Lexicon reverbs: a brief bestiary

[Warp69]

Originally Posted by seancostello

.............224XL algorithms in general seem to have a different sound than the 480L and later units (with the exception of the 480L Classic Cart).

Was David responsible for the Classic card for the 480L???

[seancostello]

Originally Posted by Warp69

Casey described a constant density plate algorithm in another thread - not exactly like an EMT250, but similar.

Yep. I consider the "constant density" to refer to the fact that the echo density would not increase with time. This is found in algorithms with parallel combs with no cross coupling and no embedded or nested allpasses. The original Schroeder algorithms were constant density, as well as the EMT250-esque algorithm Casey describes. The "plate" implies a fairly high echo density, which can be realized in a number of ways - lots of output taps, series allpasses, lots of parallel combs, etc.

Not many algorithms from the 224 has survived - only the Concert Hall?

I am unsure how much later plates and chambers have in common with the 224XL. As far as the original 224 (no X or XL), apparently these algorithms sound fairly different from the 224XL. One of the Concert Hall algorithms in the 224XL is apparently the same as the original 224 Concert Hall. Some of the other 224XL Concert Halls seem to add chorusing to additional locations within the algorithm (Dark Hall), which implies that the 224XL had a few more cycles than the original 224.

Sean

[dale116dot7]

Originally Posted by seancostello

Some of the other 224XL Concert Halls seem to add chorusing to additional locations within the algorithm (Dark Hall), which implies that the 224XL had a few more cycles than the original 224.

The 224XL service manual indicates that the program is 100 steps. However, the microcode actually could do 128 steps - the program counter is cleared by writing an address with bit 3 set to the I/O space. The 224 (non-XL) version shows the same sort of thing - the WCS resets the PC. Though many of the Lexicons have a 128-step program counter, not all of them actually use all 128 steps.

To multiply with sufficient precision to do a linear interpolate, I would guess that four lines of code are likely needed for each chorus certainly on something like the PCM70 or 480L. But the 224 uses a 6-bit multiplier which would be on the ragged edge of being able to pull off a chorus in two lines.

[seancostello]

Originally Posted by dale116dot7

To multiply with sufficient precision to do a linear interpolate, I would guess that four lines of code are likely needed for each chorus certainly on something like the PCM70 or 480L. But the 224 uses a 6-bit multiplier which would be on the ragged edge of being able to pull off a chorus in two lines.

Ragged edge is probably good enough for this application. The SST-282 that Casey brought up used NO interpolation for its modulated feedback taps. It just went ahead and moved them.

Casey's interpolation example used 32 discrete steps between samples. This was probably not meant to be an ideal example, of course. It seems like the 224 might have 64 possible steps for its interpolation if using the two line chorusing, which may produce acceptable results.

Sean

[dale116dot7]

Originally Posted by seancostello

Ragged edge is probably good enough for this application. The SST-282 that Casey brought up used NO interpolation for its modulated feedback taps. It just went ahead and moved them.

Chris didn't like that but it would be difficult with the SST architecture to do anything much about it without adding many more PROMs to the design. He even commented on it in the brochure. You'd have to make the gain prom pretty big, and run another state machine for tap fadeins before moving the tap.

Casey's interpolation example used 32 discrete steps between samples. This was probably not meant to be an ideal example, of course. It seems like the 224 might have 64 possible steps for its interpolation if using the two line chorusing, which may produce acceptable results.

If you updated the chorus gain coefficients and addresses on every sample, that would be the same granularity as a 4 Hz chorus of a 3ms depth, even if you had a more precise multiplier you'd have that granularily anyways. That's a faster chorus than I would expect in a reverb, but a two-line chorus would be at least not in the grunge area of granularity.

[seancostello]

Originally Posted by dale116dot7

If you updated the chorus gain coefficients and addresses on every sample, that would be the same granularity as a 4 Hz chorus of a 3ms depth, even if you had a more precise multiplier you'd have that granularily anyways.

That's an interesting way of looking at it.

That's a faster chorus than I would expect in a reverb, but a two-line chorus would be at least not in the grunge area of granularity.

