Assetto Corsa: McLaren MP4-12 Imola Video

[Bram Hengeveld]

Bram Hengeveld submitted a new blog post:

Assetto Corsa: McLaren MP4-12 Imola Video

[youtube]

After the much discussed GT3 video of this car earlier this week its now time to see the road version of the McLaren MP4-12 in action on Imola.

Driving without any assists Marco Massarutto shows us how the Autodromo Enzo e Dino Ferrari looks in Assetto Corsa around lunchtime in free practice mode. The popular Italian race track, nicknamed by Enzo Ferrari as the Little Nürburgring was...

Continue reading the Original Blog Post

 

[DennS]

Not at all but, I do enjoy watching replays of races that I run. It's just that the track-side sounds have been lagging far behind graphics for . . . well, forever. We can show replays that can be mistaken for RL video but, show me replays with realistic sounds that haven't been dubbed. Nothing has come even close to date. Admittedly, I am very aural and rank sounds as a higher priority than most people. It's great to see some recent improvements in this area but, surely there are more gains to be made in terms of sound dynamics. Perhaps there just hasn't been enough market demand for better sound dynamics.

Well the thing is you can do only so much with sound. First of all everybody has different speakers, rooms or hardware in general so it will sound different for literally everybody. Next thing is most racing cars are unbelievable loud, something which adds to the sound immensely. Not even a recorded race for TV sounds remotely close that what it sounds in real life so how can it sound like real life in a sim...? Sure there are huge differences from game to game but nonetheless Simbin games, iRacing and AC (from what we heard) sound pretty good to me...

 

[RaceNut]

Well the thing is you can do only so much with sound. First of all everybody has different speakers, rooms or hardware in general so it will sound different for literally everybody. Next thing is most racing cars are unbelievable loud, something which adds to the sound immensely. Not even a recorded race for TV sounds remotely close that what it sounds in real life so how can it sound like real life in a sim...? Sure there are huge differences from game to game but nonetheless Simbin games, iRacing and AC (from what we heard) sound pretty good to me...

That very sentiment may be what keeps the sounds from progressing further. We can easily control volume - the bigger issue is to have enough contrast in the sounds to provide a more realistic sounding result. Environmental sound dynamics are incredibly complex - no doubt but, so are visual dynamics and look at what is being done there.

I believe that there just hasn't been enough public requests for sound improvements in racing sims to drive improvements in sound tech. That may be starting to change but, it would be a shame to have a title like AC not have some stellar sounds to match the rest of the features. Thus far, the sounds seem to be decent but, not near the level of RRRE or iRacing's new sounds. Even rF2 sounds work quite well but, they all could benefit from some increased volume levels and duration during departure/downshift/engine-braking. That alone could dramatically improve the realism of replays.

 

[Mr Whippy]

That sounds and looks nice to me.

Not sure if they are using gain control on audio and a high range volume though?! GT5 P had that half a decade ago nearly so AC should be able to do something similar. Anyone know for sure?


The big missing factors there though are ambience. I'm surprised there are no tweeting birds, traffic noise, wind noise, wavy grass and trees.

Lets hope they have some plans for those things.

Dave

 

[Mapu]

don't forget the crickets plzzz...

seriously, and with all due respect, the sounds are more than suitable at the moment and there are more important modules still missing on which kunos should focus first

 

[CKraegel]

has anyone listened to the recent onboard preview sounds of the 12c gt3 and z4 gt3??? They are near perfect! Don't know any sim, that has that good onboard sounds! And personally i don't care about birds singing or trees waving while watching a replay.

 

[DennS]

That very sentiment may be what keeps the sounds from progressing further. We can easily control volume - the bigger issue is to have enough contrast in the sounds to provide a more realistic sounding result. Environmental sound dynamics are incredibly complex - no doubt but, so are visual dynamics and look at what is being done there.

I believe that there just hasn't been enough public requests for sound improvements in racing sims to drive improvements in sound tech. That may be starting to change but, it would be a shame to have a title like AC not have some stellar sounds to match the rest of the features. Thus far, the sounds seem to be decent but, not near the level of RRRE or iRacing's new sounds. Even rF2 sounds work quite well but, they all could benefit from some increased volume levels and duration during departure/downshift/engine-braking. That alone could dramatically improve the realism of replays.

