System DSP roundtrip latency

C

Christian Tepfer

Sophomore
Jan 12, 2011
238
0
16
51
Hamburg, Germany
klanggestaltung.wordpress.com
While I'm capturing a trace I was thinking, we could make a documentation of processor latencies.

A colleague found out that the Mayer Galileo has a higher latency when the digital input is used from a PM5D. I try to make him post his insights here.

Interesting topic, on a small show with a digital desk it surely adds up into latencies higher than the propagation delay from the loudest stage noise emitter, compared to the PA.



So I start with what I have before me right now.



VMB ARK-48, analog I/O: 0.56ms
 
Re: System DSP roundtrip latency

Likewise, this is true for original I-Techs. Analog latency is 1.13ms, Digital inputs are 2.36ms @ 44.1/2.28ms @ 48k/1.81ms @ 96k.



Tim Mc
 
Re: System DSP roundtrip latency

Tim Padrick wrote on Sun said:
According to the specs on our Rane RPM88s, the latency of the unit analog in to analog out is shorter than the latency of the unit digital in to digital out. I don't get this.
Click to expand...

A colleague measured analog in - PM5D - Mayer Sound Galileo - analog out:



digital, 48kHz: 5.073ms

analog, 48kHz: 3.802ms

digital, 96kHz: 3.448ms

analog, 96kHz: 2.656ms



That's a drastic difference imho. It seems to raise from the SRC done in the unit.
 
Re: System DSP roundtrip latency

Christian Tepfer wrote on Sun said:
Tim Padrick wrote on Sun said:
According to the specs on our Rane RPM88s, the latency of the unit analog in to analog out is shorter than the latency of the unit digital in to digital out. I don't get this.
Click to expand...

A colleague measured analog in - PM5D - Mayer Sound Galileo - analog out:



digital, 48kHz: 5.073ms

analog, 48kHz: 3.802ms

digital, 96kHz: 3.448ms

analog, 96kHz: 2.656ms



That's a drastic difference imho. It seems to raise from the SRC done in the unit.
Click to expand...



Yes. In Robert Scovil's Complete FOH Engineer seminar he discusses sample rate conversion. I recall one of his comments as saying ''you really don't want to see these 3 letters: 'SRC'.''



There are all kinds of places for latency to creep in and they do add up. I can foresee in some unique situations where cumulative latency could ''move'' the PA upstage of the acoustic source on stage. It would take some specific staging and PA placements for this to happen, but I can see a frustrated SE trying to fix an alignment issue, desperately needing the Ultra-Speeder-Upper Pro® to spit out electrons faster than they come in and 'fix' the latency issue. Nudge, nudge, wink wink.



Tim ''Alchemy is our friend'' Mc



 
Re: System DSP roundtrip latency

Tim McCulloch wrote on Sun said:
Christian Tepfer wrote on Sun said:
Tim Padrick wrote on Sun said:
According to the specs on our Rane RPM88s, the latency of the unit analog in to analog out is shorter than the latency of the unit digital in to digital out. I don't get this.
Click to expand...

A colleague measured analog in - PM5D - Mayer Sound Galileo - analog out:



digital, 48kHz: 5.073ms

analog, 48kHz: 3.802ms

digital, 96kHz: 3.448ms

analog, 96kHz: 2.656ms



That's a drastic difference imho. It seems to raise from the SRC done in the unit.
Click to expand...



Yes. In Robert Scovil's Complete FOH Engineer seminar he discusses sample rate conversion. I recall one of his comments as saying ''you really don't want to see these 3 letters: 'SRC'.''



There are all kinds of places for latency to creep in and they do add up. I can foresee in some unique situations where cumulative latency could ''move'' the PA upstage of the acoustic source on stage. It would take some specific staging and PA placements for this to happen, but I can see a frustrated SE trying to fix an alignment issue, desperately needing the Ultra-Speeder-Upper Pro® to spit out electrons faster than they come in and 'fix' the latency issue. Nudge, nudge, wink wink.



Tim ''Alchemy is our friend'' Mc
Click to expand...

Right, Tim.



I basically want to be aware of the combined latency in my signal chain in order to chose the right tools for the job.



So maybe NOT EQ the nearfills through the system processor, even if there's enough I/O and DSP because the latency there would be too high to get them where they belong.



Or use one additional AD/DA conversion to get the latency under the hurting value.



Alternatevely I take the time machine. Let's give Doc Brown a call...
 
Re: System DSP roundtrip latency

Christian Tepfer wrote on Sun said:
Tim Padrick wrote on Sun said:
According to the specs on our Rane RPM88s, the latency of the unit analog in to analog out is shorter than the latency of the unit digital in to digital out. I don't get this.
Click to expand...

A colleague measured analog in - PM5D - Mayer Sound Galileo - analog out:



digital, 48kHz: 5.073ms

analog, 48kHz: 3.802ms

digital, 96kHz: 3.448ms

analog, 96kHz: 2.656ms



That's a drastic difference imho. It seems to raise from the SRC done in the unit.
Click to expand...



There's also a ''feature'' in the Galileos where the right channel is a couple samples late relative to the left channel when using the digital inputs with a 44 or 48k source. Meyer knows about this. The information they gave me was to put the delay on the left, but actual measurements had me put a .02 ms delay on the odd #'ed inputs on my Galileo's.



In most situations it's probably not a big deal, but if you start summing a stereo pair of inputs in the Matrix you could potentially run into some weird destructive interference on anything panned up the middle.



Dave
 
Re: System DSP roundtrip latency

...what I have before me right now (beeing bored
icon_sad.gif
):



LAKE LM-26, analog I/O: 1.02ms



more to come...
 
Re: System DSP roundtrip latency

Dave Stagl wrote on Sun said:
There's also a ''feature'' in the Galileos where the right channel is a couple samples late relative to the left channel when using the digital inputs with a 44 or 48k source. Meyer knows about this. The information they gave me was to put the delay on the left, but actual measurements had me put a .02 ms delay on the odd #'ed inputs on my Galileo's.
Click to expand...

This reminded me that the processing applied and whether delay compensation is employed can also make a difference in both the overall latency and the relative latency between any two outputs.
 
Top Bottom