Optimod 8600 comparison clips

[rorban]

Audio clips comparing the 8600 and 8500 are now available at:

http://www.orban.com/products/radio/fm/8600/comparison_audioclips/

These compare the 8500 GREGG and 8600 GREGG MX presets using a variety of well-known "challenging" material.

Bob Orban

 

[chriscollins]

Thanks for that. I'm downloading now.

My least favorite thing about my 8300 is the IM distortion. Especially when I use the bass clipper on hard (which I need to for the hip-hop content I play).

Now, is this new 8600 firmware? Seems like someone (maybe you) had mentioned that at the moment, the 8600 and 8500 are pretty much equal, with many new features to be enabled down the road because of the extra DSP horsepower in the 8600.

 

[rorban]

Thanks for that. I'm downloading now.

My least favorite thing about my 8300 is the IM distortion. Especially when I use the bass clipper on hard (which I need to for the hip-hop content I play).

Now, is this new 8600 firmware? Seems like someone (maybe you) had mentioned that at the moment, the 8600 and 8500 are pretty much equal, with many new features to be enabled down the road because of the extra DSP horsepower in the 8600.

This represents the current 8600 firmware. The 8600 and 8500 are not equal, although the 8600 offers all of the 8500 processing algorithms so that people with favorite 8500 custom presets can run them and so that stations requiring off-air headphone monitoring for outside broadcasts can switch to an 8500-style algorithm (which has much lower delay than the 8600 algorithms) for these situations.

The 8600 algorithm in these audio clips uses the new 8600 back end, which is very compute-intensive. In essence, this replaces the 8500's bass clipper, HF clipper, distortion-cancelled clipper, and overshoot compensator with entirely new code that introduces some novel structures and algorithms that have never been used in Orban processing before. We believe that it considerably improves the loudness/brightness/distortion tradeoff. It also improves transient punch while lowering distortion, which was a difficult design challenge because most low-distortion peak limiting algorithms wash out transients.

Compared to the 8500, the only change to the processing upstream from the back end has been to add a third choice for the band 1/2 crossover frequency, so 100, 150, and 200 Hz are now available. We were satisfied with the sound of the AGC and multiband compressor/limiter in the latest version of the 8500 and decided not to fix what was not broken. For example, by virtue of the design of the compressor/limiter coupling in the multiband compressor/limiter, this processing automatically adapts to the density of the input program material (and does so without adding extra artifacts), so there is no need to add this as a "new feature" when we have already been doing this for years.

In the future, the main addition that we expect to make is adding a soft knee option to the multiband compressor, which is something we did with our digital media processors (1101, 6300, 8585, 8685) some time back. However, soft knee is more useful for mastering and for very gentle processing than it is for the type of processing typical in FM radio, so this addition will mostly be of interest to those running "purist" formats. In the next year, we also expect to add some other features that our customers have been requesting. But the basic "sound" of the 8600 processing for typical FM applications is already there in our current release.

Bob Orban

 

[oldiesstation]

Now the 8600 has the "WOW" factor over the 8500.a vast improvement.This box has come to play.Congrats to the Orban team.Really enjoyed the clip comparisons.Now maybe we can get some clips of the Omnia 11 and the new Vorsis Air Aura,( which i hear is amazing).Good job,Mr. Orban,

 

[Lazy J]

I listened to the comparisons and I think the Green Day song (Track 7) gives the most noticeable difference. I did pick-up on some distortion with U2 - Electrical Storm (Track 2) though. Is that the song, the processing, or my ears?

 

[rorban]

I listened to the comparisons and I think the Green Day song (Track 7) gives the most noticeable difference. I did pick-up on some distortion with U2 - Electrical Storm (Track 2) though. Is that the song, the processing, or my ears?

As you start to push for maximum loudness, any processing algorithm will eventually start to produce audible artifacts. These clips represent a very "competitive" loudness level and use program material that is notorious for challenging older processing algorithms, including ours. As you can see from the image included with the clips, the peak meter on the Belar monitor used to make these clips stayed at 99-100% all of the time.

The 8600 was designed to have lower distortion than the 8500 while not introducing a new set of artifacts. As I mentioned in my post above, trying to reduce distortion to its minimum possible level seems to produce a sterile, washed-out sound that loses punch and adds unfamilar set of artifacts when you push for high loudness. The 8600 is designed to trade off punch vs. distortion in a more favorable way than any previous Orban processor and when the distortion is audible, it is designed to be as acceptable to the ear as possible.

