Late Not Great, Yamaha CL1 vs Behringer X32 Subgroup Latency (Public)

[@mattcraig9433]

First time I ever tried doing parallel anything on a CL things got weird. So much so I hit up Yamaha CA to ask if there were fixed values so I could at least compensate. They had no idea what I was going on about. So glad to see I wasn’t going crazy! Thank you for these!

[@dylanbarber6899]

Same here. I remember trying to run a parallel drum buss and just ditching it off in soundcheck because it sounded so weird. Following research I ended up just sending my drums to two busses, to keep them aligned, even sending the dry buss through a bypassed insert rack just to keep it as tight as possible.

[@DaveRat]

🤙🤙🤙

[@pressorv]

@@dylanbarber6899 to me desks should adjust timing to compensate like DAWs, or at least make it a point to alert using parallel path or any ways it could happen to warn the user with a session/global override. some sort of idiot light if you don't balance it..

[@akusbass]

@@dylanbarber6899 This is the one and only way to do that with Yamaha consoles. It’s a shame!

[@DaveRat]

It's a challenge between minimal latency, or fully compensated or having latency keep changing with routing. No solution is ideal. Minimal latency is desirable but it requires the engineer to be cautious with routing.

[@briankinnaman6532]

This is a very interesting topic and one that maybe doesn’t get thought about enough. As always, thanks so much for sharing, Dave! 🤘🏼

[@DaveRat]

Thank you Brian and agreed.

[@danialdevostmusic]

Interesting to see the two manufacturers have different priorities. Yamaha developed A/D and mic pres with perfectly linear frequency response, but didnt implement latency compensation. Behringer chose otherwise. Honestly in this case I would say Behringer made the better choice

[@jean-lucbattista2492]

And the sends on output faders/mix on input faders. Really missing that on the Yamaha's

[@DaveRat]

🤙🤙🤙

[@robertlofgren6834]

The hi-end rolloff is due to the ad-converters anti-aliasing filters up to the nyqvist freq. I find this gentle slope to be quite pleasant since it limits hi-end buildup that otherwise would sound harsh(er).

[@danialdevostmusic]

Robert Löfgren oh i did not consider that, you are probably right!

[@martheunen]

Very interesting! I used to do ch>LR and same ch>bus>LR with variations with fx and or compression quite often, but have been using it less and less with digital consoles. Sometimes it just didn't sound right and I new it was because of phase shifts and this is a perfect illustration of that. I work in theatre nowadays with pretty much a complete yamaha dante setup (CL3, rio, mtx and other stuff) and it's a treat (when i actually get to do the mixing myself that is), but still try to avoid abovementioned phase shifts. So no more (or very much less) 'doubling' of signals. You win some, you lose some i guess... Love your insights, 'tutorials' and experiments! Keeps me up to date and still learning!

[@DaveRat]

Awesome thank you Mart!!!

[@artysanmobile]

Any device that ‘mixes’ must employ automatic delay compensation, yet few do. As an everyday user of DAWs, I am amazed by this, as simply disabling it turns my meticulous work to utter shit. A ‘mixer’ by its very definition MUST correct for all contrasting delays incurred by its mechanics. The result of NOT doing this can be distortion so gross as to render the mixer useless for music. The result of DOING this causes a single comparatively meaningless penalty; the final combined output is delayed by some indistinguishable quantity. That’s a pretty powerful argument for mandatory automatic delay compensation. Even better, it is not particularly difficult to implement. Each digital process in a piece of hardware causes a precisely known and repeatable delay that can be reported to a DSP, and there is a finite number of these processes, actually a pretty low number so the needed code is minimal. Sound that is late is a daily part of the human experience. As we step away from a source we are perfectly aware the source is the same thanks to our brain’s happily adjusting. Sound that is grossly distorted by haphazardly combining different delays is a totally different beast. That is an effect. An example is the reflections off a tiered surface such as the stairs in a stadium grandstand, which we’ve all heard. If your mixer is going to give you this all on its own, you don’t have a chance.