I agree with you that there is still a lot of possibilities to improve the sound in racing games as I said before but it is not as "dramatic" as you make it sound... And you really can't control volume because I don't want to wear earplugs , sound is such a huge part in (real life) racing and there is nothing on this planet that compares to you standing at a circuit hearing a Ferrari V12 (or something similiar ) screaming past you, nothing......

Sure RRRE and iRacing's exterior sounds are great, especially Simbin games always had great sound... But when we talk about interior sounds where we actually heard real ingame sounds (not filmed with an iPad) I'd say AC sounds better than iRacing simply because of the additional effects which add a lot to the atmosphere of sitting in a car, much like in Simbim games...

 

[RaceNut]

I think that most of the in-car sounds in most of these games are very good, it's the replay sounds that haven't had as much attention as they deserve IMO. The in-car sounds of iRacing still standout as being the best to date IMO. They are very clean dynamic sounds that represent the RL sounds quite well and the replay sounds have more rasp and power in the new samples.

I would gladly trade bird, cricket sounds and the like for better engine/exhaust sounds in replays but, that's just me. The sound that racing engines make is music enough for me.

 

[philippeapo]

I think that most of the in-car sounds in most of these games are very good, it's the replay sounds that haven't had as much attention as they deserve IMO. The in-car sounds of iRacing still standout as being the best to date IMO. They are very clean dynamic sounds that represent the RL sounds quite well and the replay sounds have more rasp and power in the new samples.

I would gladly trade bird, cricket sounds and the like for better engine/exhaust sounds in replays but, that's just me. The sound that racing engines make is music enough for me.

Inside perhaps, because outside (like AC mp4 12 c) i can't ear your music ;-)

 

[RaceNut]

I can't say I have heard the McLaren sounds first hand but, the Cadillac, Corvette and most of the other V8 have very good external sounds. I'm not claiming that they are ultra-realistic but, they are a step in the right direction. The new Z4 external sounds in RRRE are also very good. All need more intensity on the departure side though.

 

[Hannes-Joshua Thor Barthel]

That sounds and looks nice to me.

Not sure if they are using gain control on audio and a high range volume though?! GT5 P had that half a decade ago nearly so AC should be able to do something similar. Anyone know for sure?

gain control? whats that? every sound that goes louder and quiter has "gain control" and "high range volume" ??? what does that mean ? high dynamic range? or the opposite heavy compression to get quiter sounds louder and let them duck away through louder sounds?


never heard that terms especially high range volume? gain control I heard but thats just make somehting louder or quiter lol for example volume automation of a singer or such things...

sounds to me as a soundguy like some simple things putted into a good for marketing bla bla word...



to the the other point I read here.. like the comparsion of athmospheric graphical gimmicks and such things...

it's alot differend with audio... yes ofcourse they could add bird sounds and crowd sounds and all that kind of stuff.. thats very very simple...

but for example... things like shadows or reflections and so are in the graphical thing quite easy... because the game is in a graphical 3dimensional world... that is nowadys possible through the immense powerfull graphiccards...

but we don't have the hardware yet to provide that same 3dimensional world in sound... you would need a sound processor/card which is as strong as a graphic card for that... which is not available...

soundwise we are in the time where the graphics were in the early / middle 90s ... we don't have the dedicated power to built a complete 3dimensional world of sound...

so you need to do a lot of tricks.. for example because you can't built a complete 3 dimensional sound world... you need to trigger differend "canned" effects like reverbs and echos at differend parts of the track... that works for example for the driving view...
because in the driving view you are at the sound source... so you travel together with the sound source so you can trigger these effects...


but for replays you need a complete differend system...
in replays with tv cams and so... you are not the sound source... but the sound source is moving somewhere and so should the soundwaves do.. till the ear or microphone receives them...
and there it's getting very complicated... because the sound of the moving source spreads in all directions... for example... the source (car) come through a turn with a wall behind it... and then driving through an area with trees towards you... with for example another wall at the next turn.... now you would need to hear the echos of the walls but delayed in relationship to the sound source and the listener.. and you also need the flattering echoreverb of the trees.. but they also need to stay in the right relationship....

