SH-101 Low Pass filter thoughts

[jsrobinson]

A lot of folks seem to commonly mention that the Roland SH-101 low pass filter is 18dB/octave. From what I've read in one thread here, (a single post referencing the Roland documentation,) the filter is actually 24dB/octave.

A lot of folks recommend "TAL Bassline" as the best VST emulation for the SH-101. TAL is actually 18dB.

- I'm wondering if the real SH-101 filter is indeed 24dB/octave? It seems from my reading the 303 is 18dB, but not the SH-101 as commonly mentioned on the `net.
- I'm also wondering if any soft synth by now gets reasonably close? Can Omnisphere with its soundsources? Does DCAM Strobe?

I'm curious to clear this up and see if anything is worth trying that can get the SH-101 sound within 90-95% accuracy or so.

Thanks for any discussion/thoughts!

[audioconsult]

the 101 has the standard roland filterchip that is used in most roland synths. so 4 pole 24 db...the 303 has a moog style solid state filter with a little variation to avoid patent problems..but that has´nt worked .. i think they got sewed anyway and used after the 303 only theire ota based filter chips..

the 303 filter is a 4 pole like in a moog but 1 pole is tuned an octave lower so its around the resonant point 18 db and on lower frequencies 24 db..

[jsrobinson]

Brilliant, thank you.
Does Omnisphere, DCAM Strobe, or anything else I might not have heard of get reasonably close to the SH-101's sound?

Omnisphere seems like it can do a lot from what I've seen around, but I'm not experienced enough to know if there are limitations it or Strobe might impose as opposed to the real SH-101.

[soundxplorer]

It used the IR3109 chip and I've always heard it was 4-pole. I've never heard anyone suggest it was 3-pole.

In reality the slope of a filter is one of the less-distinguishing characteristics. The tonal differences come more from the types of circuitry used. The IR3109 was a cascaded OTA design, as was the SSM 2040. The 303 filter was a diode ladder, and the Moog was a transistor ladder. These things all describe the resonance feedback loop, so the tonal differences become more evident as the resonance is increased.

I think the number of poles became a unique identifier back when the original Oberheim SEM was released. People made a big deal about the Minimoog being 4-pole and the SEM being 2-pole, so a lot of people assumed that was the entire reason why they sounded so different. But it isn't. A 2-pole Moog transistor ladder still sounds like a Moog filter, not an Oberheim. It just sounds a little brighter...but still like a Moog.

[clusterchord]

Originally Posted by audioconsult

the 101 has the standard roland filterchip that is used in most roland synths. so 4 pole 24 db...the 303 has a moog style solid state filter with a little variation to avoid patent problems..but that has´nt worked .. i think they got sewed anyway and used after the 303 only theire ota based filter chips..

the 303 filter is a 4 pole like in a moog but 1 pole is tuned an octave lower so its around the resonant point 18 db and on lower frequencies 24 db..

303 has a copy of EMS VCS3 filter, just like System101. it is a 4pole ladder that is correct, but built with diodes. and it is precisely due to their characteristics on all four poles, that net result is 18dB/Oct, not because 1pole is "tuned octave lower" - never heard of such interpretation, are you sure of this ??

EMS VCS3 (the early version) AND System100 had exact same thing: 4pole but 18dB. interestingly enough, the second version of VCS3 features a 5-pole filter.. and you guessed it.. that one has 24dB/Oct slope.


SH101 has standard cascaded integrator type chip IR3109 (Rolands copy of Doug Curtis 3320 which was again a copy of Dave Rossums SSM2040) , which is the basis for the filter, however surrounding circuit topology is designed so it has much drier, agressive/raw sound than for example Juno60 or Jupiter 8, which also feature 3109. this give it its characteristic acidic edge, or BoCish colour if you tune it near selfoscillation.

if you look at schematics youll see most Rolands from that era have variations in the filter circuit - for example JP6 and JP8, both w IR3109, but arent very similar at all. its not all in the chip.

[Rogue Ai]

Originally Posted by clusterchord

303 has a copy of EMS VCS3 filter, just like System101. it is a 4pole ladder that is correct, but built with diodes.

This could be why my diode ladder vcf module sounds a little like 303 filter though the resonance is a little different.

[audioconsult]

Originally Posted by clusterchord

303 has a copy of EMS VCS3 filter, just like System101. it is a 4pole ladder that is correct, but built with diodes. and it is precisely due to their characteristics on all four poles, that net result is 18dB/Oct, not because 1pole is "tuned octave lower" - never heard of such interpretation, are you sure of this ??

just look in the schematics.. a filter pole is a filter pole.. each does 6 db.. but they only sum up when they are tuned to the same frequency.

and its a diode ladder build with transistors.. but thats also covered in the original moog patents..

you can google booth.. the moog filter patent and the 303 schematics

[acreil]

Originally Posted by audioconsult

just look in the schematics..