[@DaveRat]

Hey hey peter, I agree except for the exceptions. Here is a video diving into a common situation where latency is an irreversible issue that parallels the issue of routing latency. When we hear our own voice late through in-ears or headphones, there is no fix. kzbin.info/www/bejne/d4HNYY2Xgt5njtk

[@artysanmobile]

@@DaveRat Agreed to the ears comment. I keep showing my studio roots!

[@jean-lucbattista2492]

What about channel processing? FX insert? FX return? Matrices? It should all be clearly stated onscreen manufacturers!! It is that important! Say with a tunable delay that cannot be deactivated with it's minimal value corresponding to the added latency in that path.

[@DaveRat]

Yeah or it should not exist. Either way, awareness is important

[@tobitweaks]

Good question! I would put the name of the audio alteration „Latency“ simply in the delay category - it essentially is a delay that can cause chorus, flanging or phasing when you add it to a non delayed signal :) As it does not distort the waveform I would not call it distortion no. Thank you so much for your work it it soo valuable!

[@DaveRat]

🤙🤙🤙

[@Solmbarnes]

Very surprised that the Yamaha desk doesn't have a fixed latency. I would be very interested in seeing you investigate this more. Particularly in relation to how this could potentially course issues with real world use.

[@steakikan]

I think they only started doing fixed latency on the Rivage family.

[@DaveRat]

I think as long as you avoid recombining the same or similar signals, you should be good

[@murraywebster1228]

Would be interesting to know which desks have latency compensation to achieve phase alignment at final output stage, like what many DAW‘s do.

[@DaveRat]

Yes, agreed. While I have no desire to test a pile of desks, I did do the test with simple cheap gear so anyone wanting to test can do so.

[@artysanmobile]

The list is vanishingly small. S6 and S3 from Avid are notable. Without reading the manuals, I have not found any other claim, though I suspect the very high end from Calrec, LAWO, SSL Sony Oxford are likely. I don’t know of any other device advertising the same.

[@VertedMonk]

Drinking some bourbon listening to the King.

[@DaveRat]

Good times! Audio and alcohol do mix

[@artysanmobile]

It would be a great service to list the devices that offer automatic delay compensation. I am going to make my best effort to do this. I’m sure it will be an ongoing project.. In my opinion, these are the only devices for serious music mixing.

[@DaveRat]

This was shared during a discussion on that topic in one of the recent channel member zoom chats www.prosoundtraining.com/digital-mixer-study/

[@artysanmobile]

@@DaveRat Thanks very much for this, Dave. Some fine work (and lots of it) went into this but the bottom line is still well hidden, as shown by this grid investigating only a few of myriad scenarios. I’ve been working all day on this and have so far found only Avid advertising hardware that implements “full auto delay comp”. It is so hard to find manufacturer detail on this the only reasonable conclusion is that no one else is doing it, but I can’t accept that. Surely the Oxford OXF-R3 must have, and I’m hoping for SSL to be transparent about the Live series. I have friends at some of the manufacturers who will help in my search. Lawo, Sony, SSL, Calrec are on my checklist. Until we can freely assign and process without gross errors, digital mixers are letting us down. Only by shedding light on this issue will things percolate down the $ scale.

[@KSJAFN]

Really interesting stuff - thanks. From memory, a lot of these consoles (eg Avid SC48 springs to mind, but don’t quote me) claim to somehow “manage” routing latency internally. I recall when behringer launched the x32 its low latency was a bit of a selling point - shows that one can trust vendor marketing at least sometimes.

[@DaveRat]

👍

[@artysanmobile]

They don’t just claim to, they actually do it and I have explained in general terms how it’s done in the comments here.