You can get some nice effects with faster chorus settings. Very Budd/Eno. I think they used the modulation from a Lexicon Prime Time in their work, with a medium speed LFO, around 3-4 Hz, and ran this in feedback with an EMT-250.

Sean

[seancostello]

One more box to ask about: LARES. Do these run a 480L-type Random Hall algorithm, or something totally different?

[Casey]

Originally Posted by seancostello

One more box to ask about: LARES. Do these run a 480L-type Random Hall algorithm, or something totally different?

The reverb itself is the surround or HD algorithm found on the 480 surround cart.

The strong chorusing on the early portion of this algorithm was designed in part to increase the gain before feedback needed for the live LARES application.



-Casey

[seancostello]

Originally Posted by Casey

The reverb itself is the surround or HD algorithm found on the 480 surround cart.

The strong chorusing on the early portion of this algorithm was designed in part to increase the gain before feedback needed for the live LARES application.

Thanks for the info. This fits in with what Peter Svensson found in his research, that delay line modulation was the most effective technique of reducing feedback in reverberation enhancement systems, although phase modulation (via Hilbert technique) was more effective for lower frequencies.

If I read correctly, the LARES systems currently run on Lexichips, which would imply that the HD algorithm was ported over to that hardware at some point.

Sean

[dale116dot7]

I was flipping through my PCM91 preset list, and noticed that most 'rooms' use either the concert hall or random hall presets - a lot of random hall. Not many 'rooms' on there actually use the room/chamber algorithm. A big surprise for me was to find that 'Saxy Hangar' with its long tail uses the Ambience algorithm. Is this a common thing to do - abuse the algorithms?

[Michael Carnes]

Originally Posted by dale116dot7

Is this a common thing to do - abuse the algorithms?

You bet your bippy! The name of the preset is (and always has been) an indication of the sound, not an indication of the underlying algorithm. Lex algorithms are generally quite flexible and sometimes we get a better effect by 'abusing' the algorithm. Once you get to the 960L and PCM96, you'll even find rooms made from delay algorithms (on the PCM96, there are even rooms made from chorus and pitch shift).

I've been amused that another thread has latched onto a hall preset called "EM7", believing it to be our attempt at modeling the Bricasti. In fact, that preset is based on the new "Room" algorithm and was simply a tip of the hat. Just as a hall can be used to make a room, a room can be used to make a hall!

Confused enough?

[Michael Carnes]

Originally Posted by seancostello

If I read correctly, the LARES systems currently run on Lexichips, which would imply that the HD algorithm was ported over to that hardware at some point.

Almost, but not quite. LARES was originally a separate algorithm for the 480L (not provided in the standard unit). Dave G used a few aspects of LARES for his HD algorithm, but HD isn't a full LARES implementation.

Several years ago, LARES was ported to the 960L and has been the basis of LARES systems for about the last 5-6 years. Another version was ported to the DNA chip (another Harman chip) and is used in the Wenger practice room. For those that don't know the Wenger room, it's pretty cool. It's used at conservatories and provides a realistic rehearsal space for musicians. Simply dial in where you'd like to be and start playing.

[dale116dot7]

Originally Posted by Nobody Special

You bet your bippy! The name of the preset is (and always has been) an indication of the sound, not an indication of the underlying algorithm. Lex algorithms are generally quite flexible and sometimes we get a better effect by 'abusing' the algorithm. Once you get to the 960L and PCM96, you'll even find rooms made from delay algorithms (on the PCM96, there are even rooms made from chorus and pitch shift).

Makes sense, if the parameters are set right. On the PCM70, 'Tunnel' which sounds like it shold be a reverb algorithm is really just a six-voice delay. I guess I never paid much attention to the actual algorithm used until we've started these discussions. I just dial up 'vocal magic' or 'just plate' or whatever, and either just use it, or diddle with some of the settings a bit. Usually I pick a preset and diddle with either size or RT or the RT EQ, that's normally enough.

[zmix]

Originally Posted by seancostello

That's an interesting way of looking at it.