if you would just fade through these effects it would sound wrong... because you are not the source so you need to have a dynamical delay including filters for air dampening and so on... and if you want to do it very correct you would also need to get the reflections of the between the walls and the trees too... which are then no longer from the "actice" sound source (the car) but from passive one (the reflections from the trees to the wall)...

to get a picture you could imagine a cannon on a car which shoots alot ALOT of gum balls in all directions.... and now you see where thy are bouncing off every surface (forget about ballistics in this case )

that would be the correct 3d sound world... (which is like I think you can understand not possible on todays hardware)

so for replays it's very hard to trick around in a way that sounds at least a bit realistic...
one solution would be to forget about distance and sound travel... so that the delays of the the walls and trees just stay all the same all the time...

so you would need to do a fixed reverb echo "scene" for every cam position... which just blends with the dry car sound... a pre fader send for example.... for people who are not into sound...: the source (car) is sending the sound into the reverb generator before the actual car sound is volume controlled... so if the car is far away and the volume of the car is still quite low... you still hear x amount of reverb... and when the car is near.. and the volume of the car is louder.. you still hear x amount of reverb but the ratio between the reverb and the car is now differend... so lets say... reverb sound is x and to put it into a random number that is now 50..... when the car is far it is for example 10... so you have 10 amount of car.. and 50 amount of reverb.... if the car is near it's for example 90 then you have 90 amount of car and 50 amount of reverb... and while the car travveled away again... it's again far let say again 10 amount of car and 50 amount of reverb....
that would be "physically" more or less correct (if you forget about the real physic like sound travling delay etc)...

the thing is... our ears work a bit differend they are compressing at high sound levels.... so even if that would be some kind of physically correct.. there would still be a little touch what could make it more exciting....

for that there are now 2,5 additional solutions... one way would be to also controll the send amaount... or wet/dry level... so if the car is far away you have for ecample 50 reverb... if the car is near you only have for example 20 reverb and if it is far again you have again 50 reverb...

secound solution would to put a compressor on that... so the reverb gets "compressed away..." by the sound level of the car... that could sound more intense in some cases but could also couse a bit strange effects at volume peaks like gearshift bangs for excample...

I think a combintion of the both would get a quite good result... so you let the send amount or wet dry amount.. controlled by the distance of the source.. (car works like somekind of fader)
and then put a tiny tiny bit of compression on the summing of that..

so ... and now... you have only one car.. so you need to make a channel for every car (but ok well every car already has it's own channel) and every car is it's own send knob/fader to the reverb generator...... but would work I think....

but like you see that is quite differend to what you can do with graphics... because all the reverb thing and so is canned... it's like blending a texture in and out... because there isn't the power to do a real 3d sound enviroment yet...

however if that is done right.. (good reverb echo modeling and such things) I think it can sound quite good... however a good reverb generator alone is not very cpu friendly.... ok in this case it doesn't need the abbility to do long creamy spacey reverb tails... it just need to do reflection stuff... so it doesn't need to fight with long clear reverb tails (which are often the most cpu consuming part of reverbs)

then there is the choice between convultion reverbs and algorithmic reverbs... (convolution is you could say kind of a audio photo of a scene and alghorithmic ones are processed ones)... I would choose algorithmic over the convultion ones... because you don't have to travel to a specific place and build up a sound and recording system to get that "photo" (impulse response) .. with a alghortimic one you could just programm it till it sounds quite right... so you are much more flexible..


however... long post again like everytime some one feeds me with sound things


but well... I don't think AC gets that kind of sound FX implemantation... if it would .. I would love to the reverb modeling !

 

[Stereo]

Heh, your long post has got me wondering how much CPU power a 'pretty good' method would need, where you limit it to a finite number of moving audio sources with their own reverb attached, that spawn new sources when they hit some predefined objects, and maybe limit it to fast reverb methods (iirc convolution is as fast as the FFT, so a bit slower than O(n) where n is the length of the impulse response). You do have a bit of room to do computations because the speed of sound delays arrival at the camera, so maybe there are tricks to do it efficiently. For example, as soon as a source plays, you can calculate how loud it'll be at every object it echoes from, and when it'll arrive at them.

