In-Room Recording

[RayDunzl]

I've taken part of a track, and recorded it using my measurement mic. Played in stereo, recorded in mono.

Then, mixed the same track from CD into mono, and time-aligned it with the in-room recording.

Here is the zipped audacity project, 38meg:

https://www.dropbox.com/s/amz174w1l7mr3mr/Audacity Project for Live vs Recorded Compare.zip?dl=0

Open Audacity and open the project file.

Click Mute on the first track, and play.

Click Solo on the first track to seamlessly switch back and forth between the in-room recording and the original (but mixed to mono) file from CD.

I find them to be acceptably close. The second file may need to be dropped a couple of dB, not sure.

Try a listening compare if you like and give a silly comment.

 

[Blumlein 88]

I am sorry Ray. I mean it can probably be fixed up. I think we could do it for $200k. Maybe a little less. Just goes to show you what we can get used to if we don't know what the reference really is.

 

[Blumlein 88]

Now my prior post was in jest of course. So don't sweat it.

We need to get FAS42 to give his remote system critique. I said I would record something for him a week or two back, but haven't done so yet. Glad someone did.

 

[Sal1950]

OK, I'll bite.
I'm not sure how your determining track 1 from 2 but in my window the top file had a level about 3db to hot and I lowered it to get a subjectively equal level.
After that when clicking back and forth I'd be hard pressed to hear any differences.
Listening on my laptop and Senn HD650 cans.

Edit, believe to hear a sharper more detailed base line on track 2, but it's close.

 

[RayDunzl]

I'd be hard pressed to hear any differences.

Yes, I have the same problem - casually flipping back and forth, there's nothing jarring in the switch.

in my window the top file had a level about 3db to hot

Played back here, I thought the second one a little louder. I haven't tried the headphones yet (also 650's here).

Thank you for playing...

 

[fas42]

Ray, thanks very much for doing that ... I have a hate hate relationship with Dropbox, it always decides to be clever in such a way that I end up getting nowhere fast - its latest trick is pretending to download the file, without telling me where it's going to, and I end up with no actual file on my hard drive. I will have to waste some time sorting out what nonsense is going on now, unfortunately ...

 

[fas42]

Okay, solved the problem - typical, dopey programmers in the Internet era assume everyone has a setup that matches their own - which means that if one is using the 'wrong' browser then things ... just ... don't work! Switched from Firefox to Chrome, and calm is now restored ...

Again, Ray, thanks for going to all that effort, including the bundling it up into an Audacity project. Not my favoured player, so exported to distinct files, to use with Media Monkey.

On first listens, very impressive - everything in order ... the main distinguishing elements are some loss of sense of space, and transient punchiness. Considering the extra path travelled by the recording signal this is no big deal - a tick from me .

 

[fas42]

Ray, this was very brave of you, because it allows one to easily compare the before and after - which can be a scary thing! Naughty me, I've started using Audacity to see what the waveform is doing on its journey down that track - at times it is remarkably faithful, and at others you might be looking at a completely different piece of music! Most of the discrepancy appears to be in the bass, which would be expected - I stripped out everything below 400Hz, and then one gets a decent correlation - but, there are still some interesting anomalies ...

Perhaps you've already done some investigation along these lines yourself - I won't say more until you add something here ...

 

[RayDunzl]

Perhaps you've already done some investigation along these lines yourself

I've done a lot of stare and compare between what the mic picks up vs what was sent to the hardware to be output.

 

[fas42]

A pretty obvious thing to do, of course - one peculiarity is that the volume envelope seems to fluctuate, quite dramatically at times ... resonance, room echos, ... ???

 

[RayDunzl]

one peculiarity is that the volume envelope seems to fluctuate, quite dramatically at times ... resonance, room echos

Yes. However, my Impulse Response shows the reflections to be down 25 dB or more from the direct.

The addition of the direct + reflected under that assumption:

Coherent: 85 + 60 = 85.475dB
Incoherent: 85 + 60 = 85.014dB

That takes some of the steam out of worrying about reflections. But it doesn't jive with a spectrum, so, still a curiosity:

Using the files above:
Top - CD
Bottom - Room

Audacity doesn't seem to have a way to process the microphone calibration file, so there's a little hump in the highs that may not really be there, and maybe a little sag in the lows. The 46Hz room/listening position hole is evident, but I'm pressed to identify it when listening to music. Below 20hz is the environment, seems inaudible, ever-present, in varying degrees, distant Interstate, helicopter rotor blades, air conditioners, etc, maybe some whale songs.

_____

A repeating 5 ms log sweep source pulse:
Top - signal
Middle - with DRC
Bottom - without DRC

The timing here tries to preempt the wall reflections (after 6ms), but the 1 ms reflection on the back of the couch still disrupts the direct wave I'm interested in. I need to buy a mic stand for a little more experimentation with the direct wave here in the room.