Some of the early comments we have been getting from customers are that the air sound seems cleaner, more open, and less processed-sounding compared to the 8500. That was our goal -- to create well balanced processing that improves on the positive qualities of the 8500 without adding new negatives.

 

[fm-engineer]

I listened to these in the lab using reference monitors and equipment. The Green Day clip really tells the story. What an improvement on dense material. The other clips also have less audible IM artifacts, especially as you cross the mid to presence range. Nice sparkle on the highs!

 

[WNTIRadio]

Put them all up in my home studio on the big monitors... what a difference in the midrange. It's clear, and the highs don't have that squashed 8100A/1 HF limiter feel to them anymore. Of course, there will be some moaners who will say "but I WANT that sound". Well, to heck with 'em. Sounds markedly better.

I don't remember the website, but there was an aircheck posted from WYNY in 1981 and 1982 playing "Sailing". It pumps, squishes, swirls, distorts all over the place. Don't know exactly what they were using at the time, but it sounded like some DAP's with an 8000 to my ears. Anyone who thinks the art of processing hasn't come a long way is smoking their socks.

 

[chriscollins]

There is a substantial difference in those clips. You guys did a great job on that, Bob.

On a side note, I always enjoy the detail in the posts you make. I love reading about audio processing.

I do wish that you guys would use a few more Hip-Hop/CHR songs in the demo, as that stuff is all flatlined with clipping from the label. I would happily transfer you some problematic tracks in .wav format if you were ever interested. I play so many things that get affected very badly with IM distortion.

I guess I also have to accept that my 8300 has much less horsepower than the 8500 or 8600.

 

[gtomas]

I appreciate the clips, but I would have liked that the two presets used (8500 Gregg and 8600 Gregg MX) sounded more alike. There's a difference in texture in these two presets and they sound slightly different (in bass and overall), making it harder to compare the actual improvements in the HF. But for proper comparison one needs to get hands on the box anyway...


Regards,
Goran Tomas

 

[fm-engineer]

I would guess the new clipping and distortion management has something to do with the bass difference.

I appreciate the clips, but I would have liked that the two presets used (8500 Gregg and 8600 Gregg MX) sounded more alike. There's a difference in texture in these two presets and they sound slightly different (in bass and overall), making it harder to compare the actual improvements in the HF. But for proper comparison one needs to get hands on the box anyway...


Regards,
Goran Tomas

 

[rorban]

I appreciate the clips, but I would have liked that the two presets used (8500 Gregg and 8600 Gregg MX) sounded more alike. There's a difference in texture in these two presets and they sound slightly different (in bass and overall), making it harder to compare the actual improvements in the HF. But for proper comparison one needs to get hands on the box anyway...

Regards,
Goran Tomas

It is indeed true that a proper comparison requires access to the processor and the best that comparison clips can do is to give an impression of the improvements. Because 8600s are now shipping, users should soon have this ability.

The GREGG MX ("MX" indicating that the new 8600 back end is used) preset was based on the GREGG preset, but the thresholds of the various bands in the multiband were slightly tweaked to keep the overall frequency balance as close as possible to the original GREGG preset in the midrange and bass. To do this, we used a Stanford Research Systems SR785 dual-channel dynamic signal analyzer, programmed to display the difference in the spectral balance between the two processors via a third-octave real time display.

Some retuning of the multiband was necessary because the new back end processing provides a different frequency balance than does the 8500-style back end when both are heavily driven, as they are in these clips. However, all of the MX presets exploit the new back end's ability to stay clean while providing 2.5 to 3 dB more clean HF energy as indicated on the SR785, so there was no effort to dial back the HF energy in the GREGG MX preset to match what the GREGG preset provides in the 8500. For 75us preemphasis at least, the challenge has always been to get the HF spectral balance as close as possible to the spectral balance of the source material given the huge handicap that the 75us preemphasis/deemphasis process creates. The 8600 gets closer to this goal than any previous Orban processor.

As for the bass, the 8600's back end handles bass quite differently than the 8500 back end does. The 8600 has two controls that allow the user to trade off bass energy against distortion control, so the amount of bass will depend not only on the program material but also on the setting of these controls. Given that FM listenership on small radios (like clock radios) is much higher than on radios with deep bass response (at least according to CES category sales reports), I believe that it is important that the midrange stays as clean as possible because this is the attribute of the sound that will benefit the largest percentage of a station's audience. The 8600 has several MX presets that are tuned with this philosophy. There are also MX presets tuned to produce more low bass; Greg Ogonowski considers low bass to be important and contributed heavily to these presets. The great thing about digital processors (as opposed to processors like the XT2) is that we can create presets for all tastes instead of having to lock down the sound because giving users that kind of flexibility is too costly with analog technology.

