R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 3 [All]   Go Down

Author Topic: Why different results when mixing down ITB with different interfaces, buffer settings etc  (Read 7989 times)

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Hi this has been bugging me for years. I've been mixing down my sessions using different buffer settings on the DAW and each setting makes the resulting track sound different.  A low buffer setting on one soundcard makes everything sound more forward but unnatural, whereas a high setting is more laid back but blurry.  Then in order to check if it's just my soundcard I tried to mixdown the identical session using four other soundcards at different settings.  The result is every soundcard yielded different sounding mixdowns even though the mix was all digital.  For example an EMU 1616m with Protools HD converters and was the least wobbly but also the narrowest and harshest.  Using the laptop's soundcard the sound was less wobbly with strong lower mids, but it's less 3D and the highs are rolled off.

Also I often "mixdown" mono files for each track within a session to save CPU and make things neater like a tape machine.  And it often sounds better too for some reason.  I've found that different buffer settings and also different soundcards affects the resultant file tremendously.  I've played back the tracks using different soundcards and you can always hear the characteristic of each soundcard.

Are these issues caused by jitter?  I was always under the impression that clocking or jitter only matters when conversion is occuring, but maybe I'm wrong.

Any wisdom would be greatly appreciated.  Digital is supposed to be more convenient but now it seems more straightforward to me when I was tracking and mixing on tape.

Andy
Logged

Waltz Mastering

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 912

Have you tried identifying the mixes that you bounced at different buffer settings and using different sound cards in blind listening and/or null test?

Patrik T

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 833

I'd advice you to run test tones with different buffer settings on different cards and measure the output voltage on the DAC in use.

This is about the only way to actually confirm that:

- there is a problem. If so, voltage would alter with altered buffer. Does it?

- the different DAC's might output different voltages for the same digital ref. level. If so, audible confusion might arise.


Good Luck
Patrik
Logged

Greg Reierson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 425

Have you done null tests?

There is NO jitter ITB.


GR
Logged

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Yes we've done blind tests with different people in two separate studios.  Also did a blind test at Metropois Studios in London with a mastering engineer called Tim Young.  We played him three digital mixdowns using three different soundcards and he heard the difference too.  He actually preferred the mixdown with the cheaper Tascam FW1804 soundcard, it was more grainy and less sterile (but wobbly).

At our studio we did one test just mixing down files with different buffer settings on the same soundcard (and repeat with alternate soundcard).  Then another test using 5 different soundcards on two separate computers.  We mixdown the same session on the same laptop with different soundcards engaged.  Then we listen back to the different mixdowns from the different soundcards on the same soundcard.  

The most disturbing is when we mixdown a mono track to a mono file.  The results are different with every soundcard.  Some lose bass, some add harshness (digidesign!), some lose depth (the laptop's own soundcard), some add bass (Native Instruments).  One thing we did find is all the firewire interfaces sound a bit blurrier than the USB, PCI, Cardbus cards.  Could be a coincidence, but maybe the firewire card can affect the sound?  Also noticed that the soundcards with Cirrus Logic branded converters tend to be less harsh than AKM, but again could be a coincidence.  The converters shouldn't matter anyway as we're mixing in the box?
Logged

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Haven't done null tests at home, but have brought it to three different mastering houses in London and they can all hear the difference.  Apart from Metropolis we also went to Alchemy Mastering.  There they could immediately tell that one mix sounded narrower than another.  Then we told them they were identical mixes using different soundcards and they were a bit miffed...
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

The vast majority of DAW apps these days involve the soundcard only in playback and recording - and don't have any processing done with the soundcard or any effect by it at all when they are bouncing to disc.

I've had this confirmed to me directly to me for the DAW app I use (SAWStudio) by the author of the app himself - and it can be easily further confirmed with null tests.

You need to do some null tests.  If different bounces turn out to be data identical then I'd say the reasons you are hearing differences are most likely due to effects detailed at http://www.ethanwiner.com/believe.html - and yes, mastering engineers are effected by this just as much (and sometimes more, as they tend to be very attuned to minor differences) as others.

Best regards,
Steve Berson  

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

Greg Reierson wrote on Tue, 08 December 2009 10:24

Have you done null tests?

There is NO jitter ITB.



But there can be jitter at the DAC, even if playback is coming directly from a DAW.

