Announcing Poly-Ana Beta 0.11 Special offer save $60

[fred-hal]

The sounds are absolutely amazing, but the gui is a pain to work with..

..but another synth added to my 'what to spend my money on'-list..

[aciddose]

do you want bugs posted here?

[bduffy]

@Admiral: Well, you know, that's just my first reaction. I've defended your unusual GUI in the past and I still feel the same. Now that pointed out the tooltip up top, that's already better! And with a complete manual, I'd bet it wouldn't matter that much once you're used to it; after a year, you're probably pretty set on this architecture anyway!

Yeah, this thing sounds awesome; your year has paid off, large, by my estimation. I totally agree this is in the G-Media/Arturia league, and even better than that, it's an original. Can't wait for the next versions, keep doin' what yer doin'. Crazy like a fox.

[AdmiralQuality]

do you want bugs posted here?

Sure (so everyone can know what's been reported already), but try to be brief. If you need to get into detail PM or email me.

[aciddose]

ok, this one is probably not a big deal for you, but the filter doesnt accurately track the keyboard even when set to exactly 100%. i'm not sure if you mind having it like that or not, but when you want to do organ sounds with high res, or use the filter to add an additional partial having perfect tracking is important. try setting res to 100% and tune it to a position where it doesnt "burble". play different keys and you'll get different patterns of "burble" from the saturation. that is caused by inaccurate tuning. it isnt "ok since analogs are like that too" because analogs are not like that. the problem with making certain types of patches is what concerns me, otherwise i wouldnt give a damn about it. also its a bit annoying for me that you have a fixed amount of saturation happening.. that means you cant make clean tones with the filter, and your saturation is a loooot more than the typical analog has. you're in line with the wasp filter or something simmilar.

second bug i found is, your voice stealing will drop each note when it cycles through to the last voice if you have it set to only one voice. i assume this is because you're doing if (current voice > voices) current voice = voices; noteon(current voice, etc), where you end up with it always being equal to voice 1 (index 0 in the code i assume).

i'll keep playing with it and find some more if they're in there. sounds really good so far, that phase thing, whatever it is seems to work well combined with the naive oversampling. better than i would have thought oversampling could be actually.

[diverdee]

Haven't tried it yet, althoug I undoubtedly will.
I'm glad that people are liking the sound of it.
Tooltips (or something) would likely be useful, it's beautiful for sure, but strains the old eyes.
Looking forward to hearing demos people do.
It's even moe cool that it has no internal effects & presets are blowing people away.
Well done AQ

[aciddose]

having the dual signal path makes for some pretty good depth in the sounds, comes at a cost of cpu of course though :)
it kills my machine (p3 550) to the point where i couldnt stand using it in a track, so hopefully that will improve over time.

found another "bug".. i notice you do not filter the gate signal when used on the amp stages. you should probably use a quick attack, just enough not to make a click (2.5ms?) and a moderately fast release (30-50ms?) instead of just the plain sharp edges. every analog i've used does it like that, although that is not really a bug either, its an issue of taste.

by the way, do you want me to mention these subjective things, or just stick to technical problems?

[AdmiralQuality]

ok, this one is probably not a big deal for you, but the filter doesnt accurately track the keyboard even when set to exactly 100%.

Known issue. Read the "docs". Turning the quality up to 8X will dramatically improve this. But yes, I do hope to be able to compensate for it someday.

second bug i found is, your voice stealing will drop each note when it cycles through to the last voice if you have it set to only one voice. i assume this is

Switch to unison mode if you want to play Mono. Polyphony knob just affects the number of voices.

i'll keep playing with it and find some more if they're in there. sounds really good so far, that phase thing, whatever it is seems to work well combined with the naive oversampling. better than i would have thought oversampling could be actually.

It's amazing what a little brute force will do. And as we all know, THAT is why the real item sounds better than most softsynths... because a real synth isn't limited to some low Nyquist like software is. Well, we still have Nyquist in Poly-Ana, but it's pushed WAY WAY up there. Go into the options and pick 8X oversampling and listen to the difference. You'll probably even find your filters track (almost! would have to be infinite for it to be perfect.)

If you have any advice about scaling the cutoff to compensate for that, I'd love to hear it btw (in private or another thread, don't want to pollute this thread too much) but we already went through that last year to no avail. It's non-linear is the problem. But I'm sure it's not TOO hard a curve to calculate. Room for improvement...

Thanks AD.

[AdmiralQuality]

found another "bug".. i notice you do not filter the gate signal when used on the amp stages. you should probably use a quick attack, just enough not to make a click (2.5ms?) and a moderately fast release (30-50ms?) instead of just the plain sharp edges. every analog i've used does it like that, although that is not really a bug either, its an issue of taste.

by the way, do you want me to mention these subjective things, or just stick to technical problems?