Seems like it's mostly important to have good samples in the first place - if it doesn't sound like an MP4-12c, then adding careful reverb of the track won't do much for that. Not sure what an adequate collection of sounds are to reproduce the original accurately - really you want to know what it sounds like at any set of inputs (rpm, throttle, load, etc.), with an ability to transition smoothly from one state to another.

 

[Mr Whippy]

Just like everything you do enough to convince the viewer and nothing more.

Josefhead, by gain control I simply mean that the volume of a car at 10m away should be able to vary depending on the other sounds.

Ie, if you race a Prius vs an F1 car, you probably won't even hear your Prius in real life... how do you do that in a sim though? The Prius would have a volume setting so low as standard you wouldn't hear it.

So you simply have sounds that all add up for a given observer, that might be really loud with a grid of F1 cars, then you divide it down to fit into the normal volume range you want.


We do this with HDR now, we crunch the HDR lighting of real life into a monitors 255 instances of brightness (ish)
We should be doing it with sounds now.

We can then tweak the way we compress the wide ranges of 'reality' into our simulated versions in different ways to suit our speaker set ups, our monitors, our preference, and even to suit what is going on.

Ie, for a replay camera we can make it look/sound like a real video camera and audio.

For the driving camera we can make it respond more like our own eyes/ears!



Stereo, ideally you need full load, zero load, and full braking load, then you can check the engine state position 'in sim' and just interpolate between them.

The problem is recording all those noises hehe.

Another thing is perception though, and that comes back to how you crunch the real value sounds into 'simulated' desktop speaker ranges or for headphones etc.
GT5 to me sounds very right, but many people have a problem with it. OK some car samples are terrible but plenty are very good and sound great.

So how you balance the sounds is important. For me in my road car everything but the engine is dominant until I put my foot down, at which point I don't hear the tyres, wind noise, or anything else.
There is no way to do that in most games these days because we have this narrow flat dynamic range we work within.


No signs yet if the AC system uses an un-bounded audio environment and then compresses it or not. If it doesn't then the sounds will always feel a bit 'flat' no matter what you do imo.


Dave

 

[Hannes-Joshua Thor Barthel]

Just like everything you do enough to convince the viewer and nothing more.

Josefhead, by gain control I simply mean that the volume of a car at 10m away should be able to vary depending on the other sounds.

Ie, if you race a Prius vs an F1 car, you probably won't even hear your Prius in real life... how do you do that in a sim though? The Prius would have a volume setting so low as standard you wouldn't hear it.

thats simply compression and masking if a sound in the same frequency range is 18db quieter you won't hear it... things like that are called psycho accoustic (also things like longer sounds are received differend in volume in the brain then shorter sounds and so on.... btw for the masking thats one point of how the mp3 codec works..... so if you would have the differend levels for example very high volume for f1 and very low volume for prius thats just masking... but because we don't want to crank up our audio systems to volumes like 130 db spl (even if that would be the most accurate and authentic way lol) compressors are used so you still hear quiter sounds.. but if a louder sound comes it reaches a threshold at which the sound "is stopped of getting much louder" (this example explains more a limiter but it's basicly the same.. a limiter is also a compressor just with high ratios)... so the louder sound start to "push " the quiter sound away.. because the compressor is reduces the gain

So you simply have sounds that all add up for a given observer, that might be really loud with a grid of F1 cars, then you divide it down to fit into the normal volume range you want.


