Analog delay using allpass filter to avoid latency phasing

[Rogy]

Hi all,


I'm struggling with the following problem:

I'm the monitor engineer for a 70-piece symphonic orchestra. For years all musicians have an own three-channel analog mixer with the following features:

Channel one is a direct out of the built-in mic preamp in which their close mic is patched. This way they can adjust the level of their own instrument.

Channel two and three are two channels to which i send two mixes.

Channel two is a mix which contains all instruments in the same section as the musician in question; if the musician plays 1st violin, this mix contains all other string players (violin, viola, cello, dbl bass). NOTE: The musician's own instrument is also present in this mix.

Channel three is a mix of all instruments in different sections as the musician in question (eg the brass and wood players).


This setup works nice, using an analog monitor desk.

Now we want to replace the analog desk by a digital one.

Problem: The mixes sent back to the musicians arrive a couple of milliseconds later than the direct signal of their own mic, due to the A/D and D/A conversion time and processing time in the desk. Total latency is about 3 ms, and remains constant.

Is it possible to build an analog delay to delay the direct out of the musicians to get channel 1 and channel 2, which both contain the same instrument, back in phase?


Thanks in advance,


Rogy

 

[bcarso]

Analog delay, broadband, of that sort of an amount is a bee-atch. You would do as well imo to do similar A/D-D/A processing, made a little faster than the desk so you can always insert some shift register or equivalent memory delay as required.

It's going to bug the musicians in either case though---maybe not as much as the undoubtedly bizarre comb filtering you are probably getting now.

Too bad you can't convert the desk to faster converters. Perhaps some of the delay is from the type of other DSP going on, and maybe some of that could be rewritten to for example replace some long FIR filters with IIR ones that have lower latency. Maybe someone would be interested in making a product with a fast channel in parallel---of course that won't solve your problems now.

 

[Guest]

[quote author="bcarso"]Too bad you can't convert the desk to faster converters. /quote]
good old cheap SAR converters solves all...
And newer units go 18 bits. It is not much for marketing prospects,
but too much for Neumanns

Not much DSP for symphonics, only add. All filters you use
must be minimal - phase IIRs for low resolution.
You cannot use paragraphic equalisers, only DSP IIR allpasses and
adding. Much of programmers work, but it would be good to design good
DSP mixing desk with that properties.
And you need only 1-st and 2-nd order IIR digital allpass, add,
good DSP and 10 years of work for 10 peoples.
Would be superior and cheeper than analog.
I mean, that analog is somewhere superior than digital only
because digital is used improperly and programmed by lazy personas.

I would make it now, but no BGA soldering plant at home :-(...

xvlk

 

[bcarso]

What he said....

Wish I could learn Czech xvlk. You clearly know what you are talking about :grin:

 

[mike_relay]

just wandering.... what is the digital console you are using?
-mike

PS I didn't know about any orchestras with monitoring systems like this? Is this some sort of film scoring gig?

 

[Rogy]

Hi all,


I'm talking about the DiGiCo D5/D1 series.

IMHO the best sounding live digital mixing system.


About the orchestra monitoring system:

Every year there is a three month tour through europe of a classical orchestra accompanying some pop/rock artists. The show is called Night of the Proms (www.notp.com), and visits concert halls with capacity up to 20.000 people in four different countries.

Acoustical circumstances in some of the halls are far from great, that's why a decent monitoring system was developed.


Thanks for the reactions,


Rogy

 

[MHanson]

How many i/o's do you have and how many are you using? From the console standpoint, to keep everything in time, you need to take all your inputs of the orch and direct out them back down a snake to the respective party. This way the individual signal goes through the console, the mix is through the console and everything in theory should line up. But this also means you've just added another whopping breakout to the whole orch. (You'll have 4 lines going to each player: Mic in, Mic out and mix from console)

Beyond that, building a delay in the samples -> ms range is going to be a pain also. I didn't see a throughput delay listed. Do you actually know what it is?

Sorry, it's late and I'm probably ranting.

Michael

 

[jhaible]

I fully agree with the other posters: analog solution would not be good here.

Just to provide some numbers:

I have build a reasonably flat analogue delay line for 1ms, 5kHz Bandwidth.
This uses 25 inductors (and capacitors).

So for 3 ms and full audio bandwidth, a rough estimation would be 25 x 3ms/1ms x 20kHz/5kHz = 300 inductors.

You don't want to do this.

JH.

 

[Samuel Groner]

I have build a reasonably flat analogue delay line for 1ms, 5kHz Bandwidth.

How did you do that? Just a Bessel filter or some all-pass filters as well?

Samuel

 

[jhaible]

[quote author="Samuel Groner"]

I have build a reasonably flat analogue delay line for 1ms, 5kHz Bandwidth.

How did you do that? Just a Bessel filter or some all-pass filters as well?

Samuel[/quote]

In that specific case, copy and paste from a Hammond Line Box.

In general, you can approximate a transmission line (i.e. the delay from a wave running along a cable) by breaking it down to tiny LC elements. Drawback is that the LC defines your delay time as well as your bandwidth, so for high bandwidth and a delay in the millisecond range, you need a ridiculous amount of elements.

The whole thing is a gigantic LPF - if Bessel I don't know. Propagation delay in the passband is very flat, at least.

JH.

 

[bcarso]

[quote author="jhaible"]

The whole thing is a gigantic LPF - if Bessel I don't know. Propagation delay in the passband is very flat, at least.

JH.[/quote]

Yep. Good explanation jh. The Bessel is about the ideal approximation to getting time delay without preshoot/overshoot.

Crown uses a 30kHz 8 pole Bessel iirc to give one of their power amps a bit of look-ahead for various reasons.

To paraphrase the wonderful old movie with Alec Guiness playing 8 characters, Kind Hearts and Coronets: "Ah, time delay...one thing the bitheads do really well."

Although sometimes it explodes in their faces.

 

[12afael]

look this

http://www.diystompboxes.com/sboxindex.php?topic=25168&postdays=0&postorder=asc&start=0[/url
I make some simulation with rc filters but I need like 16 stage for 0.5 ms for a bandwith of 10 khz if I remember well.