Best regards,
Steve Berson

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

That's what the mastering engineer at Alchemy said.  But we did the same mixes again right there and then (so no one can accuse us of changing the mix), one with the laptop's sondcard and another with a high quality cardbus soundcard, and the two mixdowns sounded different.  I've never done a null test and am not sure how to go about it, but we can all hear on different systems that these soundcards are making a difference to the resultant mix.  The ME at Alchemy gave us all this theory about how this shouldn't be the case, but all of us including himself could hear that one mix was narrower and harsher than the other one.  Even my girlfriend could hear the difference when played the mixes on different monitoring systems and different soundcards.  I suspect it could be something to do with firewire or usb as we do all our mixes on laptops?
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

A Deian wrote on Tue, 08 December 2009 11:07

That's what the mastering engineer at Alchemy said.  But we did the same mixes again right there and then (so no one can accuse us of changing the mix), one with the laptop's sondcard and another with a high quality cardbus soundcard, and the two mixdowns sounded different.  I've never done a null test and am not sure how to go about it, but we can all hear on different systems that these soundcards are making a difference to the resultant mix.  The ME at Alchemy gave us all this theory about how this shouldn't be the case, but all of us including himself could hear that one mix was narrower and harsher than the other one.  Even my girlfriend could hear the difference when played the mixes on different monitoring systems and different soundcards.  I suspect it could be something to do with firewire or usb as we do all our mixes on laptops?


To do a null test:
bounce to disc the same mix starting and ending at the same exact place (with in this case the single variable of what sound card or device or buffer setting you are testing).

Line the two versions up in separate tracks of a DAW, starting and ending at the same exact times.
Reverse the polarity (aka "flip the phase") of one of the 2nd channel.  
Bounce to disc once again.
If the two files are data identical the resulting file should have no audio on it whatsoever.  If not - the resulting file will show the differences.  

IF they are data identical then there are 3 reasons that are most likely why you might be hearing a difference:
* a jittery playback (generally cured by using a better DAC)
* different frequency response at the sitting position you are currently at versus the one where you were listening previously due to comb filtering in the room
* placebo effect

Best regards,
Steve Berson  

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348

A Deian wrote on Tue, 08 December 2009 09:31

Yes we've done blind tests with different people in two separate studios.  Also did a blind test at Metropois Studios in London with a mastering engineer called Tim Young.  We played him three digital mixdowns using three different soundcards and he heard the difference too.  He actually preferred the mixdown with the cheaper Tascam FW1804 soundcard, it was more grainy and less sterile (but wobbly).

Regarding the influence of buffer size, I don't know your DAW, but I use Samplitude and it's a fact that changing the buffer size changes the behaviour of plug-ins,EQ's, dynamics (in particular those with look-ahead capability) and reverbs.
In Samplitude the project buffer size can be adjusted independantly of the soundcard buffer size; maybe in your app, you don't have the choice...?
Are you sure you're mixing 100% ITB. Seems to me this could happen if you looped the soundcard's outputs and inputs for printing your mix.
Logged

Andrew Hamilton

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 573

I thought I heard a difference with two different buffer settings in PTLE during a bounce to disk.  However, a null test proved that both mixes were bit-identical.  They still "sounded" different - as long as I knew which mix was which.  But I already knew that they couldn't be (different sounding), since x=x.  );  


Andrew

Logged
www.serifsound.com
premastering for CD and DVD-A.  Featuring FTP load in and delivery as well as analog tape transfers.

subvertbeats

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 77

Andrew Hamilton wrote on Tue, 08 December 2009 21:06

I thought I heard a difference with two different buffer settings in PTLE during a bounce to disk.  However, a null test proved that both mixes were bit-identical.  They still "sounded" different - as long as I knew which mix was which.  But I already knew that they couldn't be (different sounding), since x=x.  );  


Andrew




Out of curiosity I did the same with Cubendo a couple of years back, and had the same results.
All ITB of course.
With the proviso that I dont have any of these files to test and listen to, I think Steve is on the money in this thread.

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Thanks for the helpful feedback I'll try the null test tomorrow.
Logged

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Oh and yes everything was mixed ITB.  The idea that buffer settings can change the way Samplitude reacts to plugins is interesting.  Did a lower buffer make mixes sound more "forward"?

Did a null test tonight after all.  I lined up two mixdowns of identical mixes (they were mixed to 256kpbs mp3s using the same laptop and program) and clicked on the phase switch for one of them.  One mix used a usb interface, the other used a firewire interface.

Bounced 17 seconds of audio and uploaded the result here:

http://www.sendspace.com/file/upyk62

I sincerely wish it did, but it doesn't sound like it nulled...


Ps: Just in case I also bounced two mixdowns that used the same soundcard with the same settings, and this time it nulled.

Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

A Deian wrote on Tue, 08 December 2009 20:39

Oh and yes everything was mixed ITB.  The idea that buffer settings can change the way Samplitude reacts to plugins is interesting.  Did a lower buffer make mixes sound more "forward"?

