Orban’s first audio processor for FM broadcasting was launched 50 years ago. To mark the anniversary, the company is sponsoring a series of interviews of Bob Orban in conversation with Radio World Editor in Chief Paul McLane. (Read the series from the beginning here.)
Last time, Bob discussed Orban’s tenure as part of AKG Acoustics and Harman International Industries, and the introduction of the first DSP-based Orban processors. Below we talk about a new owner, digital radio and the introduction of the 8400, an all-in-one audio processor for analog and digital broadcasting.
Paul McLane: In 2000 Orban was purchased by CRL Systems Inc. from Harman. How did that come about?
Bob Orban: We were a very small tail on a very large dog with Harman, and they decided that they didn’t really want to put the effort into managing such a small part of their business. So they decided to spin off the Orban operation.
Around that time, Jay Brentlinger acquired the assets of Circuit Research Labs. Ron Jones had passed away, and it had become a non-viable business. Jay used it as a corporate structure in order to acquire Orban.
McLane: How did that affect you personally, did you have to move?
Orban: We had already moved out of San Francisco and were in San Leandro on Alvarado Street. We didn’t have to move out of that facility. CRL itself was based in Arizona.
Today most of our engineering staff work from home. The main office is in Pennsauken, N.J., and there’s an office in Germany.
McLane: How well did you work with Jay Brentlinger?
Orban: We got along well. Jay’s a nice guy. We had our disagreements, inevitable in any business relationship. But we got a lot done during that time, especially in product development.
We did have fewer resources and a much smaller budget, because it wasn’t a mega corporation like Harman. Eventually we had one DSP engineer in the United States who did the 56k code, and another in Germany who came with CRL’s acquisition of Dialog4 System Engineering.
But we were seeing competition in the processing arena, particularly from Telos, and it was clear that we couldn’t keep the 8200 going forever. It was manufactured for 10 years, a pretty good run for a product. But I’d started work on the 8400 during the last part of the Harman era, along with several other DSP engineers.
The Optimod-FM 8400 was designed for lower distortion. We introduced anti-aliased clipping, using an algorithm that I’d developed because we were getting beat up over that by our competition. And we were able to incorporate HD Radio processing as well as better PC remote. Harman had also gone in for a fairly expensive package redesign.
[Read a Radio World review of the 8400 from March 2001.]
The half cosine interpolation composite limiter was an important feature. We’d avoided composite clipping because of problems it caused with distortion and reduction in stereo separation. So I examined how you would do composite limiting without having these downsides.
We applied for a patent on the half-cosine interpolation composite limiter, which was granted as #6,434,241. That was the first thing I prototyped in Fortran, because it required a fair amount of subjective assessment.
McLane: That redesign you mentioned was distinctive.
Orban: We got that famous 8400-style package, with the little protrusion and the big control knob that stuck out the front. I think it cost around $50,000 to develop that. We kept it right through the 8700.
The 8400 addressed a number of competitive issues, and in terms of sales, it picked right up from where the 8200 left off.
It had a 32 kilohertz-based sample rate, like the 8200. This limited us to about 15 kHz audio bandwidth in the HD processing, and we got some competitive heat for that. In the 8500, which we introduced three years later, we doubled the base sample rate to 64 kHz. That brought the supported audio bandwidth of the HD up to 20 kHz.
There were other minor algorithmic improvements in the 8500, but mostly it was a tighter and more mature HD integration compared to what we had in the 8400.
McLane: Was there demand at this point for HD Radio processing? It seems like that was early days for the technology.
Orban: Major markets were starting to convert. They had to deal with the problem of unobtrusive cross-fading between the analog and digital channels, which required accurate time alignment and similar processing textures. It became obvious that the best solution, by far, was to have the HD and the FM processing coupled in the same box.
McLane: There were a lot of things happening at the company around the same time. Were you involved in the rollout of the Opticodec line of codecs?
Orban: No, that was done by the group from Dialog4, which CRL acquired in 2002. The Opticodec was developed by them in Germany.