[@Mtaalas]

This is why Behringer bought Midas. Any modern console should be completely phase coherent, meaning that the latency is fixed across all buses and groups and routing. Only way to get phase incoherence introduced should be an AD and DA conversions (which are in order of microseconds). So you go out of the console and come back in, that introduces just a tiny bit of latency... but you should be able to compensate for that using console's delays. This _might_ come with a cost of a small amount of overall latency that you can't get rid of (let's say 0.5ms for the arguments sake), or by overall latency increasing as routing increases, but it's (or should be) basically inaudible since it only means that the source of your audio has moved like some centimeters (few inches) further from you, depensing... basically non existent change. If one's using digital snakes, even that should be phase coherent in general, meaning that what comes out of the stage box is in same phase and delay as what comes out of the console's local output... you might want to test that as well? :)

[@DaveRat]

Agreed and unfortunately the x32 and m32 add latency when imternal plugging or efx are added to the groupa and also when inserts are used. Optional aligned latency and minimal latency modes would be good.

[@Mtaalas]

@@DaveRat as an electronics designer, the issue with latency compensation is the need for very fast and low latency memory. Every sample of delay needs 24 or more bits and if you introduce let's say 100 sample's of latency on one routing, you need to store 2400 bits * number of channels and buses etc worth of data. Let's say we have 32 channels + 16 busses + 4 subgroups + stereo main output, that's 2400*(32+16+4+2) bits of data or 129600 bits, that's ~16kB of data or so. Seems like small amount, until you understand that it just can't necessarily be stored in some block of DDR-memory that can be gigabytes, but in most cases has to be stored within the DSP-processor(s) in very fast registers or local cache since the timing requirements are so tight that monolithic DDR isn't fast enough because we talk about hundreds of microseconds of latency here... a real-time processing issue. So there needs to be a limit somewhere or the costs of the hardware can skyrocket pretty fast because memory like that needs bigger DSP's and tighter timing etc... So it's understandable that there needs to be a limit... I mean DSP processors themselves are expensive, so you try and reduce the amount as much as possible in a product like x32, and developing multi DSP systems is expensive and time consuming so yeah, most of this memory has to be found from one or two DSP's that do the grunt of the work, so it could need megabytes of local cache in larger consoles or longer latency compensation, and memory takes up space from the silicon. :)

[@artysanmobile]

The words “fixed” and “latency” don’t really belong together. Every process unavoidably incurs some delay, particular to that process. So, inherently, a mixer with a dozen or so different processes, needs to be compensated automatically for each of them every single time one is invoked. It’s not rocket science, though it is a difficult thing to enforce and to do without creating noises. Much math is involved and a system-wide approach with firmware updates to accompany feature changes. Our awareness of this will force the hands of mixer builders.

[@artysanmobile]

@@Mtaalas Memory speed is definitely not the dealbreaker. Consumer DRAM works at rates well above 10TIPS. 10,000,000,000,000 throughputs per second. Your 16KB would move in and out of memory in .00000006 second if done serially, and that’s not the way it’s done anyway. Memory, I assure you, is no part of the problem.

[@cesarperez-cardenas97]

You're a wizard, Dave. So informative watching your videos. Trying to break into the production and pro audio industry. Hopefully bump into you one day. Keep rocking and rolling. Much love all the way from Miami! 🐀

[@milmar_echoes]

0.2 ms delay onthe grp CH equals 5000Hz correct me if am wrong, a quater wavelenght has .2ms means 4*.2=0.8msLampda=1250Hz (It takes .2ms for a quater wavelenght of 1250Hz to pass.) now both waves r on the way to master bus, one passes the deay the other not, they will cancel each other out partially i guess is that what happens when mix busses and channels on the yamaha unfortunatly i did too much rock n roll and the feds hunting all my earnings down….still can t afford a consoles to dive into experiments like this. To see you reminds me of the gd ol days in the live company where i learned sound engineering, i could spent all day with experiments like this. great job 🐀

[@DaveRat]

Yes, .2ms is 5000hz full wavelength. Max cancellation is at 1/2 wavelength, so cancellation would increase as the freq approached 2500 and then there would be comb filtering as the freq increases.

[@brandonstorms3186]

Subbed!! Love your content and gremlin hunting!