Did a null test tonight after all.  I lined up two mixdowns of identical mixes (they were mixed to 256kpbs mp3s using the same laptop and program) and clicked on the phase switch for one of them.  One mix used a usb interface, the other used a firewire interface.

Bounced 17 seconds of audio and uploaded the result here:

http://www.sendspace.com/file/upyk62

I sincerely wish it did, but it doesn't sound like it nulled...


Ps: Just in case I also bounced two mixdowns that used the same soundcard with the same settings, and this time it nulled.




What application are you using?  Did you confirm that the two mixes were lined up exactly down to the sample?  It helps to have a timing spike (i.e. a super short test tone) at the head of the file to help with making sure they are correctly lined up.  If this is confirmed to be tested correctly then sounds to me like it's time to change what DAW software you use - it's absurd to have a bounce to disc routine actually be effected by what sound card is used!

Best regards,
Steve Berson

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

One other possibility not covered yet - some plugins, such as reverbs, have randomizing elements to their algorithms - which means that no two bounces are exactly the same when using them.  fwiw - the file you posted sounded mainly like difference in the ambience.  So perhaps you're hearing the effects of this (although the fact that you didn't in the "control" track is still suspicious).  Perhaps it's worth doing a null test without any reverb plugins loaded though.  

Best regards,
Steve Berson

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

For those mixdowns we used Cubase.  The mixes do really sound like the output of each convertor though, like the 1616m mixdown sounds like the DA of the 1616m.  And the laptop soundcard's mixdown sounds like the laptop's soundcard.

We did the same test with no reverb and the mixdowns still didn't sound the same so it's not the reverb.  There's definitely a link between the soundcard's output sound and the mixdowns but I can't work out why.

Thanks for the help though, Steve!
Logged

subvertbeats

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 77

A Deian wrote on Wed, 09 December 2009 13:24

For those mixdowns we used Cubase.  The mixes do really sound like the output of each convertor though, like the 1616m mixdown sounds like the DA of the 1616m.  And the laptop soundcard's mixdown sounds like the laptop's soundcard.

We did the same test with no reverb and the mixdowns still didn't sound the same so it's not the reverb.  There's definitely a link between the soundcard's output sound and the mixdowns but I can't work out why.

Thanks for the help though, Steve!


Could you share the full Cubase project folder?
I'll happily try some tests for you and try and work out why you are seeing (hearing) these things.
As I wrote before Ive tested Cubase and Nuendo and have never seen anything like you are...

UnderTow

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 393

I am betting it is operator error combined with the placebo effect. I mean come on, someone comes here with this whole story and never once mentions which DAW it is until someone explicitly asks? Yet he does mention "Pro Tools HD converters" on his Emu card? (A marketing lie btw). Clearly someone isn't thinking particularly logically.

Why not go to the Steinberg forum and ask for some advice there? If it has really been bugging you for years...

Alistair
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

UnderTow wrote on Wed, 09 December 2009 10:51

I am betting it is operator error combined with the placebo effect.


fwiw - the file he posted from his null test showed that the 2 files are not nulling.

Best regards,
Steve Berson

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

Alistair I'm sorry we didn't mention DAW.  We've used Cubase, Nuendo for mixing, while tracking was done on Protools HD3 through a MCI desk.  I'm not sure what you mean by "someone coming up with this story" as there's nothing to be gained by making something up here.  We didn't think it's the software's fault as the same Cubase session at another studio (actually they ran Nuendo but the session is the same) with PCI desktop setup and Lynx Aurora didn't experience these mixdown problems.  The problems only arise when using firewire or USB interfaces with our laptops.  We've tried two laptops, both windows XP.  FYI we have contacted Steinberg and they said it's definitely not the software.

And The Emu card does have the same AKM converters as Protools HD.  Of course that doesn't mean it sounds the same as we all know there are other factors involved in sound quality.  In my opinion neither the EMU nor Protools HD3 sound particularly great, everything sounds inherently "modern" through them.  Especially after hearing the Prism converters at mastering.

Subvert Beats - Thanks for the offer. The problem is not session specific, it even happens with mono WAVs.  Would be happy to send you files although like I said we did a mixdown at another studio (Strongroom) with a desktop setup and there were no problems.  If you're by any chance in London perhaps I can get in touch?
Logged

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

It's been suggested that we should try a newer laptop and a different firewire card so we'll do that and see if it makes a difference. Though the current laptop is Core2Duo 2.4 with 3gb ram so it shouldn't be the issue, unless the culprit is Windows.
Logged

UnderTow

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 393

I really think you are wasting your time by coming to this forum with this problem. You need to find out exactly what the problem is and the only way to do that is to go through a very specific step by step methodology.

