Jump to content

AN1x filters are not correctly modeled


Don Solaris

Recommended Posts

  • Members

I don't like finding failures, but after JP-8000 / 8080 filter distortion issue i think it is better to publish information, seek for possible solution or beg big companies to do something about it, than just to sit and do nothing. ;)

 

an1x.jpg

 

This have been bugging me for quite a long time. When playing Yamaha AN1x i've noticed that when switching from 12 dB to 18dB and specially to 24 dB, the change is minimal. Yesterday i decided to clear this out once and for all. Testing was done using AN1x white noise oscillator. Filter was set to near half position (60/128), resonance 0, filter keytracking 0 and filter env depth 0. Other oscillators, ring mod and feedback were set to 0. Effects were off.

 

12 dB filter:

 

An1x12dB.gif

 

Difference between one octave (450-900 Hz) turned out to be 12.75 dB. This is more closer to 13 dB than 12 dB.

 

 

18 dB filter:

 

An1x18.gif

 

Difference between one octave (450-900 Hz) turned out to be 15.75 dB. This is not good! You want to emulate the TB-303? Forget it. Looks like you will need to use the 24 dB one (see next filter type).

 

 

24 dB filter:

 

An1x24dB.gif

 

Difference between one octave (450-900 Hz) turned out to be 19.15 dB. This is far too much difference to be tolerated. If you want short Jupiter 6 type 24 dB basslines or arpeggios - forget them. The filter is too shallow. It is more like a 18 dB filter, so ironically if you want a 18dB filter, you need to select a 24 dB one.

 

For comparison on JP-8080 (and other synths) when you switch from 12dB to 24dB you can immediately hear the difference. On AN1-x it sound like a small change, and when switching from 12 to 18 almost no change at all.

 

In short

The Yamaha AN1-x does not contain true 12, 18 and 24 dB filters. Instead it offers some "half-filters" with values of 13, 15 and 19 dB. This can be interesting for building classic 12dB (Jupiter 8) type of pads and strings, but when you need a strict 24 dB filter for percussive instruments, dance basslines or JP-6 arpeggio lines, the provided 24 dB one simply doesn't cut enough.

 

If anyone is interested in doing more tests, please use the above settings and same frequency range when making spectral analysis. (This however doesn't change the fact that AN1-x offers a huge sound palette thanks to its deep and complex routing - it's a really nice synth, perfect for sound design and special effects.)

Link to comment
Share on other sites

  • Replies 76
  • Created
  • Last Reply
  • Members

It might be worth running the same tests against a couple of other synths for comparison

Yup, already done it.

 

Roland XP-30 12 dB filter:

12db.gif

Result is exactly 12 dB.

 

 

Roland XP-30 24 dB filter (using structure):

24db.gif

Result is exactly 24 dB.

 

 

Of course i can make few more tests, but it won't cure the problem. :( I can't figure out why they put "24 dB" mode when it sounds like 18 dB - in fact, results show it is a 18 dB filter (well, 19.15 dB to be exact).

Link to comment
Share on other sites

  • Members

Yup, already done it.

Roland XP-30 12 dB filter:

This graph misses the spike you get when you clonk the resonance up to the horribly squealing noise they still haven't fixed.

 

Of course i can make few more tests, but it won't cure the problem.
:(
I can't figure out why they put "24 dB" mode when it sounds like 18 dB - in fact, results show it is a 18 dB filter (well, 19.15 dB to be exact).

I guess whoever made the TB-303 kind of got a "whoops" feeling about that, too.

 

"Whoops". :D

Link to comment
Share on other sites

  • Members

Hey Don,

 

Very cool analysis. Do you find that the slopes on the AN1X filters are dependant on the amp feedback parameters? (I think you would agree that the perceived resonance and cutoff drop ... and more so on the "24dB/oct" filter) when using amp feedback. I am just wondering if the slope is affected also?

 

I can hear the low frequencies rise in volume as I adjust the amp feedback, and I wonder if there is a magic number where the 24 db/oct filter is really what it says it is. Of course if that is the case, there is not much resonance available on the filter.

 

Best,

 

Jerry

Link to comment
Share on other sites

  • Members

Interesting, I always thought the 24db sounded weak and I always had a lot of difficulty emulating the AN1x on another synth.

 

I'd be real curious to see where the EX5, CS6x and motif series are at... Have to say though that after dumping it I've never wanted it back.

Link to comment
Share on other sites

  • Members

 

Interesting, I always thought the 24db sounded weak and I always had a lot of difficulty emulating the AN1x on another synth.

