MeterMan

Here ya go

http://www.youtube.com/watch?v=SvC2JEHyGDI

Cheers and thanks to all the guys who contributed in the
thread up in the Drawing Board,

Lukas

Nice one!!

I don't mind the meter but I do love track a lot... ;D

Very cute.

What's the relative level of the peak/average meters? In other words, if you feed it a sine wave, how many dB difference between the bar and the dot?

JDB.
[I'm afraid I have to agree with the comment at 1:28, though]

Nice work..

I obviously like the concept...  It was my first patent back in 1979 (US04166245  Roberts).

In the '80s I built up a version on the bench using a 3915 duplexed for input and mode. I used CMOS transfer gates to do the switching. I used the 3915 with logarithmic LED spacing so you could directly calculate crest factor from distance between peak dot and VU bar anywhere within the range of the display (equal 3 dB spacing).  I never found the LM39xx cost effective for real production.

Yes, it can be done using a microprocessor, with more bells and whistles like highest peak during VU, highest peak hold, optional modes etc. Some console glue can be incorporated into the meter micro (like PFL, emergency alarm, etc). There's even stuff in there I don't talk about.  ;D

http://www.johnhroberts.com/christmas_PK_VU.wmv
http://www.johnhroberts.com/lazy.wmv

The song tracks are from a friend's website. Not presented as his best work but OK for demonstration purposes.  FWIW the crest factor or distance between peak and average is more pronounced on raw individual tracks than compressed final mixes.

FWIW I went through the trouble of coding up a RMS conversion for the VU display, but didn't see any significant difference on the bench so removed it. I may revisit RMS for a premium version. 

JR

PS: My patent is now 30 years old so the approach is public domain. I've been messing with meters for a while.

Well happy 30th!

I'm glad to see I'm not the only one nutty about meters
Thanks a lot for the input back then, it was mainly you I owe
it to that it's working.

But really, it indeed isn't a solution for whatever market when
you can just take a 2-buck-50 chip and flash it. Space would
be a direct second consideration (though you could probably
do a nifty little SMD sure enough)...

There is this gratifying thing though about getting nuts and
bolts working like that...just going from idea to "hey wow
and that's it".

@JDB, haven't gotten anything solid to speak of except for
the peak which will work to the specified 0.2dB error margin
worst case - but as for average, I have more of a "kitchen
logic" approach...was there a specific thing you had in mind?

livingnote said:

@JDB, haven't gotten anything solid to speak of except for
the peak which will work to the specified 0.2dB error margin
worst case - but as for average, I have more of a "kitchen
logic" approach...was there a specific thing you had in mind?

Click to expand...

Nothing specific, I was just surprised at the apparently high Crest factor.

JDB.
[and if I wasn't so lazy (and not tied to the computer with working SPICE but broken Acrobat) I'd have looked up the dB-per-LED figure for the LM391x parts rather than vaguely recalling it to be ~3dB]

On a finished mix I rarely see more than a few dB of crest factor.. That meter looks like a very slow release time on the peak (slower than my peak hold) and perhaps an odd scale.

I always try to use a log scale so the meter peak to ave spread  looks similar for any loudness of the same track.

JR

It's the 3916 with the 3915, I believe it's +2 to -3 in one dB steps, then -5, -7
and -10, with the 15 taking over from there in 3dB steps, I dropped the -20 light.

The ballistics come from a 100µ cap in there, and the crest really can't be taken
from the meter as of yet because it's not yet set consciously w.r.t. time constant
and amplification - do you know if the average detector in the datasheet from
NatSem is chosen to be a happy compromise to hopefully hit close to rms most
of the time?

RMS does not have a specific time constant.

I have not looked at the National schematics in a long time.

I have settled on roughly 200 mSec attack and release for VU/average. 4 mSec  attack and same 200 mSec release for peak. For best viewing of crest factor you want the peak and average release to fall at the same rate when signal drops. Of course if you use other than logarithmic (dB) level scale, this is less important.

Peak hold time is subjective, and I don't remember what I used.

JR