If Steinberg can't or won't help you, ask on the Steinberg forum. You will find knowledgeable Cubase users. (Is it Cubase or is it Nuendo? Which version? Etc etc. Why all the vaguery? If you tried both, why not make that explicitly clear?) They might know of some setting that needs changing or you might have discovered an unusual set of circumstances that reveals a bug in Cubase or Nuendo. Either way, a mastering forum isn't the right place to get trouble shooting. Ask Steinberg or ask on the Steinberg forums. (Unless you can't post there. Rolling Eyes)

It seems you have a technical issue. The only way you are ever going to solve it is by being technical and explaining every single step in minute detail. Every setting you use. Every step you made and you have to explain it to people that are familiar with Cubase or Nuendo. Not present your problem to a group of people that use a variety of DAWs that do not behave in the way you are experiencing.

Alistair
Logged

A Deian

  • Newbie
  • *
  • Offline Offline
  • Posts: 14

I'm here because I respect mastering forums, the average gy here should have better ears.  A few "engineers" we've played the tracks to can't seem to hear the difference.  They just dismiss it with the usual "placebo" effect or "you're just imagining it" quip.  Admittedly the difference is small but big enough to make a difference to our sound.  It's the same with pan laws and truncating, some people don't care saying the difference is negligible.  But mastering guys care about this stuff.

Honestly though I don't see why you insist that I'm hiding something, or "can't post on the Steinberg forum".  Sorry for taking your time.
Logged

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651

i don't know of any bug like this for cubase. it is possible if the hardware buffer is
set too low, some plug-ins which have high latency may not perform correctly. that
is the only plausible explaination i can think of at this time.

jeff dinces

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

A Deian wrote on Wed, 09 December 2009 16:13

I'm here because I respect mastering forums, the average gy here should have better ears.  A few "engineers" we've played the tracks to can't seem to hear the difference.  They just dismiss it with the usual "placebo" effect or "you're just imagining it" quip.  Admittedly the difference is small but big enough to make a difference to our sound.  It's the same with pan laws and truncating, some people don't care saying the difference is negligible.  But mastering guys care about this stuff.

Honestly though I don't see why you insist that I'm hiding something, or "can't post on the Steinberg forum".  Sorry for taking your time.


I don't agree with Alistair that your concerns aren't valid here.  Understanding what is happening "behind the scenes" in the DAW apps we use can often be critical to delivering a good mix these days.  I think it would be good to post the 2 files that you derived the file that showed lack of a good null from so that we can verify that the null test was done correctly in this case - the 17 second length is fine for these purposes.

If the null test was correctly done then it points to the possibility that Cubase's bounce to disc routines could be compromised.  Imo a well coded DAW app's bounce to disc routines should be completely unaffected by what sound card or other audio interface device you use.

Best regards,
Steve Berson

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

cerberus wrote on Wed, 09 December 2009 16:46

i don't know of any bug like this for cubase.  it is possible if the hardware buffer is
set too low, some plug-ins which have high latency may not perform correctly. that
is the only plausible explaination i can think of at this time.

jeff dinces


Thinking about it plugins that can have random results, or buffer settings that are set too low for the plugins to function correctly definitely are more likely the culprit here than a bug in Cubase's code.

Best regards,
Steve Berson

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651

hey steve;  thinking more, that should not affect offline exports;
only if one does the export in real time (optional)...

also there is a "constrain delay compensation" option in cubase... pdc can be
turned off partially ("constrained") or disabled. in that case, exports could still
employ pdc.  (would be pilot error, not a bug).

jeff dinces

Geoff Emerick de Fake

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 348

A Deian wrote on Wed, 09 December 2009 11:27

It's been suggested that we should try a newer laptop and a different firewire card so we'll do that and see if it makes a difference. Though the current laptop is Core2Duo 2.4 with 3gb ram so it shouldn't be the issue, unless the culprit is Windows.

What's the FW card got to do with it? Have you made a test WITHOUT any soundcard?
I've listened to the "null" file; it is clear that the differences lie in the reverbs; the random modulation is clearly audible.
Logged

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651

Geoff Emerick de Fake wrote on Wed, 09 December 2009 17:04

 it is clear that the differences lie in the reverbs; the random modulation is clearly audible.

