I don't see why the h3000 factory should sound different than the hardware, it's the same algorithms. There's a shootout video on YouTube as well, they sound exactly the same. You don't get all the features of the hardware though.

Does Eventide claim that the algorithms are the same? The output is unquestionably not. I've put the hardware right next to the software and there's a wide hole between them in terms of quality of sound, and it doesn't require much work to recognize. I very much wish that weren't the case. I think I saw the video you're referring to and the sound was not the same; even the comments were relatively unanimous about that.

Well, they say 'The H3000 Factory is a re-creation of several key algorithms from the H3000 Harmonizer® effects processor." I guess I took that to mean it's a 1:1 port of those algorithms. What's the alternative, they don't have access to those algorithms so they had to do it all over again and tried to get it as accurate as they could??

My guess is that they were programming the initial batch of these plugins at a time when CPU horsepower was much more limited than it is now and they felt the need to cut corners. If that's the case, I wish they'd revisit those great algos and rethink the plugins. On the other hand, they're still a hardware company, so they don't want to rob Peter to pay Paul, at least not too much.

I heard about this thread, so I thought I'd drop in for a correction or two. There's apparently a video critique in which the video maker claims the plugins don't match the PCM96. I'm afraid he made the mistake of not checking to assure he was comparing the same algorithms. Parameters are not identical between the algorithms on either the box or the plugs, so it's easy for even an experienced person not to notice just where they are. But as I've said many times, the software matches the hardware. There may be some reference level differences on the DA/AD that cause apparent sonic differences, but once that's accounted for, they should really sound the same or very very close.

I've never hidden this, but I should remind folks that I didn't retire from Harman. They retired from me. Along with hundreds of other Harman people on the same day, I was called into the office and sent packing. They were very nice--no security walk-out or anything like that. But I think they believed that someone in a lower-cost area could do my work and that of many talented others who were called into those meetings. I suspect that a visible termination like mine was also intended to send a message to the others in my group (who indeed did leave in the months that followed). I carry no anger or resentment about any of that. And any personal distress I may have felt about the current state of the Lex plugs has long passed.

After a year of contracting during the day and experimenting at night, I realized I could make reverbs with the characteristics I wanted without encroaching on Harman IP. Over the next years, my creativity went into Exponential Audio. If I was capable of 'new and exciting things' you'll have to look at my output there.

But I am now truly retired from tech. My long career was quite unexpected. My training was in classical music and I was lucky enough to be given chances to figure out what I was doing with computers. I've now done just what I planned over all those years. I'm again part of the local classical community in many ways--as a recording engineer, board member, and composer. I'm happy with what I accomplished in tech (there was a lot more than just audio), but I always intended to keep a little in the tank for myself and my family.

I wish you all luck in your attempts to prod Harman/Samsung or to find substitutes that work for you. It's certainly unfortunate, but audio companies do come and go. That was true 40 years ago and it's true now. Stay safe!

Hi. Thank you for sharing this and I wish all the best for you. Cheers

Thanks for the added clarity, Michael!

No doubt about it, Exponential has always felt like an evolution of your previous work. A next step up the ladder. A new and exciting thing, for certain. The retirement I was alluding to above wasn't in reference to leaving Harman, but in reference to the more recent announcement. The EA products are great achievements, but as I've mentioned, I wonder what else might have come from your mind. More new and exciting things. I'm thankful for what we have. You've left quite an impressive bundle of products in your wake, many of which are still in use today. Enjoy the much deserved you-time!

I submitted some questions to Eventide a while back when the factory plugin came out as it didn't line up to the hardware in certain ways that made me think it wasn't a port (not sure they ever said it was a 1:1 anyway, but stuff like LFOs being the wrong polarity and the filters not sounding close when at the same resonance)). Can't remember what the response was though. Plus the converters and surrounding hardware probably play a bigger role in the sound.