You can get some nice effects with faster chorus settings. Very Budd/Eno. I think they used the modulation from a Lexicon Prime Time in their work, with a medium speed LFO, around 3-4 Hz, and ran this in feedback with an EMT-250.

Sean

Sean,

I've enjoyed reading your posts on the subject of artificial reverb in these many threads, thanks for contributing..! What's your background in DSP? ( EDIT: I read your blog.. we must speak! )

Every track on those Eno / Budd records used a different signal chain.. nothing to my ears as warbley as 3-4hz.. plenty of modulations and recirculated signal routings, to be sure. Possible the most interesting soundscapes I've encountered, though!!! I'd love to see a gearlist for the "Ambient 2" and "The Pearl".. right? I have some doubts about the use of an EMT 250 simply due to it's scarcity and price (given the budget of those records). In a related bit of trivia, the 250 has some very very slow modulation in it's algo. A remarkably effective and simple algo it is (4 delay / diffusor stages)...

CZ

[seancostello]

Originally Posted by zmix

I read your blog.. we must speak!

So you're the one person I got hits from!

I'd love to see a gearlist for the "Ambient 2" and "The Pearl".. right? I have some doubts about the use of an EMT 250 simply due to it's scarcity and price (given the budget of those records).

I was basing this off of a Daniel Lanois interview I read:

Music Glob » Blog Archive » Daniel Lanois Interview


The later ambient work that Eno and Lanois worked on apparently made heavy use of the Lexicon 224. The Concert Hall algorithm was used for its modulated ambience, and the 6-voice Chorus added modulation to long feedback paths. The various Lexicon Concert Hall algorithms I have heard (PCM-70, PCM91) have internal modulation, which creates an effect similar to the Budd/Eno work.

In a related bit of trivia, the 250 has some very very slow modulation in it's algo. A remarkably effective and simple algo it is (4 delay / diffusor stages)...

Casey has written about an EMT-esque algorithm here on Gearslutz:

Reverb Subculture

Probably not the exact EMT algorithm, but still interesting.

Sean Costello

[zmix]

Originally Posted by seancostello

I was basing this off of a Daniel Lanois interview I read:

Reverb Subculture

The later ambient work that Eno and Lanois worked on apparently made heavy use of the Lexicon 224. The Concert Hall algorithm was used for its modulated ambience, and the 6-voice Chorus added modulation to long feedback paths. The various Lexicon Concert Hall algorithms I have heard (PCM-70, PCM91) have internal modulation, which creates an effect similar to the Budd/Eno work.

Casey has written about an EMT-esque algorithm here on Gearslutz:

Reverb Subculture

Probably not the exact EMT algorithm, but still interesting.

Sean Costello

I'm quite familiar with the EMT 250 and have spoken with Barry Blesser about some small aspects of it. I find early digital reverb research fascinating stuff.

As for the Lanois Interview.. both of those links you provided point to the same thread, which I had already seen, but I am interested in reading the interview you cited.

[seancostello]

Originally Posted by zmix

As for the Lanois Interview.. both of those links you provided point to the same thread, which I had already seen, but I am interested in reading the interview you cited.

Sorry - yesterday was a long day. The proper link is

Music Glob » Blog Archive » Daniel Lanois Interview

I fixed my post as well.

Sean

[Honch]

I've been readin this thread for quite a while, with great amusement and find most of it quite intriguing. I find it peculiar though, that no one has mentioned the "ECM reverb" yet.

To those of you who works at Lex, can you hear what preset they use on ECM recordings? I e by engineer Jan-Erik Kongshaug and/or Manfred Eicher?

For those who doesn't know:

ECM is a SMALL record label (German/Norwegian) specialising in chamber jazz music, and uses a lot of silence and - especially - space in their recordings. I e lots of ambience added to their production. Known artists are Keith Jarrett, Pat Metheny and so on. Critics have been labeling the label such as "the most beautiful sound next to silence" and "the only label not in need to ever remaster their back catalog". But that's another thread, topic and discussion.

