Bricasti M7 Native Plugin

[izonin]

M7 v.1 doesn't have modulation, v.2 does.

Both v1 and v2 have modulation.

Most people associate modulation with chorusing, and v.1 doesn't have that. v.2 introduces Lexicon-ish type of verbs.

[Warp69]

Most people associate modulation with chorusing, and v.1 doesn't have that. v.2 introduces Lexicon-ish type of verbs.

I'm not in a position to comment on people's definition of modulation, but almost every hardware reverb company (except TC Electronic) haven't used chorusing as modulation in their high-end dedicated reverb units since 1985.

I'm not sure what you mean by v1 and v2 regarding Bricasti M7 - OS version or algorithms (Bricasti refer to algorithms)?

[izonin]

I'm not in a position to comment on people's definition of modulation, but almost every hardware reverb company (except TC Electronic) haven't used chorusing as modulation in their high-end dedicated reverb units since 1985.

Yes, the oldschool Lexicon verbs.

I'm not sure what you mean by v1 and v2 regarding Bricasti M7 - OS version or algorithms (Bricasti refer to algorithms)?

There's a lengthy discussion about this on gearslutz.


ps: So when can we expect a Relab MX7? You're an expert at reverse engineering and own the hardware.

[Andrew Souter]

anyway, the point is, all these talks about "custom DSP chips" being better than a modern CPU are, in most cases, hogwash.

agree...


The only rational explanation for custom hardware is that it can not be pirated. i.e. it is a business decision, not an engineering one... Nothing wrong with that (wise in some ways), but the latest Intel CPUs are certainly NOT lacking in power...

B2 1.2.0 incidentally can use easily more than an order of magnitude more than 30sec of total delay memory... (assuming that is what this "30 sec" stat is measuring?)

..and it can also "just barely fit" on a single core, if you start maxing out all the various settings.

...or run several hundred instances on the latest CPUs using modest settings...

[Uncle E]

Ummmm and how exactly are you getting modulation inside your IR tails? If you mean on some pseudo chorus thing before or after IR signal, then no my friend that is far far away from real reverb tail modulation and IT IS NOT how original M7 or any other reverb (for that matter) behave..

Maybe not but some good sounding reverbs could probably be created by layering several instances of an impulse, each modulated independently. Is there currently an impulse plugin that allows impulses to be used in this way?

[Uncle E]

You have GPU's for a reason, it is a processor highly optimised for a task at hand.

It'd be interesting if the new Mac Pro's GPU could somehow be harnessed by Bricasti.

[ObsoleteAcc99]

[DELETED]

[valhallasound]

Most people associate modulation with chorusing, and v.1 doesn't have that. v.2 introduces Lexicon-ish type of verbs.

I'm not in a position to comment on people's definition of modulation, but almost every hardware reverb company (except TC Electronic) haven't used chorusing as modulation in their high-end dedicated reverb units since 1985.

This is true of the 480L, and this is part of the reason I'm not much of a fan of the 480L and later Lexicons.

The original reverb algorithms for the 480L (the ones that aren't Random Hall or Ambience) have NO modulation, and just "sit there" to my ears. These algorithms seems close to the Rich Chamber and Rich Plate on the 224XL, but the 224XL used chorusing modulation for all the reverb modes.

The 480L and some of the other high end Lexicons (including my 300M) used different modulation types. Random Hall and Ambience don't use identical techniques, but they are both closer to "granular" than chorusing, in my opinion. I know that people love these algorithms, and they do a better job of preserving the pitch of the input, but at the loss of a great deal of richness.

I'm not putting Warp69's LX480 in this category. IIRC, the LX480 has the option of adding chorusing to some of the reverb modes that didn't have modulation in the hardware 480L.

The 480L Classic Cart added Concert Hall and Rich Plate. I am presuming that both of these use the 224XL type chorusing. The studio I was in Monday had a 480L with Classic Cart, but I spent the day focusing on the 224XL. I LOVE the 224XL. Not just for the chorusing, but for the diversity of algorithms. The 224XL has the 224 algorithms, as well as Rich Plate and Rich Chamber, and a bunch of oddball variants of the 224 algorithms.

As far as other high-end reverb units since 1985, what other units do you (Warp69) consider "high end" from that time period?

Sean Costello

[valhallasound]

anyway, the point is, all these talks about "custom DSP chips" being better than a modern CPU are, in most cases, hogwash.

agree...


The only rational explanation for custom hardware is that it can not be pirated. i.e. it is a business decision, not an engineering one... Nothing wrong with that (wise in some ways), but the latest Intel CPUs are certainly NOT lacking in power...

B2 1.2.0 incidentally can use easily more than an order of magnitude more than 30sec of total delay memory... (assuming that is what this "30 sec" stat is measuring?)

..and it can also "just barely fit" on a single core, if you start maxing out all the various settings.

I remember programming 256 randomly distributed delay taps on a 600 MHz single core Blackfin back in 2005, from a 1 second stereo delay buffer. IIRC, it took about 10% of the processor. It's not just about delay memory - it is how many taps you can read from that memory. The M7 could easily be reading out several thousand taps from the delay memory, which would put far more stress on the cache of an Intel processor than having several dozen to a few hundred longish combs (and a Hadamard matrix combining them) that take up the same amount of delay memory.