One capacitor is a different value, but it doesn't mean that there's one pole tuned to a different frequency. The stages aren't isolated from each other, so you can't determine their frequencies independently. And anyway they all move to the same frequency when any negative feedback (resonance) is added. They're all 4 pole, 24 dB.

Tim Stinchcombe - Diode Ladder Filters (including the pretension to 18dB)

[Boddhisattva007]

I also wonder about the OP's other question... The one about AU/VST substitutes for an SH-101. I read on the Internet somewhere that D16 were making the SHIOITOR (don't know if that is spelled right) that is supposed to be (according to them) a near flawless recreation sound wise.... I wonder if that project is still alive.

[audioconsult]

Originally Posted by acreil

One capacitor is a different value, but it doesn't mean that there's one pole tuned to a different frequency.

no..it means just that.. that the poles might interact dont changes that the caps tune the poles..

""Aside from these differences, superficially they both look similar to the transistor ladder filter structure, but the move from transistors to diodes has implications in the way the circuit operates, and in one sense this leads to the loss of a certain amount of 'elegance' too: the resistor chain used to bias the transistors in the Moog ladder means that the voltages at each filter section are separated, which effectively means that the sections are buffered from each other; this 'isolation' is simply not present in the diode ladder, giving it a quite different transfer function (it is much harder to derive), and which in turn means the pole placement and their subsequent movement with increasing resonance is also quite different from the transistor version.




this just says that its different because there is no isolation in the diodelatter.. but all the graphs in the artical just show the non resonant state..

i tried different values in the caps when building my xoxbox to find out myself what the differnt cap is doing...and the fx of matched high quality 1% caps is a more narrow frequency cluster on the resonance that gets stronger when having all 4 poles the same cap value

changing the one pole as roland does gives a smother resonance that lets pass more of the original sound and is rhicher in overtones.. so like one of the poles is not resonating on the same frequency as the others..

for me that sounds like detuning one pole..

so however complex the poles might interact, changing the cap values does tune the poles somehow..at least when you get in the resonating state..

the math in the article is impressive.. but checking the circuit in practice with other caps just shows that the filter was modded by ear..it really sounds better the way roland did it and i guess that wasnt calculated ..

the goal was probably that the 303 has in all settings a sound that still dont kills speakers..and detuning the one pole just results in that..

at least thats the conclusion on my experiments.. and as long there are no simulated graphs for the resonant filter sweep on various cap values i tend to belive my own ears and not impressive mathematical formulas that dont necessarily contradict my impressions..

[JEBEQ]

As far as VST's go I haven't heard anything come close to the 101. I have even compared the TAL and that doesn't sound anything like it. Below is some audio of my 101, it just has character in the sound that I have not heard in software, You be the judge and tell me if you have heard a VST come close.
http://soundcloud.com/jebeq/roland-sh-101-acid-bass

[audioconsult]

have you ever heard a filter plug that can compare with real analoge filters?

would be nice.. but in reality every analog filter goes on resonant sweeps well above 50 k.. so with samplerates under 192k its probably difficult anyway to have them properly emulated..

[Acid Hazard]

As already stated, the IR3109 VCF Chip is 24db, not 18. As far as softsynth 101 replicants, we're all waiting for D16's LuSH-101 =o]

[acreil]

Originally Posted by audioconsult

changing the one pole as roland does gives a smother resonance that lets pass more of the original sound and is rhicher in overtones.. so like one of the poles is not resonating on the same frequency as the others..

for me that sounds like detuning one pole..

so however complex the poles might interact, changing the cap values does tune the poles somehow..at least when you get in the resonating state..

Wrong, offsetting the poles just reduces resonance (or rather increases the amount of negative feedback needed for resonance). The slope is slightly different only when the resonance is at minimum. They're all going to be slightly offset anyway, even when the capacitors are all the same value. But add enough feedback to bring the poles together, and it's exactly the same from there on out, regardless of the pole locations in the open loop case.

This is well known stuff, here: Root locus - Wikipedia, the free encyclopedia

Also, a single pole by itself doesn't resonate. It's always two complex conjugate poles.

[natefrogg]

imho, with software like tal, it is nice and all, but when a person jacks the resonance up and starts moving through the filter, it becomes very evident that it's not the real thing, there is a major lack of girth, force, strength, raw power, tearing through the space/time continuum haha

man those red 101s sure are lovely, i'd trade my 202 in a heartbeat

[audioconsult]

Originally Posted by acreil

Wrong, offsetting the poles just reduces resonance (or rather increases the amount of negative feedback needed for resonance). The slope is slightly different only when the resonance is at minimum. They're all going to be slightly offset anyway, even when the capacitors are all the same value. But add enough feedback to bring the poles together, and it's exactly the same from there on out, regardless of the pole locations in the open loop case.