But we acquired a couple of very good engineers as a part of that deal. They’re still with us to this day and have been contributing to the development of new Orban platform products.
An interesting thing about Orban is that the staff seem to like it here. Having people who have been with us for over 30 years is a rarity in today’s environment. They are loyal and seem to find it interesting and challenging.
McLane: The year 2004 brought the Optimod-FM 2300 and Optimod-TV 8382 introduced at the NAB show. Were some of these upgrades more incremental?
Orban: We were adding value-priced processors so people could afford Optimod processing even if they couldn’t go with an 8400 or 8500.
The 5500, which we introduced in 2010, was particularly successful. It replaced the 2300 and 5300 models, putting five-band and two-band Optimod processing into a single rack unit package. In terms of units sold, I believe that was our most successful digital audio processor. It had a price that medium and small markets could afford, and it still had very competent processing.
We had done TV processing, starting in 1981 with the Optimod-TV 8180A, which was an 8100 without a stereo generator. Then two years later the 8182 added a CBS loudness controller, licensed from CBS Technology Center, as well as the Hilbert-Transform clipper, which is optimized for low distortion on speech.
We were very successful with that product in the analog TV days. So we wanted to follow that up with a digital TV processor. The first one of those was the 8585 in 2008. It featured two-band and five-band audio processing for 5.1 and 7.1 surround sound broadcasting, netcasting and mastering.
Its weakness was it didn’t have SDI input/output, so we followed in 2011 with the 8685, which had HD-SDI I/O and was more suited for the facilities at the time.
That was an interesting project because I had to deal with designing audio processing for surround for the first time,
McLane: I was going to ask you about surround. What did that entail?
Orban: A whole bunch of thought processes! There were a lot of technical innovations in the 8685. How do you optimize for sound-for-picture style audio, which is obviously different than the program material that’s usually passed on radio? How do you deal with dialog intelligibility in particular?
I came up with a way of making sure that the dialog didn’t get buried and that audio in one channel didn’t objectionably modulate the audio and other channels. Multiband compression would automatically re-equalize muddy-sounding dialog, and there was also a program-adaptive high-frequency enhancer available that could dynamically boost high frequencies appropriately.
Additionally, a loose interchannel coupling process in the multiband compressor allowed the center channel in each frequency band to take additional gain compared to the other channels when the center channel had too little energy.
The SDI implementation took a large amount of engineering resources because we had to do it with FPGAs — the product was developed just before dedicated HD-SDI chipsets became available. Eventually we had problems getting certain parts and we had to discontinue the product.
McLane: Was radio still in the era of loudness wars, or were things becoming more nuanced?
Orban: We had to offer a product that could compete in the loudness wars, if people wanted to do it. But with all the consolidation, there was less pressure than there had been, say, in the 1980s for every station to be loudest in the market, because that’s self-defeating. Every station obviously can’t be loudest, and eventually you get a race to the bottom in terms of quality.
Groups like Clear Channel and CBS now could dial it back a bit. But nobody wanted to fall off the dial. We still had to maintain competitive loudness.
McLane: In 2010 Orban introduced a new flagship, the Optimod-FM 8600. I see in my notes that it had five-band and two-band processing for both analog FM transmission and digital media. Why does that project stand out for you?
Orban: We again were getting serious competition from the Telos product line, and I was looking for the next step in peak limiting.
I spent about a year researching and finally came up with a rather complex block diagram, which we called the MX limiter. It used a psychoacoustic model, similar to a codec, to estimate whether distortion was psychoacoustically masked in various frequency bands, and to take mitigating action if the distortion otherwise would have been audible.
We introduced that with the 8600 and it made a big difference. It was a noticeable step up compared to the peak limiting in the 8500. It had almost 3 dB higher high-frequency power handling capability, better transient preservation and lower perceived distortion. It also made an exceptionally good tradeoff between speech and music processing. We didn’t have to compromise one for the other. Again this was a very successful product for us.
McLane: At this point, was it you and a team of engineers, or mostly a solo show? How is development working?
