signal 0dB or -6dB and -6dB is very easy maths, the whole word can be shifted 1bit towards the LSB.
For signals up to 23bits, no actual information would be lost, for 16bits there will be no loss at all.

Steve.

Thanks Steve.

You can do 0.5bits though. The 24bit value is really 48bits internally (floating point I believe) they subtract 3dB (or even 1dB, what ever) and dither it. I don't know if dither is done before or after it's converted back to integer.

This is why I did the +2.5dB test. The math is fine (no audible harm).

It's the loss of MSB bit(s) that I believe is causing the difference.

I thought doing an "even" bit would help too. Which it partly why I used to run at -3dB. But 0dB (in MConfig) was better (and some folks were indeed doing that). But +3dB is optimal. I always knew +3dB was the real 0, but didn't know it was reachable.



- Fiddler (IMHO)

If you think listening to music is fun, try learning an instrument.
HT: G61RSL, 4x5K, 5.5KHC, 2xSC6000, Transporter, HD621, Philips BDP7501 4K, Tivo, 65" LG OLED 4K C6
Computer: Foobar2000, 5K, Mojo, HiFiMan HE1000 V2
Joined: Feb 2002
Posts: 8,816
Likes: 87
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: Feb 2002
Posts: 8,816
Likes: 87
I wonder if there is a relationship to this 'tweak' and whether you get more/less benefit when it is also apodized?

e.g.

If the tweak does yield slightly more detail, air, transparency (put whatever words you like here) the would it also make the pre-ringing fix that the filters do more easy to spot ?


Rgds,
Nick


Joined: Dec 2000
Posts: 866
Likes: 2
Pan-dimensional being
Pan-dimensional being
Offline
Joined: Dec 2000
Posts: 866
Likes: 2
Originally Posted by ncpl
All very interesting to get into the bits and bytes theory. The empirical question is:-

Why does it sound different - at normalised volumes ?
We just tried to explain the why with the bits and bytes ...

@George, Meridian is not using floating point, they have always claimed to do everything integer, mostly using 48bits (2 registers) or even 72bits in the 518.

Joined: Sep 2007
Posts: 1,986
Likes: 55
Knows where his towel is
Knows where his towel is
Joined: Sep 2007
Posts: 1,986
Likes: 55
Originally Posted by George Mills
You can do 0.5bits though. The 24bit value is really 48bits internally (floating point I believe) they subtract 3dB (or even 1dB, what ever) and dither it. I don't know if dither is done before or after it's converted back to integer.

I have to disagree, the notion of half a bit is impossible.
1 bit is equivalent to 6dB (or x 2), 3dB isn't equivalent to 0.5 bits though.

If you had a full scale 24 bit signal;

FFFFF (Hex) = 16777215 (Dec) = 111111111111111111111111 (Bin)

-0.5dB is the same as x 0.944, so applying this to our full scale signal;

F1A9FB = 15837691 = 111100011010100111111011

The MSB is still used.

-3dB is the same as x 0.708, so applying this to our full scale signal;

B53F7C = 11878268 = 101101010011111101111100

The MSB is still used.

-6dB is the same as x 0.5, so applying this to our full scale signal;

7FFFFF = 8388607 = 011111111111111111111111

The MSB is lost, and as 011111111111111111111111 = 11111111111111111111111, it can be seen that the equivalent of 1 bit has been lost, or 24 bits has been reduced to 23 bits.

I'm sure you know this, but thought it might help explain it to those who didn't know.

Steve.


Joined: Feb 2002
Posts: 8,816
Likes: 87
President of the Imperial Galactic Government
President of the Imperial Galactic Government
Joined: Feb 2002
Posts: 8,816
Likes: 87
Hi Evdberg,

I'm not sure I agree. There's a disagreement above on what bits may/may not be being manipulated. HOW does this impact the sound (if it does) in the way we might perceive?

In other words, if we are talking about 23bit vs 24bit, then it seems to be a fairly significant impact on the presentation of the sound. Is that possible ?

Put another way around (and with respect too) I never found your 24bit conversion program from 16bit ever made this type of difference ? The logic cannot be so remote ?

thoughts ?


Rgds,
Nick


Page 4 of 23 1 2 3 4 5 6 … 22 23

Moderated by  Carl, Duncs, ncpl 

Link Copied to Clipboard
Who's Online Now
3 members (JCD, RobertW, IanM), 736 guests, and 2 robots.
Key: Admin, Global Mod, Mod
Newest Members
Kiran1103, Tony999, PhilP, Grover, Humberto
5,503 Registered Users
Top Posters(30 Days)
HugoF 10
Top Posters
VirusKiller 12,733
ncpl 8,816
Carl 8,506
Fiddler 8,364
Ian 7,950
Forum Statistics
Forums18
Topics31,253
Posts297,452
Members5,503
Most Online4,788
May 22nd, 2026
Meridian  |  Media Centre  |  Support  |  Firmware Release Notes  |  RSS Systems  |  RSS Streaming  |  RSS Restaurant

Website Provided by Mr Tech Guy - Resolute Audio Visual Limited © 2026

Powered by UBB.threads™ PHP Forum Software 8.0.1
(Release build 20251126)
Responsive Width:

PHP: 8.5.12 Page Time: 0.049s Queries: 39 (0.040s) Memory: 0.7113 MB (Peak: 0.8151 MB) Data Comp: Off Server Time: 2026-11-17 17:28:49 UTC
Valid HTML 5 and Valid CSS