This is well known stuff, here: Root locus - Wikipedia, the free encyclopedia

Also, a single pole by itself doesn't resonate. It's always two complex conjugate poles.

ok.. maybe so.. but why has roland done it this way and hasnt just reduced the max feedback amount ?

in any case the behaviour on resonance has leaded to the 18 db filter assumption because it sounds like a 18 db filter than.. just try the various digital 18db filter emulations.. they are there to do 303 alike sounds.. and they actually do sound closer to the 303 resonance than the 24 or 12 db settings.. acording to the other link you posted the 303 filter is a pure 23,xy db filter.. right? so why the 18 db resonance sound?

isnt that contradicting a little that its just less resonance.. ?

from my experiments i actually had the feel that the resonant sound itself was different and not just less.. but ok.. that was no a/b test of cause...

[seancostello]

Originally Posted by audioconsult

ok.. maybe so.. but why has roland done it this way and hasnt just reduced the max feedback amount ?

The more feedback needed for a given Q factor, the higher the internal signal levels, and the more possibilities for tasty distortion. Not saying that this was the goal, but it seems like this was the result.

The SH101 uses the same IR3109 chip as the Juno-60, but has back-to-back diodes going to ground in the feedback loop, which will limit the self oscillation volume, and produces a raunchy sound as the resonance is turned up. The Juno-60 filters always sounded tame to me, in comparison to the SH101.

[audioconsult]

Originally Posted by seancostello

The more feedback needed for a given Q factor, the higher the internal signal levels, and the more possibilities for tasty distortion. Not saying that this was the goal, but it seems like this was the result.

The SH101 uses the same IR3109 chip as the Juno-60, but has back-to-back diodes going to ground in the feedback loop, which will limit the self oscillation volume, and produces a raunchy sound as the resonance is turned up. The Juno-60 filters always sounded tame to me, in comparison to the SH101.

ok.. would explain the different sound and more overtones to just less resonance

[robkramble]

Originally Posted by natefrogg

imho, with software like tal, it is nice and all, but when a person jacks the resonance up and starts moving through the filter, it becomes very evident that it's not the real thing, there is a major lack of girth, force, strength, raw power, tearing through the space/time continuum haha

Listening to some of my favorite electronic tunes and A/Bing them with my own has made me realize that, aside from my arrangement and idea woes, there's one piece of the picture in particular that, even with the right source material, still sounds shitty ITB; the filter.

And it's exactly for the reason you stated. VST filters are entirely reductive, or painfully reductive with added "resonance sound" rather than actual resonance.

For this reason I'm investing in a Doepfer modular setup based around a couple SSM 2044 filter chips for some fun, lush and hands-on stereo filtering.

[arabesque]

I used TAL bassline for a good while before owning a 101. It's similar in a far far far off way, but sonically, not in the ballpark, not even in the same city as the ballpark.

The CLOSEST "similar sounding" but not "sounds just like a 101" vst's I've A/B'd myself, are Vaz Modular (after many hours of patching all sorts of things in to try to create the subtle characteristics), and uhe ACE (for lower res glidey stuff).

Beyond that, I agree with one the the posts above, the 101 has this quality, like many other analogs, but it has it's own character, that I have just not heard in software yet. Especially the chirpy-nasal'y type sounds, that some software can get sorta close to, but lack the depth and body the 101 provides. However with eq/fx/compression you can lose a lot of the 101's character and at the end of the day, if you're just trying to make sounds you're happy with, there's plenty of vst's out there to get you close enough to make a good song with good sounds.

On a side note, it was amazing to me - after just a few months of finally owning a 101 - how many "classic sounds" from the 80's/90's were a 101 that I had originally thought were 303. There is a very distinct difference in the OBVIOUS 303 sound of course, but there is some overlap, especially with distortion. It's just something I feel I never would have "appreciated" had I not finally owned a 101.

I've owned lots of synths, digital and analog, and the 101 is hands down one of my most favorites. IF for no other reason, the immediacy and the sliders, it's just seconds before I have something that sounds "great." But the synth is SO flexible (not even to mention how much more with mods!) AND also STABLE, and typically in much better working order/condition than a lot of other analogs purchased from CL and ebay.

But the sound --- it's just "there."

[blinky909]

to toss some gas on the fire...

i have owned two 202's and a 101. i sold my 101 and still have a 202. to my ears the 202 has a more raw, less precise sound than the 101. the 101 had a brittle, FM metallic edge that made it less chewy. the classic Roland hollow squarewave sound seems very deeper, and broader in the 202.

looking at the filter schematics, the 202 uses a slightly different feedback setup for the resonance which might explain the difference i hear.