We do this with HDR now, we crunch the HDR lighting of real life into a monitors 255 instances of brightness (ish)
We should be doing it with sounds now.

that is very very common to do that with sounds...it's done since radio transmission was invented but that was to keep the bandwith low..
but also in an "artestic" and "creative" way it is already done long long time...and in the late 80s early 90s a loudness war was released.. which ended up in having us nowadays a worse dynamic range then these old gramophones things (that turntables with that horn thing)

but yes if it's done right compression is a wonderfull tool to get a sound alive and intense

and yes most games could do a bit more with compression... but some games are already extremely let me say experimentation happy like for example battlefieldf I think it startet with bad company 2 .. that wartape option... it was a very very heavy compression... with heavy pumping release times...but they have done it better in bf3 where it doesn't pump so extreme... however even if I would throw it directly into the garbage can if it would be a music thing... it was quite exciting and intense in an action war game...

however overdone compression will do very much the oposite.. it will make the sound lifeless and booring then no compression is often better.. you need some kind of dynamic to bring the sound alive... and usually the louder you can play / hear... the less compression is better ( more intense) ... for example... the differend versions of movies... in a cinema you often have a very big dynamic range so you hear when they talk quite good... like a usual talk volume and when an explosion comes it's getting realy loud at that moment... and at home.. then you wether have less dynamic range.. or you have that.. oh they talk make the volume up... ohh explosion turn the volume down

but yes like I said you can add some intensity with compressors

We can then tweak the way we compress the wide ranges of 'reality' into our simulated versions in different ways to suit our speaker set ups, our monitors, our preference, and even to suit what is going on.

Ie, for a replay camera we can make it look/sound like a real video camera and audio.

For the driving camera we can make it respond more like our own eyes/ears!

well tweaking by the end consumer? that won't work well.. you need somekind of knowledge and experience to dial a compressor in so that it sounds good and right and intense..

but you or they could built presets for example.. desktop speakers.. hifi speakers.. headphones and such things... and maybe let the tweaking also open for modding guys too

Stereo, ideally you need full load, zero load, and full braking load, then you can check the engine state position 'in sim' and just interpolate between them.

The problem is recording all those noises hehe.

Another thing is perception though, and that comes back to how you crunch the real value sounds into 'simulated' desktop speaker ranges or for headphones etc.
GT5 to me sounds very right, but many people have a problem with it. OK some car samples are terrible but plenty are very good and sound great.

So how you balance the sounds is important. For me in my road car everything but the engine is dominant until I put my foot down, at which point I don't hear the tyres, wind noise, or anything else.
There is no way to do that in most games these days because we have this narrow flat dynamic range we work within.

the dynamic range is in almost every game the same.. 16 bit (which result in 96 db dynamic range (cd standart) or maybe some have already 24 bit or maybe even 32bit... however doesn't make much a difference because you mostly don't play at peak level of 96 db SPL.. (maybe with headphones... but 16 bit is already enough to get a very dynamic sound) ...

so we don't have that narrow flat dynamic range... we have indeed the opputunity for a very big dynamic range.. the problem is if you want hear quiet things... you need to crank up the volume.. that is ofcourse unpracticable.. so you need to reduce the dynamic range.. I understand now that the gt5 guys mean with high dynamic range sound... it is indeed low dynamic range sound but that ofcourse doesn't sound good as marketing bla... so they twist the words into something like they have crunched the high dynamic range into a low dynamic range...
and that is like I said above very very simple... just put compressor on top and tweak it to the right thing...

they just need to level the sounds quite right in game.. so that not everything is at the same volume (also called normalizing) .. but to have ingame a with out a compressor a quit dynamical sound... so like you said for example wind is quiter then a big hp engine on throttle etc.. and then let the compressor at the end do the rest...

from a dynamical point of view that is possible since 10-20 years... since 16bit became standart... from processiong power lets say 10 years... the thing is most developers just didn't do it yet.. or if they did only a little bit to prevent it from clipping and such stuff.. but didn't used it creativily
[/quote]

No signs yet if the AC system uses an un-bounded audio environment and then compresses it or not. If it doesn't then the sounds will always feel a bit 'flat' no matter what you do imo.