 

[fas42]

Another fascinating characteristic, which I mentioned already, is the time domain variance in the waveform fidelity, again looking purely above 400Hz - all below is brickwalled out. In short bursts there will be remarkable correlation, almost a copy and paste fit - then, the signal virtually loses the plot, and it is difficult to see any relationship between CD and room - only for an almost perfect matching to be restored, once again, a split second later. One way of interpreting that, is that the distortion behaviour is extremely dynamic - one moment, high distortion; the next, very little - for no obvious reason, on first viewing. I have seen this type of thing before in comparing the waveforms of source, and reproduced capture, although at much higher levels of mismatch: the replay "slips in and out" of accuracy, constantly. IOW, it is not possible to make a meaningful, static measurement of distortion ...

 

[fas42]

On another look through, it is evident that there is significant amount of momentary compression occurring - related to human physiology, the system has poor "lung capacity" - when it needs to speak loudly the first utterances are full bore, but the energy stores start to dwindle rapidly, and the longer a sustained volume is called for the more disconnected the acoustic output becomes to the source signal; a tiny reprieve of low level input allows the energy levels to rebuild, and the next utterance is back up again to full voice, at the beginning at least. This is still for frequencies above 400Hz only, I glanced at the bass area while this was going on, and see nothing interesting there ...

If this sounds like power supply behaviour to you, well, it does to me as well ... I've been there many times in my adventures.

If people are interested I could post some pics of what I'm seeing, if Ray is OK about that - or, perhaps he may be able to add his perspective and thoughts ...

 

[RayDunzl]

If this sounds like power supply behaviour to you

I don't think I have any power issues at the few watts being generated to create the recordings above.



I see the direct wave modulated by delayed room acoustics and some low frequency transducer phase delay (partially corrected).

What you see as a period of low-level signal giving a chance for power supply recharge I see as a period for the dissipation of stored energy in the room, so the next impulsive wave looks nearly perfect until the new set of room reflections begin to modulate the direct wave at the microphone again.

 

[fas42]

But you said above that the reflections were down 25dB or more - which doesn't jibe with the level of changes I'm seeing ??

 

[RayDunzl]

But you said above that the reflections were down 25dB or more above - which doesn't jibe with the level of changes I'm seeing ??

I mentioned a discrepancy, above.

Impulse response says -20dB or more (mostly more) in the reflections, but the frequency response disagrees where there are holes due to room acoustics.

So, it is something I haven't reasoned out yet.

As for any of my power supplies being unable to maintain the necessary output voltages and currents at the interfaces due to some starvation issue at the levels being played, I don't think so. Maybe at your place.

 

[Blumlein 88]

As for any of my power supplies being unable to maintain the necessary output voltages and currents at the interfaces due to some starvation issue at the levels being played, I don't think so. Maybe at your place.

Everybody knows those Krells, especially in monoblock form have such wimpy power supplies it could be an issue.

 

[fas42]

Please note, this is not a point scoring exercise - I have had plenty of experience of hearing "monster" amplifiers having audible problems, and "shrimps" doing extremely well; many times there are no obvious reasons why there are various behaviors, the exercise is to understand what's going on ... in some curious places this is called, "engineering" ...

Ray has provided an excellent resource for comparing how well the output of an audio system matches the source data - and at times I can see a brilliant correlation between the two in the waveform, only for that to disappear a very short time later. The interesting thing to understand, for me, is why that's happening - if the system is functioning as theory would have it then there should be a consistency in behaviour, with minimum variability.

 

[RayDunzl]

Ray has provided an excellent resource for comparing how well the output of an audio system matches the source data - and at times I can see a brilliant correlation between the two in the waveform, only for that to disappear a very short time later. The interesting thing to understand, for me, is why that's happening - if the system is functioning as theory would have it then there should be a consistency in behaviour, with minimum variability.

Lets make a test:

I created a 5-second complex tone from four individual tones, played it, and recorded it. The signal tone is repeating but unrelated to the power line frequency.

If "at times I can see a brilliant correlation between the two in the waveform, only for that to disappear a very short time later" occurs, the comparison of the recorded tone at two different times should be quite different.

Top - signal
Middle - recorded and time aligned with signal
Bottom - copy of the recording and offset by about 2 seconds, so that the display below is the signal and the recording at the 4 second mark, with the copy of the recording offset and displaying the area at the two second mark.

There are no significant differences between the offset and the original recording wherever in time you look at it, after the initial stabilization period of a few millisecods.

Conclusion:

The observation "at times I can see a brilliant correlation between the two in the waveform, only for that to disappear a very short time later" is due to the time-offset modulation of the direct wave by the earlier signal being reflected back to the microphone by the room, and not by some electrical phenomena, in my case.

If a steady tone is being played, the subsequent modulation is the same (repeatable and repeating), if music is being played, the modulation varies.

To be clear, this is how the tracks are arranged, and the display above is taken at the 4 second mark.

 

[fas42]

Ray, to give you a visual of what I'm looking at, here's a sample:

As implied, the top waveform is all frequencies with everything below 600Hz lopped off - very good correlation between CD (not shown) and Mic. The bottom 2 tracks have only the frequencies 150 to 600Hz included, and this seems to be where this behaviour is happening - in particular, what's going on at 32.060?

The prior time slice, up to 32.090, looks good, and the one after is pretty decent in the matching - I don't "get" room effects causing this ...