In the thread about the PCM package there was a lot more in depth discussion about how the plugs were made though, so different story trying to say they aren't the same code.

Thanks for the post Michael, hope you're enjoying your current endeavours.

Help me understand this better. Let's say Lexicon or Eventide wants to move the algorithms from their hardware into a plugin and their goal is to make them sound as close to identical as possible. I don't think it's just a matter of taking the original algo and copy/pasting it over to the AAX/AU/VST/etc. form, right? At least not in most cases. And I know this procedure varies depending upon the architecture of the hardware vs. the architecture of the software, but in *most* cases, aren't you required to emulate the behavior of the hardware in a different way within the confines of the software? I'm using general terms so that the specifics don't cloud the question, which, I know, will be tough to answer simply, and also ignoring the effects of any D/A converters or even the sound of digital outs.

I'm trying to understand how someone like Weiss can claim a 1:1 port when other companies seems to spend a lot of time trying to figure out how to recreate their products in software form.

Think it's as you mentioned already. Depending on the architecture some instructions can be more directly ported. Probably not as simple as a copy and paste but generally running the same processes.

Others probably need to be recoded by scratch to mimick what was happening.

And some might need to emulate the processors themselves, maybe others don't.

The Weiss says it's a port which just means it's adapted, line by line to work on a modern computer processor. So it might not literally be the exact same lines of code, but the same operations are happening.

For instance to print 'hello' in BASIC is different than doing it in PHP. But you could port the print 'hello' command from one to the other to achieve the same result. The exact line of code will be slightly different but the outcome is the same.

Then you have something like a ROM emulator for gaming which does use the exact code from the ROM, but emulates the processor to run that code (as opposed to a port of a game where it's re-coded to play the same but doesn't use any of the original code at all)

I'd imagine most plugins are doing the former, just porting the lines of code to modern language to do the same thing. Then on top of that they might emulate the DA etc... to add artifacts (like how a game emulator will add scanlines to look like it's running on an old monitor)

Off the top of my head, other ones that claim to be a 1:1 port besides Weiss/Softube is SSL Duende, from their broadcast consoles, and Sonnox from the Oxford console.

It's an interesting point, the ambiguity on what it could possibly mean. Isn't even having AAX, VST, and AU plugin formats the same issue? It's not exactly the same lines of code, because it's a different environment. The algorithm itself though, what operations it performs on the input, probably does have to change somewhat in different environments, for example a modern 64bit architecture. I never really thought about that, I just took Lexicon's and Eventide's word that it was the same. Even if they don't explicitly claim it's the same code they still promote these things as interchangeable with the original.

One last remark before I depart this thread. There's a pretty straightforward answer to your question. DSPs like the SHARC family have a limited amount of fast onboard data memory--just a few tens of K (more in some models). That's often sufficient for many processes like dynamics and EQ, but it's not nearly enough for a reverb. You can basically just port the code over for low memory processes. But for a reverb or delay, you'll have to use offboard memory (SDRAM or DDRAM). Accessing it directly (as you might in a plugin) is pretty slow and bogs down the processor, so you typically use DMA (Direct Memory Access) which uses a dedicated sub processor on the chip. It's slow too, but not quite as slow. Each pass of the process will handle a block of samples at once (for speed) and will also set up the chain of DMA operations for the next block. The most efficient use of such a processor may mean you're doing the number-crunching part of one process at the same time the DMA is running for another. You basically alternate between the two. That actual number-crunching code can be identical to the same code in a plugin, but setting up the DMA can be a bear unless you've been doing it for a while.

Intel (and ARM) chips have much larger caches and much faster access to off-chip memory if you just keep a few things in mind. This is why you can run so many more copies as a plugin than you usually can on a supposedly-faster DSP. That's been true for more than a dozen years.

So should the Lexicon suite be pulled from sale?