Orban: For the 8200 we hired a young DSP engineer named Paul Neyrinck. He coded the DSP in the 8200 and subsequent processors, then left to start his own company.
Harman hired four other DSP engineers who also worked on DSP for the DSE-7000 workstation, which had the equivalent of plugins for compression, EQ and so forth.
I never learned to code in 56k assembler, which is what the Optimods used at the time, though I did teach myself DSP, as I’ve mentioned. So I was developing algorithms and prototyping them in modern Fortran, which originally stood for “Formula Translation.” With its computational efficiency, type safety, automatic interface checking and support for vectorization, Fortran is a good language in which to express DSP, to the point that later, when we did our x86-based software audio processor, the 1600PCn, the DSP was done in modern Fortran and the rest was done in C++.
At the time I was prototyping things like the MX limiter and the half-cosine interpolation composite limiter. I had a development system that allowed me to listen and freely make changes, as I would do on the workbench in the analog days. I would give our DSP engineers the Fortran prototype. They would translate it into 56k assembler, and it would go into the products.
McLane: Also in this time we’re starting to see the impact of streaming audio. How did that change the dynamics, if you’ll excuse the expression?
Orban: We had hired Greg Ogonowski as a consultant for the 8200 project. Later he came on as VP of new product development. His goal was to develop the streaming business, and there were two tracks there.
The hardware Opticodec was the German product from Dialog4. Separately, there was a software Opticodec, for streaming, developed by Greg’s group. He was working with a very talented coder, and we supported what is now called HE-AACv2 early on.
On the hardware side we developed Optimod-PC, the first sound cards with onboard Optimod DSP. We had the 1100 and then the 1101, which were PCI and PCIe cards with 56k-based Orban processing, to be used in Windows computers. It was basically the HD processing that we had developed for the 6200 and 6300.
The Opticodec processing software would run on the computer and receive the processed audio from Optimod-PC’s onboard DSP.
McLane: In 2008 you introduced Loudness Meter software for Windows XP and Vista. This was an interesting twist. It showed instantaneous peaks, VU, PPM, CBS Technology Center loudness and ITU BS.1770 loudness.
Orban: As we’ve discussed, we had licensed the loudness controller and loudness meter from CBS Technology Center in the early 1980s. But by now the patents had expired.
We had never put the meter into production because the analog version would have been extremely expensive, using lots and lots of precision parts because there were big high-order filter banks in it. But it occurred to me that Intel and AMD processors for PCs had gotten powerful enough to run a DSP model of the CBS loudness meter.
I created a digital model of that complicated and expensive CBS prototype, of which I think there are only five copies ever made. Fortunately, I had one of the analog units in the lab. I could make sure that the model agreed with what the analog was doing.
At this point, ITU-R BS.1770 loudness had become an international standard, so we had to include that in the meter. True peak had become a hot topic, so we put a true peak meter in it. And we put a VU meter and a PPM peak program meter in to make it more complete.
To make people more aware of our audio processing, particularly in television, we decided that the loudness meter would be a freebie.
McLane: That promotional tactic was effective, I think.
Orban: They’re well known and are available to this day. They’re very accurate. And they are the only software or hardware that offers the CBS Technology Center loudness algorithm in addition to the ubiquitous BS.1770.
I thought that Ben Bauer’s group’s original work at CBS Labs in the 1960s and then Bronwyn Jones’ and Emil Torik’s 1980 update were important and worthwhile, and I feel honored to be able to keep it alive.
We had also used the CBS loudness controller in our DSP-based television processors, because we feel that it gives better results than the BS.1770, particularly in terms of having a correct loudness balance between speech and music.
In fact when I worked with the AES committee that brought forth the TD1008 loudness control for audio streaming white paper — which eventually became AES77 — we recognized for the first time in a standards-based document that you need to put a correction loudness offset in BS.1770 of approximately 2 to 3 LU to get the best balance subjective between speech and music.
McLane: What else do you recall from this very busy stretch in the first decade of the 21st century?
Orban: We had to economize compared to the Harman years, doing more with less. But we managed to keep the Orban standards up and develop important products during that time.