A Deian wrote on Wed, 09 December 2009 11:27

It's been suggested that we should try a newer laptop and a different firewire card...

andy: from a scientific perspective: one ought to verify that exports from  
one system are identical before comparing exports from different systems.

jeff dinces

UnderTow

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 393

TotalSonic wrote on Wed, 09 December 2009 22:51


I don't agree with Alistair that your concerns aren't valid here.


It is not that they are not valid here. It is that this is not the best place to solve the problem. This is first and foremost a technical problem with a technical answer. Not a listening problem. It can not even be determined by listening. Even if people claim they hear a difference, you still don't have any proof of anything!

Before even looking at the exported files, the first step is to look at the export methods. That is step 1. If it can be determined beyond a doubt that there are no errors in the export methodology then you can move on to further tests.

Step one hasn't even been touched on let alone determined beyond doubt. The fact that I seem to be the only one even questioning this shows that this is the wrong forum...

But OK, fair enough, lets assume there are no errors... Then you move on to the next step in a process of elimination. Most DAWs have a way to load a project without any plugins loading. So load the project without plugs and bounce with various sound cards. Does it null now?

If so you know it is a plugin issue and not Cubase as such. So you keep on working at it until you know what the problem is. You test on different systems. You test with different projects. You test with single files, two files, ten files,etc.

If you discover that there is a reproducible bug in Cubase, you contact Steinberg and maybe make a post on the Steinberg forums to see if other people can reproduce it too. If other users can and this is indeed a bug that affects many Cubase users, Steinberg release a patch in no time because of the uproar this kind of flaw would create. (Hopefully).

Going to a forum because you trust people's ears is a complete waste of time when the issue is a technical one. Ears can not be trusted. Certainly not anywhere near as much as good old methodological testing and a good process of elimination.

Alistair
Logged

TotalSonic

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3728

UnderTow wrote on Wed, 09 December 2009 19:14

TotalSonic wrote on Wed, 09 December 2009 22:51


I don't agree with Alistair that your concerns aren't valid here.


It is not that they are not valid here. It is that this is not the best place to solve the problem. This is first and foremost a technical problem with a technical answer. Not a listening problem. It can not even be determined by listening. Even if people claim they hear a difference, you still don't have any proof of anything!

Before even looking at the exported files, the first step is to look at the export methods. That is step 1. If it can be determined beyond a doubt that there are no errors in the export methodology then you can move on to further tests.

Step one hasn't even been touched on let alone determined beyond doubt. The fact that I seem to be the only one even questioning this shows that this is the wrong forum...

But OK, fair enough, lets assume there are no errors... Then you move on to the next step in a process of elimination. Most DAWs have a way to load a project without any plugins loading. So load the project without plugs and bounce with various sound cards. Does it null now?

If so you know it is a plugin issue and not Cubase as such. So you keep on working at it until you know what the problem is. You test on different systems. You test with different projects. You test with single files, two files, ten files,etc.

If you discover that there is a reproducible bug in Cubase, you contact Steinberg and maybe make a post on the Steinberg forums to see if other people can reproduce it too. If other users can and this is indeed a bug that affects many Cubase users, Steinberg release a patch in no time because of the uproar this kind of flaw would create. (Hopefully).

Going to a forum because you trust people's ears is a complete waste of time when the issue is a technical one. Ears can not be trusted. Certainly not anywhere near as much as good old methodological testing and a good process of elimination.

Alistair


Alistair -
Definitely agree with the points in your post.  I just think it's important to keep this a friendly place and point people towards testing methods that will help - and also which I think this last post of yours in fact does.

Best regards,
Steve Berson  

subvertbeats

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 77

A Deian wrote on Wed, 09 December 2009 16:52


Subvert Beats - Thanks for the offer. The problem is not session specific, it even happens with mono WAVs.  Would be happy to send you files although like I said we did a mixdown at another studio (Strongroom) with a desktop setup and there were no problems.  If you're by any chance in London perhaps I can get in touch?


Hey, no worries. Im about an hour outside of London.

But let me understand this correctly.
You are saying that a project with a single mono wav, no plugins, bounced to disk in Cubase, twice, each time using a different audio interface driver, will produce 2 different sounding files that dont null?

So here we have ruled out everything BUT the Cubase audio engine and bounce-to-disk routines.

I must say I have never experienced any such issues, but maybe you have found a genuine issue in a particular version (though you say this has been bugging you for years, so assume you have used different versions over those years?)




dubwise

  • Newbie
  • *
  • Offline Offline
  • Posts: 38

Cubendo definitely involves the sound card drivers in the mix.
A few years ago there was a bug that produced random buffer-length silence in the output file.
One workaround that always worked was to switch to the Mac's internal soundcard for the bounce.
Logged
Pages: 1 2 3 [All]   Go Up