 

Well, i was embarrassed even talking about it, thinking maybe i'm crazy, or maybe it is just in my mind sounding different, i mean, hey! this is Yamaha!, they are Big! If they write 24 dB, then i must be 24 dB!!

 

So it wasn't until i bought JP-8080 and switched from 12 dB into 24 dB position - BINGO! This is it! This is a "switch" from one sound to another, and you can clearly hear how 24 dB cuts away high partials. On AN1-x you need to concentrate deeply to spot a tiny difference.

Link to comment
Share on other sites

  • Members

Well, i was embarrassed even talking about it, thinking maybe i'm crazy, or maybe it is just in my mind sounding different, i mean, hey! this is Yamaha!, they are Big! If they write 24 dB, then i must be 24 dB!!

 

You are not crazy. You have great ears. Thanks for doing this. :)

 

Jerry

Link to comment
Share on other sites

  • Members

 

If you don't mind, can you tell us what application you are using to generate these graphs?

 

Data analysis: Audacity (freeware!) - analyized @ 16384 FFT exported as text file.

Data plot: Gnuplot (freeware!) - a command line program, but really powerfull.

 

Here is a notepad i wrote for the gnuplot in case you will do the test

 

# gnuplot for Windows

# AN1-x filter analysis

# set output "filter3.png" # uncomment this for picture export

# set terminal gif small size 400,300 #uncomment this for picture export

set terminal windows color "Small Fonts" 7 #comment this line for pic export

set xrange [450:900]

set xtics 50

set xlabel "Frequency (Hz)"

set yrange [-48:0] #adjust this to fit within 24dB for 24 dB filter

set ylabel "Intensity (dB)"

plot "filter3.txt" title "24 dB filter" with filledcurves x1 3

Link to comment
Share on other sites

  • Members

That one went over my pea brain. How would putting it after the effects chain make it useful? Kinda like a warm pre-amp on a mixing board, do you mean?

It would give you one of the sickest effects you can imagine. Endless delays, reverbs, massive wall of distorted sounds, pulsating and screaming with resonance, apply the filter and gently cut away high overtones while delay is still in the endless loop. You would have a true loop system, when you remove the dry signal from the delay, feedback is gone, so you have that endless delay texture that you can realtime modify by altering the cutoff, applying high pass and injecting more fresh sound in. Or apply two LFO's to do it instead of you, leave it in near self oscillation position, go for the dinner, come back to see what came up. You can record the whole Pighood style album that way, without even touching the synthesizer. (sorry Piggy, i couldn't resist!) :D

Link to comment
Share on other sites

  • Members

Maybe you should go measure various analog filters and see how close they come to actual spec. Filter designs in the analog domain are far from ideal in response.

 

After all, there are many 4-pole analog LPFs out there yet no two sound alike....maybe we should start a government inquiry :thu: Look at this analysis suggesting the "3 pole" filter in a TB-303 is just a botched 4-pole design.

 

It's all meaningless to me. I don't create patches on my PLG150-AN using graph paper I do it by ear. I just cycle through the filter types until I find one I like.

Link to comment
Share on other sites

  • Members

It would give you one of the sickest effects you can imagine. Endless delays, reverbs, massive wall of distorted sounds, pulsating and screaming with resonance, apply the filter and gently cut away high overtones while delay is still in the endless loop. You would have a true loop system, when you remove the dry signal from the delay, feedback is gone, so you have that endless delay texture that you can realtime modify by altering the cutoff, applying high pass and injecting more fresh sound in. Or apply two LFO's to do it instead of you, leave it in near self oscillation position, go for the dinner, come back to see what came up. You can record the whole Pighood style album that way, without even touching the synthesizer. (sorry Piggy, i couldn't resist!)
:D

 

Sounds like you are referring to delay feedback. Very cool. The AN1X allows you to get infinitely regenerated delays. There also an LPF in the delay unit for gently removing the highs. The dsp doesn't sound all that great though. Ok for Pink Floyd/ Radiohead emulations on stage.

 

It sounds to me like you have no interest in exploring the idea that Yamaha has given as a 24 db filter topology with 127 variations (depending on the amp feedback paramater, which is actually a coupled filter/amp feedback loop). I wonder what the graph would look like with the overdrive at +63.

 

Jerry

Link to comment
Share on other sites

  • Members

Don,

 

You just gave me a great idea. I am using the echo delay on the an1x and a mellow synth sound. I have the left channel on 99 feedback and delay time of 330 ms (max) right channel also has 99 feedback, but I am manually controlling delay time between (assign a knob to Data Entry in the control devices menu) sweeping between 200 ms and 0.01ms. It's got the whole grain shifter thing going. :love:

 

Jerry

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...