[@DaveRat]

Thank you Brandon! And welcome!

[@gaffster787]

Interesting. Yes, latency is distortion if the signals are recombined....including by your ears. I remember we conducted experiments with avionics communication radio "sidetone" delay. Traditionally when a pilot grabs a mic and talks to the tower, his radio wraps the audio sent to the transmitter back to their headphones so they have confidence the radio is on and working. When we started digitizing and delaying that audio....well that could be a real problem. Even a 100mS delay some folks can't even read from a book without messing up. Also if two radios are receiving on the same channel and their outputs are combined, they better have the same delay...or if not possible, at least vary it randomly on purpose to avoid getting stuck at nulls. Swishy, swishy. Radio, Radio. I wish I was in Tijuana.

[@DaveRat]

Cool info, thank you!

[@Edwin-van-der-Putten]

Great video, Dave! Very interesting! 🙂

[@DaveRat]

Thank you Edwin!!

[@michellefoster1020]

That’s because it’s Midas technology behind it takes 32 is a fabulous desk a live recording and conference situations it covers all the bases

[@DaveRat]

👍👍👍

[@ThePatboil]

Such a legend .

[@DaveRat]

Thank you Pat!

[@artysanmobile]

My long detailed comments aren’t reaching the intended. That’s my fault, so here goes. Latency is delay. I find it helps to call it what it is. Delay. Fact: Delay is an unavoidable part of digital music recording. It is unavoidable, period. Fact: It is not a single delay, but different delays for every channel and every change. EQ on bass drum. Routing drums to a buss. Running your signals down an Ethernet snake. Panning. Each of these day to day things causes its own, different delay. Fact: We combine these processes and signals when we mix music. Fact: Combining two identical signals with ANY difference of delay, even a millionth of a second, creates a gross musical distortion. It is not subtle. You can try this yourself. No amount of delay is acceptable for this scenario. Not .5mS, not .000005 mS, not a single sample either. From these irrefutable facts, you should be able to see what we face. Manufacturers can solve this by building into their products “Automatic Delay Compensation”. When all of us understand this, they will. Until then, they won’t. I welcome any argument, questions, discussion here and thank Dave Rat profusely for tolerating my use of his forum.

[@DaveRat]

Hmmmm, careful with staying the extremes as facts. First, yes latency is a delay a specific type of delay. Using the specific more accurate term of latency to refer to the inherent delay caused by digital processing is accurate. As far as 1 million of a second delay or latency having a negative impact on audio signals that have been recombined I suggest you test and rethink that statement. It is nearly impossible to recombine two signals at exactly the same time even a fraction of an inch of wire length differential will cause a shift in latency in the analog domain. These small amounts of time differential are not relevant to audio. The relevance of time offset with recombined signals is progressive and frequency dependent. It is not an all or nothing situation

[@artysanmobile]

@@DaveRat I clearly stated that delay difference for TWO IDENTICAL SIGNALS and yes, you can try it. A single sample will create a comb filter of epic proportions. I stand by that. The word latency was coined for a very specific use in science to quantify time discrepancies in such things as light propagation, current flow, memory write/read cycles, and other minuscule delays that are of no use to us in the audio world. It is an unnecessary confusion for us, having used the word delay for 50 years, to call this anything but what it is to us - delay. The word ‘latency’ causes endless confusion among the technically illiterate, and frankly, that’s most people. Why we perpetuate that confusion I will never understand. The extreme you chose, an inch of wire, results in .00000000008 seconds of delay. This is quite a few orders of magnitude smaller than my most hyperbolic overstatement, but I do take your point.

[@jamesonvogltanz1654]

Hey Dave! Really enjoyed this video. I expected better from the Yamaha! I use a CL5 every week so I'm a bit biased haha. I will note that on the Yamaha setup page, console latency can be adjusted from .25 to 5ms if you're using Dante i/o. Seems that there's a bit more latency if you're just using the direct i/o on the back! Regarding the increased latency after routing... very interesting. I'm curious whether the EQ, Dynamics, and/or Inserts on the channels contribute any further latency. Same for X32. Topic for future video? Thanks!