Modern computers tend to run floating point DSP processes. Either 32-bit (single precision) or 64-bit (double precision). There are some possible quantization and distortion issues that you can run into with single precision floating point and some recursive filters. But, for the most part, if your algorithm is theoretically stable, it will be stable and clean in floating point.

Older digital effects, for the most part, use fixed point processing, with much less precision than 32-bit floating point. The H3000 discussed above originally used 16-bit DSPs, 16-bit convertors, and wrote the delay memory to 16-bit integer buffers. This sort of processing will introduce a TON of quantization noise and error, as well as clipping (in the BEST case) or wraparound distortion (if saturation math isn't enabled).

In order to get anything remotely decent out of older fixed point hardware, all sorts of clever tricks had to be used. The order of operations was critical, as some sorts of filters and feedback operations will clip if executed in one way versus another (see: direct form I filters versus direct form II versus transposed direct form II). In order to avoid overloads or wraparounds, the gain often had to be reduced before an operation, which will result in less precision for fixed point signals.

Analog processing was often used in conjunction with the digital processing. It was very common to have pre-emphasis filters on the input (to boost the high frequencies and/or cut the low frequencies) and de-emphasis on the output (the complementary filter of the pre-emaphasis) in order to avoid distortion while reducing noise. Other weirder analog tricks were sometimes used, such as instantaneously tracking the input signal level and setting a "gain bit," in order to use a cheaper 12-bit convertor and increase the perceived signal to noise level. This can be viewed as floating point math of sorts.

Older digital boxes had almost NO signal processing horsepower compared to today. The idea that older digital boxes were more powerful than modern computers is just straight up wrong. The algorithms used in those boxes therefore had to use the less expensive computational options. The H3000 used linear interpolation in the modulated delay algorithms that worked at a fixed sampling rate, which caused a noticeable high frequency cut when the modulation depth was turned up from zero. The swept filters in the H3000 were most likely a variant of the Chamberlin state variable filters, which are computationally efficient and sound great, but will blow up if the cutoff is set above 1/6th the sampling rate. Which is probably why the filters in the H3000 max out at 7 kHz or 8 kHz (I forget which). In some older Lexicon boxes, the modulation was generated by an 8 bit Z80 processor, which resulted in a lot of noise, as the modulation waveforms had 8 bit stair steps in them.

So, the algorithm designer for older digital hardware was always trying to balance computational efficiency with sonic clarity. There were no "perfect" solutions, only compromises. The end results were as clean as they could be, given the circumstances, but they were far from pristine.

It turns out that a lot of people LIKE sounds that aren't pristine. The older digital boxes had some dirt, noise, reduced high end, stuff like that. What some people perceive as flaws, other people view as character.

So, long story short, a port of a basic algorithm topology from an older digital box to a modern plugin needs to take into account the importance of the artifacts of older fixed point processing. A floating point port might avoid most to all of those artifacts. Either the plugin needs to be created in fixed point (which is a real pain, as modern CPUs don't have the saturation features of older/custom DSPs used in audio signal processing hardware), or those artifacts need to be emulated, or the people porting the algorithms can decide "you know what? I always HATED that noise/clipping, and always wanted things to sound THIS way."

The PCM96 is a totally different story. It was programmed on a very high speed 32-bit floating point processor (TigerSHARC). Any differences between the hardware and the PCM96 plugins was either a difference in features/parameters, an error, or a bug fix when porting to the plugin. There is literally no magic mojo in a TigerSHARC - it's just a high speed floating point number cruncher.

P.S. If you want to learn more about what programming fixed point DSP was like in the 1980s, there's a great paper from the folks that programmed the H3000: https://www.aes.org/e-lib/browse.cfm?elib=5449

Thank you, Michael and Sean! That's invaluable information.

Short of a fix, a disclaimer would be the responsible thing to offer, in addition to a complete refund in the event that these plugins aren't operable.

Fascinating stuff! Thanks for your insights Michael and Sean

Hey Michael! It's great to read your words here. It's lovely to see you are now retired and fully enjoying life.