That was intentional -- a synth salesman at a local music store played with a Minimoog Voyager with me and was saying how good it was that the gates and envelopes clicked and how that's a sign of a GOOD analog synth design. So yes, it's a matter of taste. But I do also know that other's have a different taste and yes, the clicks could be rounded some. Maybe this is an option as sometimes I LIKE it (and if you run through an amp sim or some reverb, like you would in real life, it's hard to hear that stuff and the clicks can actually help in the perception that it's "cutting through the mix")

Will contine to think about it!

[mistertoast]

>>It's non-linear is the problem. But I'm sure it's not TOO hard a curve to calculate. Room for improvement...

Could it just be done empirically?

[aciddose]

it IS done empirically, and these are the results :)

the functions required to tune a dsp filter are quite complicated if you need the best tuning possible. they're difficult to find in texts if you're using a standard filter type, AQ isnt quite using anything you might call standard so he would have to work out the functions himself.

[AdmiralQuality]

it IS done empirically, and these are the results

the functions required to tune a dsp filter are quite complicated if you need the best tuning possible. they're difficult to find in texts if you're using a standard filter type, AQ isnt quite using anything you might call standard so he would have to work out the functions himself.

2:1

It was that simple. But it breaks down the lower the oversampling is set. Here I go giving away my mojo again but: the curve is flat at the low end and gets curvier at the upper end. If you push the upper end far enough past the host's Nyquist then you never know the difference, it's effectively flat through the audible range, and it acts just like a real world filter because you've made the resonance feedback loop essentially instantaneous rather than one sample long. As far as it knows it IS a real world filter! (Hooray for the brute force solution!) So it actually DOES work JUST LIKE the real world item. It gets closer to real performance the higher you set the oversample rate (and I chose 8X as the maximum oversample rate because that's where it gets to the point that you CAN play the filters in tune! Try your test again in 8X oversampling aciddose. You'll still hear the filters go out of tune a TINY bit but it will be playable across at least 4 or 5 octaves!)

I very nearly released this synth with 8X oversampling hard-wired. But at the last minute (i.e. the last 3 weeks) I realized that to some people, the aliasing artifacts of the lower sample rates will actually be considered a feature! Also, I can only get about 4 voices on my meagre machine at 8X, where at 4X (the default) I can get 6 to 8... so that's what convinced me. If I released a synth that couldn't play 6 voices, you'd have all beat me up. Compromise...

Oh man... my mojo! PATENT PENDING!!!

[Kingston]

I'm sure you could work out the *exact* math behind the filter vs. samplingrate (that you're sort of cheating with oversampling now).

...or you could do it the hard way, like me, and fork it out and plot the transform. I had to do this quite recently to a sort of a new filter design I did as I lacked the skill for the actual derivation of the filter vs. nyquist curve. Mine happened to be so damn highly non-linear that it hurt my head.

It worked out rather well in the end though. Although I have to say I only managed to mask the actual problem, not solve it (but you didn't hear me say this).

Oh and looking forward to playing polyanna (I'm in the middle of a bitchy OSX port of my plug so no time yet).

[AdmiralQuality]

Kingston. I hear ya, and that's why I say this is "room for improvement". I DO fully expect that someday Poly-Ana won't have filter cutoff shifts between the different Quality settings. I just didn't want to hold release off for another month while I fiddled with this... R&D was officially done 6 months ago. It's been all implementation work since then!

By the way, filter cutoff is only half of what we're "cheating" with the oversampling method (actually it's everyone ELSE who's cheating, we do it like REALITY!) The other half is oscillator aliasing/bandlimiting. There's no BLEPs in Poly-Ana (...yet. This may come as well if we can free up enough CPU demand. Though it will STILL be oversampled.)

But as things calm down, over the next coming ... oh, about 6 more months... I'll be able to focus on some of these little problems and most probably, beat them too!

For now... it sounds AMAZING at 8X, GREAT at 4X, not bad at 2X, and embarassing at 1X. But if you're making a filter-resonance-tracking-the-keyboard kind of sound, then 8X is about your only option.... for now.

I think I need to start an about:Poly-Ana thread in the DSP and Dev board so we can stop scaring away people with all our esoteric geek-talk. In fact... I will... Geeks go here please: http://www.kvraudio.com/forum/viewtopic.php?t=153774

[Benedict]

Congtratulations Admiral

I might just have a poke at a demo in a week or so when my partner has a wedding I'm not going to so I'll be all washed up on a Sat night and by then the Pre Roll Recorder Buddies should be all tucked in.