But to me, that label has always been the one to promote and pitch "Lexicon reverb" sound the most, to me anyway. A whole demo label for Lexicon, so to speak. Although, in some instances, too excessively. Violinist Paul Giger, and some Pat Metheny Group productions comes to mind, for overdoing it. Pat left the label when he thought Manfred Eicher went over the top adding excessive reverb to his material.

Can you hear that they're using (or was) Lexicon units at all? And what presets or algorithms?

/Honch

[zmix]

Originally Posted by honch

i've been readin this thread for quite a while, with great amusement and find most of it quite intriguing. I find it peculiar though, that no one has mentioned the "ecm reverb" yet.

to those of you who works at lex, can you hear what preset they use on ecm recordings? i e by engineer jan-erik kongshaug and/or manfred eicher?

For those who doesn't know:

Ecm is a small record label (german/norwegian) specialising in chamber jazz music, and uses a lot of silence and - especially - space in their recordings. I e lots of ambience added to their production. Known artists are keith jarrett, pat metheny and so on. Critics have been labeling the label such as "the most beautiful sound next to silence" and "the only label not in need to ever remaster their back catalog". But that's another thread, topic and discussion.

but to me, that label has always been the one to promote and pitch "lexicon reverb" sound the most, to me anyway. A whole demo label for lexicon, so to speak. Although, in some instances, too excessively. Violinist paul giger, and some pat metheny group productions comes to mind, for overdoing it. Pat left the label when he thought manfred eicher went over the top adding excessive reverb to his material.

can you hear that they're using (or was) lexicon units at all? And what presets or algorithms?

/honch

EMT 250. It has only one reverb algorhythm available (on a button marked "REVERB").

On the other hand, Windham Hill Records used a Lexicon 224 on all of their early 'new age' albums (George Winston, etc). Three years ago I nearly purchased that very 224 from Steven Miller (the engineer on all those records), but alas, he sold it out from under me...

[Honch]

Originally Posted by zmix

EMT 250. It has only one reverb algorhythm available (on a button marked "REVERB").

On the other hand, Windham Hill Records used a Lexicon 224 on all of their early 'new age' albums (George Winston, etc). Three years ago I nearly purchased that very 224 from Steven Miller (the engineer on all those records), but alas, he sold it out from under me...

Ok, it was a EMT 250... one algorithm. I thought that ECM was Lexicon all the way.
Ahha, so WH used a 224? I heard difference from them anyway. ECM and WH. They're different music altogether in spite of using mostly acoustic instruments. I think that ECM one, EMT 250 must have been swapped out over at ECM (or Rainbow studios in Oslo) a long time ago. Although ECM started out in 1969 it sounds as if - almost - they were made for big ambient production from the start on, and when EMT 250 came around, it must've been their wet dream come through. I'll bet they're using Bricasti nowadays... no seriously, they can use whatever. TC if they like. And also I think that - today - you can mimick the EMT250 algorithm inside any PC and software/convolution reverb anyway. Not to one hundred percent though, but damm near close. Get a decent one anyway, but without all the editing tweaks of a real EMT 250.

/Honch

[seancostello]

Reviving a dormant thread...

I have already asked about the difference between the 480L Random Hall and PCM90/91 Random Hall, but Casey was unable to answer, probably due to agreements with ex-employers, trade secrets, and the like. So, is anyone else able to describe the differences between these algorithms? I'm not sure if anyone who owns a 480L would buy a PCM91, but if anyone has both, can you describe the sonic differences between these algorithms?

Thanks,

Sean Costello

[DanDan]

Money doesn't talk, it shouts! I have used the 480L extensively and I own a 91.
I doubt if I am technically correct, but my opinion of the sonics goes like this:-
The 480 has say 4 engines. Two full stereo reverbs.
The 91 has two engines, which can be harnessed together in a single reverb.
These two 91 engines working together on one reverb sound like a single one of the 480 engines.
Not bad at all, but the 480 absolutely rules.
DD

[Froombosch]

Quality of reverb algorythms is not always depnding on brute CPU force. Remember that the 250 is a loved reverbmachine with relative few (K)MIPS. Good algos with low CPU power are better then a supercomputer with a lesser algo. It is the sound that counts in the end.