Dave

don't know what you mean with un-bounded audio environment.... it's just leveling of the sound and add a compressor at the end...

and at least leveling of the engine stuff can already be done in the tp in the ini file for the car sound...

the thing what they would need to do is leveling it quite dynamical and then add a compressor (sorry for sounding repretive but were the words that fitted )

if you have a sound card and a software where you can route your audio in and apply effects.. you could already do that compression part on your own...

for me for example it would look like that way if I would want to do it.. I route the sound output of the game into a mixer channel of my DAW... and then put a compressor on it... like I want...


ofcourse the leveling of the individual sounds like tires.. engine on and off.. wind and all that stuff.. is needed to be done in AC


and ofcourse that external DAW compression workaround thing is nothing to present as idea for a final release/solution... it's just an example of how you could do a compression if you miss one...


and btw... all that summing compression part and so.... is some kind of mastering... and mastering.. is always the very last step of audio work



and believe me or not compression may be a nice additional polishment (that is was mastering was/is intended to be ... not that loudness madness war)

but nothing makes a sound more alive and plastically then echos/reverbs/spatial processing...

you could take a very simple booring beep.... and create with reverbs echos reflections and such stuff a very alive and plastical sound out of it

because that is how we receive our enviroment with our brain... all the time... reverbs/echos/reflections are everywhere... thats how how brain judges deepness or distance (+ additional frequence spektrum alternations) and directions...

our brain is "calculating" (not realy calculating more kind of sorting) all these differend tiny echos and such stuff all the time... and so a sound without proper reverberation/reflection and such stuff.. sound booring flat and lifeless... because in real world there is almost no sound without reflection... even on an open field you still have the bottom ... thats also why many people get this pressure kind thing of feel in the ears when they get into a heavy accoustical threated room.. because it feels like they are floating in the air...because they doesn't get the reflections they are used to.


however I am drifting off the track here

however compression (in other words high dynamical sounds crushed into a low dynamcial enviroment) can be nice... but reflections are always more importend for the realisticness of sound


and now I got into the mood to do some very heavy sound design on reverbs and refections

 

[Hannes-Joshua Thor Barthel]

I just did a fast audio post processing to the mclaren vid... ofcourse it sounds totally stupid at the moment because at the moment the is an ipad in a room driving around an simplicified reverberated imola... so it sounds very metallic at the moment because of the dual reverberation (the room with the ipad in with the recording of speakers in a virtual reverb world ) and it was only 1 hour of work...
it may would sound a little bit better if I would use my highend reverb.. but because of the instances I think that wouldn't worked on my cpu ( at least with my first plan of doing it 1 minute long... with only these 24 seconds it may would worked) with this kind of reverb it consumed about 18 % of my I7-975

(only did the first 24 seconds or so... would be too much work now to do the whole thing) ... I created very simplicified virtual spaces for every cam view... ( I used a reverb tool where you can draw walls and objects and such stuff... but only with a global high and dispersion amount .. so a tree for example has the same values like a wall) but I tried to get the walls turns and forest stuff more or less (maybe more less then more ) in place

through that space I let the source travel and also the cam rotate the "head" to follow the car...


(and yes I did a bit (in usual master and musical terms quite alot ) of compression at the end it sounded a bit deeper without compression but I needed to do that because other wise that ipad engine of the mclaren became too annyoing )

https://www.dropbox.com/s/axke4hgu9ytd2ee/mclaren reverb.mp3

don't know if it is helpfull to mute the youtube clip and play it at the same time... or if you also get the idea of it

and yes like I said it doesn't sound good.. it's like photoshopping a very dirty pixelated image and try to do a clean beautiufll picture of it

 

[Mr Whippy]

Hi Josefhead,

Yes, running a compressor over the final sum of the volumes is basically what I mean.

I'm not sure of all the different processes or effective programatic ways to achieve compression in real-time games but needless to say the end result is that quiet sounds appear loud (like wind and birds tweeting), right up until the F1 car comes flying past at which point they are essentially at zero volume, crushed out of the way by the F1 cars volume

Ie, do we just check volumes before rendering to the 16bit data waveform, or is it cheaper to write our samples * volume to a 32bit float and then compress back to 16bit? Such things I guess are down to the programmers hehe.



I agree if we crank our speakers up that is the ideal situation but it's often not practical at all as you noted. So we compress more/less depending on the speaker set up etc.

GT5 allows you to choose the speaker set up and I assume it compresses differently based on assumed volumes you run with those different set ups!




I think it's fairly important for good sounds to run a compressor.