[@DaveRat]

Yes, the ADC and DAC are often the biggest creators of undesired latency. I will ponder diving into the impact of EQ and such on latency. I did test the X32 and the EQ on the channel does not alter latency.

[@Craig_Anderson]

That "console latency" setting for Dante is only for audio transmitted via Dante - the Rio stageboxes have their own latency spec, which is "less than 2.5ms", i.e. identical to the direct I/O on the console. So that Dante latency setting would be on top of the ADC and DAC.

[@DaveRat]

👍👍👍

[@zambotv8150]

Yes, he's back!!!!

[@DaveRat]

Yeah, KZbin broke again, maybe check for missed content as I have been posting

[@zambotv8150]

@@DaveRat Will do Dave, cheers buddy

[@MrWoodstock35]

Interesting question about if frequency rolloff from D/A should count as a form of distortion. My initial thought is that by definition it should be, but that doesn't seem like the best use of terminology. If every change to a signal is distortion there's no reason to use the word. It would be better to just use the more specific word for what we are talking about. As an aside. Could you make the case for delay/latency being distortion? It's the same thing going in and coming out but not at the same time. If everything is direct lines we can just move our window of measurement to see it's the same, but if we're combining live and reinforced sounds we can't adjust that window. Is the sound coming out of the speaker "distorted" because it's late, or is it the combined signal of original and amplified sound in space that is distorted (from phasing) compared to the original?

[@DaveRat]

While there does not seem to be a correct answer and it depends on how one defines distortion. If a signal is identical, then its not distorted, but what if the before and after signal can not be summed without issues? An identical signal can be summed with itself with the only change being an increase in level. A This does not apply to a latent signal, so perhaps time shift is a form of distortion by that definition. But EQ and compressors, gates and crossovers would also be a form of distortion. So maybe adding "unintentional" or "irreversible". But then intentional distortion would not be distortion if it is intended, Just fun with words but overall, latency should not be dismissed as it can be quite relevant and is often undesirable and in those scenarios, I believe its fair to lump it in with distortion. Also, lower distortion often comes with higher build costs and latency falls into that realm where lower latency does get expensive as an infinitely high sample rate is need to reach zero latency.

[@christopherbenedetto]

That was awesome.

[@DaveRat]

🤙🤙🤙

[@tm8473]

The loss of high frequencies close to nyquist, may be affected by the converters design which incorporates a more or less steep LP filter.

[@DaveRat]

In other videos i take a look at that and the differences

[@maustin5113]

It’s easy to make up gain but latency is a lot harder if not impossible to correct.

[@weareallbeingwatched4602]

You can delay the other sources to match, or start moving speakers around. Latency is common in the acoustic domain, and the rate of latency is about 1ms per foot.

[@DaveRat]

Yeah, with things like in ears where the sound is combined with the resonance and sound of the human body source, latency can be an issue. With wedge, latency can make things sound further away which may be undesirable But mainly, its just annoying and being aware is the key to not making combining errors

[@maustin5113]

I had a situation where I was trying to record live vocals with a few backing tracks and the delay offset was just enough to drive the singer crazy. The singer was pretty frustrated but he was super talented and goal oriented and managed to sing on key and with nearly perfect timing. I understand that you can possibly delay other sources but sometimes it’s just gonna be a problem. Whereas gain makeup is just a knob twiddle away from being fixed. I never had any of these issues with my analog consoles 😂

[@maustin5113]

@@DaveRat yup 👍🏻.

[@gregblades9373]

If latency was considered a form of distortion then every time a recording was listened to it would sound more and more different than the first. Latency is only time distortion!

[@DaveRat]

Is EQ a distortion? Is distortion determined by intent or alterations that are not optional?