[3rd world order]

Originally Posted by DanDan

Money doesn't talk, it shouts! I have used the 480L extensively and I own a 91.
I doubt if I am technically correct, but my opinion of the sonics goes like this:-
The 480 has say 4 engines. Two full stereo reverbs.
The 91 has two engines, which can be harnessed together in a single reverb.
These two 91 engines working together on one reverb sound like a single one of the 480 engines.
Not bad at all, but the 480 absolutely rules.
DD

yeah, the 91 is wimpy.... i actually liked the sound of the Nuverb TDM card more... the only thing i liked in the 90/91 better was having stereo width on a 360 degree LFO.. and that was for one crazy WTF effect.... i actually found this used on a record if anyone is curious what it sounds like...

it's on the NIN/Oliver Stone/Natural Born Killers soundtrack... i'll have to find the specific track.

I still love the PCM 60 for it's simplicity..... same reason i eventually want a princeton/eventide 2016.... one knob, one function, learn your instrument, no pages.... it also forces me to get a unique setting each time... interact with the music.. make a unique decision... instead of dialing up a preset.

it's the difference between having a modular synthesizer and a rompler.... i also like to "play" the knobs when i'm printing effects so i can emphasize lil things and make ear candy... all the type of stuff which comes out in repeated listening...

or stuff you'd do with a LARC/midi slider box.

one of the coolest things about the LXP-15 was the FIVE cv pedal inputs..... you can modulate parameters with a modular synth or cv/gate sequencer... actually, erase that from your mind... the price of LXP-15s will go thru the roof now like the jam mans did a few years ago.

these arent the droids you're looking for *waves hand*
you don't need to see his identification *waves hand*
he can move along...

[Michael Carnes]

Originally Posted by seancostello

I have already asked about the difference between the 480L Random Hall and PCM90/91 Random Hall,

They're very close, Sean. The PCM90/91 Random Hall came from the 300 Random Hall, which is very, very close to the 480L. The PCM90/91 has a 20-bit memory as opposed to the 18-bit memory in the 480L. It also has an extra bit in its coefficient architecture. These result in a slightly cleaner sound in the PCM version.

[Froombosch]

Thanks for sharing this info!

[seancostello]

Thanks, NS!

[jnorman]

i see a few comments on how lame the pantheon plugin is/was. it has now been updated with the pantheon II plugin - has anyone used it, and is ti any good? thanks.

[audio ergo sum]

Originally Posted by Nobody Special

They're very close, Sean. The PCM90/91 Random Hall came from the 300 Random Hall, which is very, very close to the 480L. The PCM90/91 has a 20-bit memory as opposed to the 18-bit memory in the 480L. It also has an extra bit in its coefficient architecture. These result in a slightly cleaner sound in the PCM version.

Nobody Special, I recently worked with a PCM 96 and found the manual very short, missing information about the preset's parameters.

Do you have a list of all the parameters of the presets in a handy excel spreadsheet on Lexicon's website?

Not everybody has time, to listen carefully to the difference between RandomHall 2 and RandomHall 3 etc.

That would be something very easy to do and of great help, also if you could include a column for remarks, pointing out those presets that were created after 480 or 960 predecessors etc.

Does this make sense to you? I find the manual disturbingly short and simplistic and when it comes to describing the presets there is nothing.

[Michael Carnes]

Originally Posted by audio ergo sum

Does this make sense to you? I find the manual disturbingly short and simplistic and when it comes to describing the presets there is nothing.

Makes all sorts of sense, and I'm sorry we don't have such a document for you. To be completely truthful, I'm not sure when we'll have it. Only thing I can suggest is simply to audition the presets with your source audio and choose what you like. In the end, that's the measure that matters.

To carry that a bit farther, the names themselves are descriptive up to a point. You know a percussion-labeled preset will be more diffuse so that transients don't rattle around. You know that a vocal-labeled preset will be less-diffuse, with perhaps a hint of discrete echo. But I'm not at all sure how I'd draw a meaningful text distinction between ElephantShoe1 and ElephantShoe2. They'll be similar, but depending on your particular snare drum (or Ophicleide) either one could be the preset of choice.

N.S.