It's just as important as how we tone map graphics from real life eye hurting brightnesses that blind us down to white on our monitor that isn't bright at all.

Ideally we'd have monitors that could blind us but again it becomes impractical for us so we tone map



I like the idea of reflections but I have no idea how you'd begin to author such things from a storing the data and creating realistic values, through to processing all the info.

For me I think it's probably enough to flag volumes of space on a track with reverb properties so as cars pass between different areas the sounds are treated differently. That way you are authoring much like you have in your example mp3 above. You roughly apply a sensible reverb to a given zone based on the properties of that area...
I assume this is how almost all current game engines work? A large marble corridor gets high reverb and a big delay, a small carpeted room gets almost zero of each.


Iirc the GT5 replays on the Swiss track seem to do reverb adjustments as you pass each pillar through the tunnel so you get a whooshing effect past each pillar, at least in a road car. A race car may sound different again as it has much more dominant engine/exhaust noise



I also think for replays at least the time taken for sound to travel is important and noticeable, but not sure how hard that would be to apply. Ie, you see an event well before you hear it on a long lens shot of cars approaching.



In any case there is loads of potential there but as long as they cover the basics I'll be happy. Given GT5 P has had reverb zones and compression on audio since the last half decade nearly I think it's reasonable to expect a game like AC to run such features

Dave

 

[Ben Lee]

Speaking of sounds, will AC model wind and road noise from external "cameras"? The one thing I notice from listening to cars out on the roads is that 90% of traffic, you can't really hear the engine or exhaust note, but what you do hear is a "whoosh" of air + tyre noise as the car goes past you. Stand by a busy fast road and actually mentally note what you can hear, its pretty much "whoosh, whoosh, whoosh" and so on. On the odd occasion you will get a sports car or something, then you get a combination of road + tyre + engine + exhaust noises. Would be great for TV cameras if this kind of thing was modelled.

 

[Hannes-Joshua Thor Barthel]

Hi Josefhead,

Yes, running a compressor over the final sum of the volumes is basically what I mean.

I'm not sure of all the different processes or effective programatic ways to achieve compression in real-time games but needless to say the end result is that quiet sounds appear loud (like wind and birds tweeting), right up until the F1 car comes flying past at which point they are essentially at zero volume, crushed out of the way by the F1 cars volume

Ie, do we just check volumes before rendering to the 16bit data waveform, or is it cheaper to write our samples * volume to a 32bit float and then compress back to 16bit? Such things I guess are down to the programmers hehe.

The function of a Compressor is a quite simple thing... if sound gose over a threshold x.. then turn volume down for y amount of time.. with the ratio of z so the output volume is only x+z

(ofcourse plugin developers do alot more nowadays to archieve a good sound etc... they go so far to modell a complete hardware like cuircit to this or that result in terms of curves and responses etc.. but the very basic function of a compressor is just make it quiter if it would get too loud)

thats an example for the usual used downward compression there is also the otherway around of upward compression.. which means if sound is under threshold make it louder


the volume sample thing is.. for just samples in a game 16bit is ok.. the question is more how the internal audio engine works... most DAWs for example uses 32bit floating point for the mixing enviroment...
in a 32bit floating point.. you will never get quantizing noise at very very low volume sounds.. and also won't get clipping at high volume sounds... because you have over 1000db of dynamic range.. and the area we work in is set in the middle.. (which uses the dBfs (fullscale) which came from the 24bit area where you had 144dB dynamic range..) so you have your fullscale meters somewhere in the middle and get around 500db of headroom before clipping and also get around 500 db befor quantizing noise would occour...