[@artysanmobile]

I would challenge that, Greg. Delay, per se, is not distortion, but this entire discussion is about non-deliberate mixing of delays and the result is a distortion you would find appalling. You may not have understood the premise here, but a very simple though experiment is to listen to two copies of a sound in your DAW simultaneously. Move one of them a single sample and listen again. That delay, what you call latency, will change your mind forever.

[@ToddWCorey1]

I think I'd take the high frequency loss over the added latency. What do you think?

[@DaveRat]

I agree, the HF loss is minimal and the latency is relevant and non ideal. That said, the durability and robustness of the Yam is superior for touring. For corporate gigs and such as well, the Yam is probably better. But I do find the attention to noise, latency and routing latency to be disappointing for a console that costs so much

[@ToddWCorey1]

@@DaveRat I appreciate that your videos give information without conclusions, that you leave it to us to draw our own conclusions. But, dammit, once in a while it'd be nice to just hear you say, "do this, don't do that!" But things are never that simple.

[@neve2254e]

interesting... This kinda stuff drives me crazy. Especially if it isn't all in time.

[@DaveRat]

🤙🤙🤙

[@twitchbook-1]

Cheers

[@DaveRat]

👍👍👍

[@treleasethomas]

Awesome stuff, Dave! The yamaha console has a latency switch in the dante setup page for matching latency with the size of the network. With 1ms delay being the default value the console chooses (and the lowest being 0.25ms), might that be the difference in latency we see between the CL and the X32? Im just not sure if this still applies for the internal latency of the console or just the dante network. Any thoughts?

[@pressorv]

interesting point. i would guess the majority of users will not assist adjust this, so a stock to stock comparison is fair.

[@rostislavkundik6154]

Dante latency settings affects only dante networking, ex when connecting rio stageboxes. In this video Dave don’t using Dante network, he is connecting straight to board. If we add corresponding stageboxes to the consoles, loss in latency of Yamaha will be more. AES50 is the fastest protocol - 60 microseconds per link, and Dante will add at least 250microseconds pre link

[@DaveRat]

yes, dante is external so whatever latency issues that are created or exist internally can be transmitted properly. ideally there would be a similar compensation internally as well

[@DaveRat]

👍👍👍

[@artysanmobile]

This corrects the delay difference between a snake input and a rear panel input and that’s a great start. It isn’t terribly common to take signals at these two inputs that need coherence but every correction is welcome.

[@tm8473]

I suppose that the m32 shows a similar behaviour as the x32. Did you tested the midas too?

[@DaveRat]

Yes, here is a link to all the console videos kzbin.info/aero/PLqsPeTx2oVyeOD8-01y-5OQwqzcxxQsoo

[@matijatatomirovic3351]

I can see some geeks running 6 ad-da on the x32 for that warm analog sound.

[@DaveRat]

Fun!

[@magicmaui77]

Some claim the M32 series to have latency compensation - I found no prove so far, because the M wasn‘t available when I needed it, so I had to get an X32. If the M does compensate, I wonder if it‘s a software-thing. Maybe the firmware distinguishes between X&M and activates some special features? 🤷🏻‍♂️

[@robertlofgren6834]

The x32 and m32 handles latency/coherency compensation exactly the same. The m32 is NOT treated in some special way. This hoax came about when the midas marketing team used different wording compared to when the x32 was released and people thought that the x32 didn’t have the same m32 processing engine/features.

[@stephenevans2518]

I saw a nice series of test measurements a while back but can't seem to find them! From what I recall, the X and M32 will be in phase with parallel bus routing until you add an inserted FX to one of the paths, but that would add extra latency and can cause comb filtering between parallel paths

[@robertlofgren6834]

@@stephenevans2518 , that is correct. The different rack-fx causes different latencies depending on which fx you insert. This holds true as for the aux inserts as well.

[@magicmaui77]

@@robertlofgren6834 Yeah, that‘s what I was thinking about (parallel compression f.e.). If they both behave the same, I‘m quite happy with my X32. Apart from the better faders and preamps you get a lot for your money.

[@DaveRat]

👍👍👍