To me and I'm sure to many other fellow gearslutz, you are truly Somebody Special.

Anyone using these in Cubase 11? They come up as invalid plugins. Wavelab 10 loads them fine (and so did Cubase 10.5 on my old Win 7 install).

update: got it working eventually after installing again.

I remember as a new GS member 10 years ago the thrill of realizing who Nobody Special was. That's probably what got me hooked on GS.

Your Exponential Audio reverbs changed my mixes forever… some of the best reverbs ever made, hardware or software. Still use them on every mix!

Buddy, do you kidding or what. In all of plugins of the Lexicon FX bundle is broken polarity switch handle (that presents in each delay line of almost all plugs of this bundle), ie it does nothing. And moreover the sound passing through the effects comes out inverted in phase. I already wrote about it five times. "Flawless".

You're wrong

The title of this thread is "Lexicon PCM Native REVERB Plugin Bundle."

https://lexiconpro.com/en/products/p...plug-in-bundle

https://lexiconpro.com/en/products/p...plug-in-bundle

With that said, are there PCM Native Reverb issues that apply to Windows usage?

If it addressed to me, i can provide proof like video with analyzer. As well, other people (for those who are interested ) already did check as well, and confirm.

PCM reverb bundle - NO, there is no such issues. But in mesage that i quoted can notice mention about fx bundle separately too.

I'm sure we can agree that it works for some and not for others and then move past this need to classify it as either globally functional or globally broken, because that seems to be the situation. Those without issues, consider yourself fortunate. Those with issues, let's hope something is done to get us back in business.

For these who at Windows at least (and use latest version of FX bundle) - it can't work , by objective reason, because this feature is broken, at plugin level. This is regardless system config, PC condition, luck, etc.
How it does at MAC, i'm don't know.

The issues are all too real.

I'm on a mac, and for years now, any time you remove any one Lexicon plugin from a session, it beachballs and hangs for a few moments. Worse, when closing a session containing multiple instances, you get that hanging X the number of instances of lexicon plugins in the session.

This means you get weird hanging every time you close any session that includes lexicon plugins.

But the 'hanging' is when it's being Nice. Because when it's not hanging on removal, it's crashing on removal.

I haven't had issues loading sets containing them, but closing sessions is always an issue: I cringe, waiting for 1) a long hanging wait or 2) a full crash.

But it also makes just trialing different plugins (after all, there's a lot of individual plugins in both collections, verb and FX) tedious, due to all the hanging.

I own a zillion other plugins and none of them do this. And it's been this way for years. So, so lame.

I'd say I hope for a fix, but why bother? After this long, I guess it's better to just move on.

After 10 years of using the plugins, I decided to "future proof" the best I could: I just bought a PCM-92.

I am running it via AES / EBU on a Dante system on Windows, and the System Architect software works just like a remote from my desk. Actually it works almost identical to the Plugin.

Do the hardware and the software sound the same? No. Not to me. I have A/B'd the same preset, using the same algo with the same source switching back and forth on headphones. Level matched.

There is a difference. This is running digital I/O for the hardware so it is not a conversion thing either. I could, blindfolded, pick out the hardware / software 100% of the time with the Random Hall algo loaded and the same preset / settings. At least I *think* they are the same settings...I did not go through every page to compare.

Is one better than the other? Again....no. I think it would depend on your desired sound. The plugin (to me) sounds ....smoother in the tail and slightly darker overall. There are some reflections or something going on in the tail of the PCM-92 that are more active and 'edgy' as the sound trails off...almost like a reverb spiccato. Poor words to describe....but the best I have.

Keep in mind, my tests were done on a solo'd source in isolation. Would anyone ever know the difference in a mix? Maybe Michael would...but I cannot imagine anyone else would be able to pick one from the other in that scenario.

I will still happily use my Lex Plugins until they no longer function...and then I will happily use the hardware until it too...no longer functions.