Bob Orban

 

[stace]

Listening to the clips I find it hard to understand why someone would choose a preset, or push a processor so hard that it creates a sound like this. Our stations have a mix of Optimod and Omnia processing, and I wouldn't choose a favourite between them.
I like a shiney consistent sound that jumps out of the speakers, not one that is grainey and LOUD. The best way I could get this out of the 8500 was to defeat the lookahead processing, (I'm not saying its impossible but) with lookahead in, it would do some strange things to some audio, even introduce distortion where it wouldn't be when turned off. I like to hear people like Orban suggest processors focus more on making their processors sound better, or more exciting rather than LOUD. It sounds like the upgrade has put some punch and sparkle back into the Optimod which excites me.

 

[fugazi]

Bob,

Why using MIPS as a benchmark?

I can very easily do 300 MIP/S with 2GFLOP/S, but i cannot do 2GFLOP/S with 300MIP/S!!!

4500MIPS is a marketing statement, it doesn't say anything about the algorithms nor it doesn't say anything about the coding style of the programmer.

The algorithm ( with the 8600 audio clips ) stands very proudly for itself. Please remove the MIPS remark from the comparison chart. Talking MIP/S is an english style of presenting a product.....

Evert

 

[rorban]

Bob,

Why using MIPS as a benchmark?

I can very easily do 300 MIP/S with 2GFLOP/S, but i cannot do 2GFLOP/S with 300MIP/S!!!

4500MIPS is a marketing statement, it doesn't say anything about the algorithms nor it doesn't say anything about the coding style of the programmer.

The algorithm ( with the 8600 audio clips ) stands very proudly for itself. Please remove the MIPS remark from the comparison chart. Talking MIP/S is an english style of presenting a product.....

Evert

I believe that it is interesting to show MIPS to illustrate how far Orban processors have come in terms of raw processing capacity. The 8600 still has significant unused capacity to accommodate future feature additions. Because the DSPs used in Orban processors are Motorola/Freescale 24-bit fixed point DSPs, "FLOPs" (Floating Point Operations per Second) is a unit of measure that is irrelevant to these processors. All arithmetic in these DSPs is fixed point, and we use single-precision (24-bit) or double-precision (48-bit) arithmetic as appropriate to complement the algorithms and structures in use.

The DSP code is written in 56xxx assembler. Thanks to some expert DSP coders in our product development group, the code is very efficient--almost certainly more efficient than code originally written in C or C++. For what it's worth, the prototype of the new clipper was about 7000 lines of Fortran 95 and took about 40% of a 3.06 GHz Core Duo processor to run in real time after compilation by the Intel Fortran compiler, which automatically vectorizes as much of the code as possible so that it can use the SSE4 instruction set of the Intel chip. The prototype, which ran under Windows XP, used floating point arithmetic for the sake of expediency (i.e., I wanted to hear algorithms without worrying about potential internal clipping), and the fixed-point DSP port (which was hand-coded assembler using the Fortran as a reference) was exhaustively checked for overflow and compared against the Fortran prototype to ensure that the DSP was running the algorithms correctly.

One of the main advantages of using the Freescale DSPs is that they require no heatsinks or fans thanks to a relatively low clock frequency. Because fans are one of the main points of failure in digital electronics, I always prefer to use as few as possible. The 8600 uses no fans at all.

Bob Orban

 

[fm-engineer]

The DSP code is written in 56xxx assembler. Thanks to some expert DSP coders in our product development group, the code is very efficient--almost certainly more efficient than code originally written in C or C++. For what it's worth, the prototype of the new clipper was about 7000 lines of Fortran 95 and took about 40% of a 3.06 GHz Core Duo processor to run in real time after compilation by the Intel Fortran compiler, which automatically vectorizes as much of the code as possible so that it can use the SSE4 instruction set of the Intel chip. Bob Orban

Geek mode on please

#include <iostream>
{
cout << " And I thought 700 lines of C++ code was a lot back in college!" ;

if (You thought the 8500 sounded good)
{
cout << "You should give the 8600 comparison clips a listen!" ; ;D
}
return 0 ;