[blinky909]

for all your Roland filter needs...

roland_filter_versions


the synths that use the 3109 all seem to have different resonance feedback stages.

[seancostello]

Originally Posted by blinky909

for all your Roland filter needs...

roland_filter_versions


the synths that use the 3109 all seem to have different resonance feedback stages.

Thanks for the links! There are a LOT of diode ladder filters in there, as well as a bunch of cascaded OTA filters with back-to-back diodes to ground in the feedback loop (SH-2, SH-7, SH-09, SH-101, MC-202, System-100M), and a few true Moog filter clones. All of these seem to be in the monosynths; the polys either use 4 cascaded OTA stages with no extra clipping (although there might be desirable clipping within the OTA stages themselves, as with the SSM2040), or cascaded SVFs in the Jupiter 6 (with the 2nd stage using diode limiting in the resonance feedback loop, a la the Oberheim SEM). It makes me wish that I had been on the lookout for more Roland monosynths back in the 1990's, when these things were reasonably affordable. I love nonlinear filters in synths.

[blinky909]

Originally Posted by seancostello

It makes me wish that I had been on the lookout for more Roland monosynths back in the 1990's, when these things were reasonably affordable. I love nonlinear filters in synths.

i'm a Roland monosynth slut. i'd love a System 100m and an SH-7.

[Llitsor]

This all makes for an interesting read.. I built a few Shruthi-1 synths that use the IR3109 IC. They sound very nice.. Im not sure diodes are used in this version of the filter though, but perhaps it could make for an interesting mod? I'll look into it!

[seancostello]

Originally Posted by Llitsor

This all makes for an interesting read.. I built a few Shruthi-1 synths that use the IR3109 IC. They sound very nice.. Im not sure diodes are used in this version of the filter though, but perhaps it could make for an interesting mod? I'll look into it!

It would make for an interesting mod. For that matter, any cascaded OTA filter would probably take on an SH101 flavor with back-to-back diodes going to ground in the feedback loop.

Mind you, not all OTAs are made equal. I am not a big fan of the CEM3320, as it sounds kinda brittle to me. The LM13600/13700, and their equivalents, apparently use something similar to the CA3080, which has a nice tanh() overdrive response. It would be interesting to compare the IR3109 filter w/diode clipping to a filter based around 4 13600 OTAS (+ external buffers) that used similar diode clipping, and see what kind of sonic differences result.

Of course, everyone has different tastes. Some people love the precise filters in the Micromoog and Multimoog, while I prefer the coarser, more distorted Rogue filters. The Monotron has one of my favorite filters EVER, and I suspect that the noise from the low voltage power supply adds to the sound for me. Generally speaking, I'd rather have a thuggish overdriven VCF than a filter with sharp resonance.

[blinky909]

I love that rounded top nasal acid sound that Roland did so well mid 80s

[Llitsor]

Originally Posted by seancostello

It would make for an interesting mod. For that matter, any cascaded OTA filter would probably take on an SH101 flavor with back-to-back diodes going to ground in the feedback loop.

Mind you, not all OTAs are made equal. I am not a big fan of the CEM3320, as it sounds kinda brittle to me. The LM13600/13700, and their equivalents, apparently use something similar to the CA3080, which has a nice tanh() overdrive response. It would be interesting to compare the IR3109 filter w/diode clipping to a filter based around 4 13600 OTAS (+ external buffers) that used similar diode clipping, and see what kind of sonic differences result.

Of course, everyone has different tastes. Some people love the precise filters in the Micromoog and Multimoog, while I prefer the coarser, more distorted Rogue filters. The Monotron has one of my favorite filters EVER, and I suspect that the noise from the low voltage power supply adds to the sound for me. Generally speaking, I'd rather have a thuggish overdriven VCF than a filter with sharp resonance.

You'd enjoy all the filter talk on the Mutable Instruments forum then! I think there are 7 different filters available for the Shruthi now. And the IR3109 version is definitely a favourite of mine.
Being that these things are generally DIY, and schematics etc are freely available, they lend themselves quite well to mods & tweaks.

Im not fully up on my electronics however (im always learning though!) but you'll see LM13700's used on many of the filter board options. Largely because there isn't a lot of choices for OTA's these days I gather(?) The SMR 4 filter uses 3 of them though, and sounds good, plus the design already allows for diodes to be added, although i've not tried it yet.

Ooops, i've gone off on a bit of a tangent!

[rockmanrock]

The MFOS 24dB filter uses a couple of LM13700 OTAs too. See how it compares to an old Roland:

[blinky909]

Originally Posted by rockmanrock

The MFOS 24dB filter uses a couple of LM13700 OTAs too. See how it compares to an old Roland:

Damn... Thanks for posting that. The Roland is a little more musical, but not by much.