An interesting test would be to run "velvet noise" reverbs on a given Blackfin and a given Intel processor, and see how many taps & what length of reverb could be run on each without choking. The Blackfin code would need to be 32-bit fixed point for audio quality, but could use 32x16 bit multiplies for the tap coefficients (technically speaking, velvet noise could use far lower tap coefficient values, but let's ignore that for now). Both the Blackfin and Intel could use block based audio processing for maximum efficiency, with the block size kept at a reasonable level for low-latency audio.

A few additional tests that could be run on each platform to test the "reverb power":

- How many modulated allpass delays can be run in series. The lengths should be in the 100 msec range, with different lengths for each allpass.
- Hadamard matrices. How large can you get?
- Other matrices.
- 1st and 2nd order filters in parallel.

By running the above tests on both Blackfin and Intel, and combining them while running at the same time, a realistic metric of the power of each platform for a given MHz & cache size could be achieved, while being relatively agnostic as to the type of algorithm being implemented.

In lieu of such a test, I'm gonna go ahead and take Casey at his word. My experience with programming reverbs on both the Blackfin and native, and my calculations of the performance of the M7 based on the published Blackfin specs, suggests that running such an algorithm in real time on any modern Intel-based computer would be problematic.

I like native reverbs. I make a living from them. However,my liking of native reverbs doesn't change the raw numeric performance of the given platforms. This is one of those areas where there are some OBJECTIVE numbers, as opposed to just SUBJECTIVE opinions.

Sean Costello

[Warp69]

As far as other high-end reverb units since 1985, what other units do you (Warp69) consider "high end" from that time period?

From 1985 - 2013? M300, 480L, 960L, S777, R880, Yardstick, M7 and many others. The original 480L algorithms do have modulation which is input depended.

Great post and I agree - it would be interesting with a test between platforms.

[Andrew Souter]

Great post and I agree - it would be interesting with a test between platforms.

Sure. Let's see them. Sean, you seem to have some inside insights having worked at AD. Maybe you want to make them?

[valhallasound]

Great post and I agree - it would be interesting with a test between platforms.

Sure. Let's see them. Sean, you seem to have some inside insights having worked at AD. Maybe you want to make them?

I don't have the bandwidth for such a project right now. I'm working on other stuff.

My point was, without these sort of metrics at hand, or a native equivalent of the Bricasti that shows that this type of sound can be obtained on an Intel PC/Mac, Casey's analysis of the subject should be taken as fact. The man knows more about the Bricasti than we ever will.

Sean Costello

[Andrew Souter]

I don't have the bandwidth for such a project right now. I'm working on other stuff.

AAX? Fun stuff huh?

Definitely superior to RTAS though...

My point was, without these sort of metrics at hand, or a native equivalent of the Bricasti that shows that this type of sound can be obtained on an Intel PC/Mac, Casey's analysis of the subject should be taken as fact. The man knows more about the Bricasti than we ever will.

Of course he does. I don't claim to know anything about Bricasti. I did not say or infer that I do.

What I said were simply two things:

1) I personally believe custom DSP solutions are pretty much obsolete... all of them. Not any one in particular.

2) If we are saying total delay memory is an important metric (I am not sure that it is), I was simply mentioning that B2 can approach something like 500 sec, which seems larger than 30 to me...

[valhallasound]

I personally believe custom DSP solutions are pretty much obsolete... all of them. Not any one in particular.

I would agree, with the possible exception of the Bricasti. The other "flagship" hardware reverbs either have equivalents in the native space (Lexicon), or use DSPs that can clearly be outpaced by modern CPUs (the 4x100 MHz 56K DSPs in the TC 6000 aren't very powerful by modern standards). The Bricasti's hardware is way above the class of any of these older reverbs.

I would also say that "obsolete" isn't necessarily a bad thing. I was using a 224XL on Monday that was almost 30 years old. The Bricasti will still be running 20 years from now. Our plugins WON'T, unless we are actively updating them. 1/4" and XLR inputs are ancient standards, that are far more robust and stable than plugin "standards" that are constantly being changed by the keepers of those standards.

AAX? Fun stuff huh?

You KNOW it! Add this to sandboxed plugins in Logic X, and the "retirement" of VST 2.4, and the result is some annoying times for plugin developers. I'm adding some new algorithm modes to these forced updates, just so I feel like I'm not just running to stand still.

Obsolete hardware is looking real good right now. I'm working on some Z-DSP algorithms, and that chip was BORN obsolete. 128 instructions per sample, but with some old skool design tricks to squeeze the most juice out of those instructions, Delay pointers are automatically incremented, 2 instruction allpasses, built in sine/cos and ramp LFOs that generate linear interpolation coefficients, that sort of thing.

You have to get into the right frame of mind to design for the FV-1 DSP used in the Z-DSP, give up a lot of your ego and expectations, and "go with the flow." It's like a tea ceremony for DSP developers. Limitations lead to creativity. Much different than the infinite blank page that greets us when we open up Xcode or VisualStudio.

Sean Costello

[Aiynzahev]

Limitations lead to creativity. Much different than the infinite blank page that greets us when we open up Xcode or VisualStudio.

Sean Costello

Dude, not to get you guys off topic, but this is why I've really been looking into rack effects. That and DAW stability and CPU overhead.

Great posts, post on!

Like me you've got music on the mind on the weekend instead of "fun" "normal" people have. haha!... unless you're getting ready to go out and I'm just a loser