keithl

I have hundreds of songs on my Jukebox that I imported from SD card. I have the MMI set to random, but I am rather disappointed in how random the playback is. First it seem it randomizes the songs every time you start the car and then I find I routinely hear some of the same songs rather repetitively days to day vs. others that I never hear. I would love to see an option to either randomize the list once when new songs are imports or a instant shuffle button.

Maverick61

Keith, you are not the only one. I wonder what algorithm Audi uses to randomize things. Like you, I noticed the same things, some songs never come up, others I can hear several times over the course of a week

qqq1970

I use it occasionally and it's seems to work.

Alphatero

I have the same problem on my outgoing A4 (new A6 on the way).

The way to make it more random, is to start your play on a different song every time (just pick one manually and then randomize). This will change the order that the Audi algorithm comes up with. Otherwise, yes, you will hear the same songs over and over.

RobC

Ditto on the S6. I too have found that there are some songs that just never get played (over a period of a year), some that get played more often, and selections from within the same album might get played (at random) within an hour.

DeerHunter

People have complained about randomness (or lack thereof) since the early days of the iPod. The problem is not one of true randomness (well, a computer can only generate pseudorandom numbers, but close enough) but our perception of what is random.

To wit: If you asked a person to randomly dot a blank page with a pencil, they will strive mightily to distribute the dots fairly evenly across the entire page. This is not random. If the dots were truly randomly placed, you would typically end up with small clusters of dots on some areas of the page. These clusters, plus the fact that humans are hard-wired to see patterns in nearly everything, are the reason we perceive a random event to be non-random. We also tend to remember when a song is played multiple times, but forget that dozens of "new" songs played in the interstitial time.

Apple got so many complaints about their random-play algorithm that they actually made it less random (e.g. it would not, ever, play back the same song twice in a row, even though this is okay for truly random events). What is now less random is perceived as more random - an okay outcome for a consumer device.

Personally, I think that there should be two random options: The first is true random (or modified random, as described above); The second being a random playing of the entire playlist, with no repeats until every song has been played. Unfortunately, unless Audi programs the latter option, we're stuck with hitting the track-skip button on songs we've heard recently.

keithl

I think the way random is implemented is what is flawed. If it uses a similar algorithm each start up it will come up with similar results. I agree with DeerHunter in that we need 2 options, "Random on start" or "Random on import" and then a "Shuffle" which re-randomizes at that point it is selected at. The Random on import should create a random sequence when new media is imported and store that play order and follow that in sequence until it makes it way through the list once and then re-randomize the list. The Random on start should re-randomize the list every time the car starts. This would allow a person to truly experience all the content in a random order without the variables of some songs hitting over and over.

ericbergan

I find it is easier to just generate a random playlist, and then use that. I actually prefer random album, rather than random song, so didn't really have any other choice. I use an Android App (Playlist Manager) and it also has options to date stamp tracks and say not to use them again for X days where you can choose what X is. It's not perfect, but it gets the job mostly done.

Richard Solomon

Well, we've been adding entropy generators to silicon implementations for a while now in order to seed crypto applications, so most systems now *could* have a fairly reasonable implementation of "true" randomness. Even without that, there are tons of techniques for seeding RNGs - e.g. measure the time between two button presses.

I think most people mean "random play" to be "randomize the order of a list of songs" as opposed to "select a song at random".

As usual the problem isn't the computer, but the person
"There are 10 kinds of people in the world - those that know binary, and those that don't!"

DeerHunter

Which just goes to show that having a little knowledge is a dangerous thing. "Danger" is my middle name.

Computers are usually never the problem (garbage in, garbage out). I tend to agree with you r.e. what people actually want (and even mentioned that option in my post). The problem, obviously, lies with whoever programmed that part of the MMI. They should have done more research about what people want (it's not like there's not a ton of user-generated feedback about this particular issue on thousands of different web pages) and done it right the first time. I can't see it being a memory or processor limitation, as the silicon guts of this version of MMI are fairly robust.

An oldie but a goodie