the only problem is then the master bus because there it clips ofcourse again at 0dB because we don't have amplifieres which could amplify a sound to infinity

however I don't know what is common to use ingame as mixing engine.. but I think nowadays a 32bit float mixing system should be standart.. however even if not.. if it is still a 24bit system you would get 144db of dynamic range (which you could only achieve with very big PA systems... in other words it is a range from totally silence to starting jet plane (in a few meters distance) and even if it is still 16bit.. it wouldn't be so worse how it might sound at the moment.. because you don't need the full dynamic range like in reallife to archieve good results... so you don't need to play a bird sound at for example 40db and a F1 engine at 130db and like you see the difference between bird sound of 40db and f1 engine of 130db would still be 90 db (that would still be in the 96db range of 16bit.. ofcourse you would need some headroom for modulations so the bird is for example sometimes a bit quiter and the engine at some point louder and so on... so it can exceed the 16bit yes... but like I said.. you don't need to work with the true levels ... otherwise you wouldn't hear the birds or if you start the f1 engine you would fall off your chair
ofcourse there also comes the compressor into play but you can already do a quit balanced mix before... the importend thing is to still have a good dynamic mix so the birds and such stuff are alot quiter.. then the engine... but they don't need to be as quite as in realife...

however.. there is another problem which occourse with a 16bit mixing system.. the problem is all sums up.... the more sounds you have the more dynamic range you need to still be able to do a dynamic mix... because more sounds means more volume ... and if you don't want to end up with having every individual sound at the same low volume level..(which would also kill the dynamic between those sounds ofocurse) you need more dynamic range

so in other words.. with all those differend sounds in todays games.. at least a 24bit mixing resolution if not already 32bit (which is in audio work already standart since the late 90s) should be standart today for the mixing stage...
but like I said that's just the mixing...

for the individual sound samples.. 16bit (or 96db of dynamic range) is enough... if one sound would vary more then 96dB I would be very surprised ( it's more or less like the range between absolute silence and a jackhammer)...

 

[Mr Whippy]

Yep, that is all pretty much as had I assumed it would work.

16 bit samples and then 32bit mixing environment, and then compress back into 16bit space once you have the mix you want.

I suppose you need only check the volumes once every 100ms or so and simply have the compressor targetting the final volume at the sample rate (say 44.1khz in this case?!)



It really is important imo to have audio that does this though. In the absence of speakers that can play very loud to capture the range reasonably without compression, then you need compression to give some realism to the sounds cars make as they go about their business


I'm glad you appreciate the benefits anyway, lets hope that AC does indeed do this because not many sims have thus far as far as I can tell

Dave

 

[Hannes-Joshua Thor Barthel]

Yep, that is all pretty much as had I assumed it would work.

16 bit samples and then 32bit mixing environment, and then compress back into 16bit space once you have the mix you want.

I suppose you need only check the volumes once every 100ms or so and simply have the compressor targetting the final volume at the sample rate (say 44.1khz in this case?!)



It really is important imo to have audio that does this though. In the absence of speakers that can play very loud to capture the range reasonably without compression, then you need compression to give some realism to the sounds cars make as they go about their business


I'm glad you appreciate the benefits anyway, lets hope that AC does indeed do this because not many sims have thus far as far as I can tell

Dave

The audio is checked at samplerate...(so at 44.1khz every round about at 0.00002 seconds) otherwise it can't work... for example.. if the detector of a compressor would only measure once in a specific time.. it could measure at that point a "wrong" value..well the value is right for that specific moment.. but a soundwave changes always it's level... otherwise it would be silence (for a "picture" if the membrane doesn't move there is no sound.. no matter if it is pulled out or in or centered in that moment) .

if the detector now just measures a specific moment... then it could be for example not the peaklevel of the wave or worse it could be at zero at that moment...

so it needs to constandly "scan" the audio and so on minimum it has to works at the samplerate...

the rms window (to avarage the soundlevel detected over specific time for not to so drastical changes) and the attack and release phase of the applied gain reduction is another thing...

there for example is 100ms for the release time quite ok I think.. for this kind of application...

and btw as you see at the time of the samplerate... 100ms is ALOT in audio... 100ms is a frequency of 100 hz... there you still have 21500 hz left of the audio spectrum (at a 44.1khz rate)

 

[Mr Whippy]

Ah ok thanks for the clarification.

I assumed it was all sampled at full rate inside the driver/sound system, but from an end user POV using something like fmod it's probably just a few compressor coefficients then that determine the effect in the end?

Ie, unless the author is making the full audio system from scratch they'll just interface through an already developed system (again which I guess most game makers do, I know fmod is popular for instance)


I wonder how AC is doing all this stuff then... would be nice to know just out of interest

Dave