Direct to channel digital exciter recommendation

[rfry]

TLC, I don't want to re-hash ten year old topics, but your post has stimulated a few questions that Harris should answer. It would appear that you work for Harris. ... True, widening the filter helped some, but the peak control is nowhere near the same as the composite input. etc

The original product releases of the Harris Digit and Digit CD were designed for, and accommodated AES-3 inputs with AF bandwidths that produced a very high-quality, analog FM transmission system.

Digital sample rates greater than 32 kHz (16 kHz AF bandwidth per stereo channel) were downconverted to 32 kHz in the Digital Input Module of those exciters.

The big "incompatibility" problem with some FM broadcast audio processors having AES-3 outputs with sample rates greater than 32 kHz was that they did not limit their L&R audio spectra to an upper bound of ~ 16 kHz, including any processing artifacts above 16 kHz -- which together produced modulation overshoots in the Digit/Digit CD.

 

[fm-engineer]

Ok Richard, when the Digit was designed, there was only one digital processor, the Optimod 8200. Let's just agree for sake that it was not the most popular sounding box, but it lead to what we all enjoy now. In 1997 along comes the Omnia FM. The Optimod was strictly band limited to 15 kHz, as was later Orban products 2300, 5300, 8300, 8400. The Omnia FM, and later the Omnia 6, were not strictly limited to 15 kHz. The filters were designed based upon the "sound." It was not a problem until you had a device that down sampled to 32 kHz. Harris would never admit they were wrong since the Optimods of the time worked fine. Others chose the sound of the Omnia group, but were not able to implement a complete AES chain due to the 32 kHz issue. At the same time, the top STL manufacturer (Moseley) began to offer 44.1 and 48 kHz linear StarLink STL’s to support all options demanded by the consumer.

To date, the Digit CD still operates at 32 kHz, even after all of these years. U10 filter IC was a slight improvement, but as a fellow engineer you know this does not solve the Nyquist folding issues at anything above 16 kHz. In 2005, Orban began to produce the Optimod 8500, and now the 8600. Guess what, they are no longer strictly band-limited to 15 kHz either. Now it is more gradual up to 16.5 kHz, which at times will overshoot the ole Digit CD. The Omnia's may be a bit wider, but it was based upon the sound the units can deliver. I’m trying to stir up old bones, but Harris did not listen to the engineers in the field, nor did they listen to some of the audio DSP experts. The Digit CD design could have been changed over that long decade and a half. What was the product run, 15 years? The Digit is still listed on the site for sale?? This is (was) just one reason why the old blue and white is not even in my top 3 anymore.

TLC, I don't want to re-hash ten year old topics, but your post has stimulated a few questions that Harris should answer. It would appear that you work for Harris. ... True, widening the filter helped some, but the peak control is nowhere near the same as the composite input. etc

The original product releases of the Harris Digit and Digit CD were designed for, and accommodated AES-3 inputs with AF bandwidths that produced a very high-quality, analog FM transmission system.

Digital sample rates greater than 32 kHz (16 kHz AF bandwidth per stereo channel) were downconverted to 32 kHz in the Digital Input Module of those exciters.

The big "incompatibility" problem with some FM broadcast audio processors having AES-3 outputs with sample rates greater than 32 kHz was that they did not limit their L&R audio spectra to an upper bound of ~ 16 kHz, including any processing artifacts above 16 kHz -- which together produced modulation overshoots in the Digit/Digit CD.

 

[rfry]

The Omnia FM, and later the Omnia 6, were not strictly limited to 15 kHz. The filters were designed based upon the "sound."

It is understood that "sound" is a very subjective topic, but how much of that sound is produced by audio content above 15-16 kHz -- that not too many adults can hear very well, and not many analog FM receivers can reproduce very well?

A Digit/Digit CD using an AES-3 program source even with a 48 kHz sample rate has no "sound" of its own, and extremely low modulation overshoot as long as the audio bandwidth does not exceed ~ 15.5 kHz.

 

[littlejohn]

As tlc points out, the initial Digit stereo composite limiter (may we say 'clipper' here?) was not designed to enhance the signal, and would in fact do Very Bad Things to it. The Continental produict, OTOH, was designed with enhancement in mind, and would in fact to a credible job of loudness before it became unhinged. It still does. Tlc and I disagree on the relative merits of the two boxes. In the instances where I've had both, the Digit was quickly relegated to the backup chain, and was >always< fed composite. It simply ain't got the poon in the digital input section... and the initial digital input was in a funky plastic box hung on the back of the thing - hardly confidence inspiring. The 802D and D1 still put out a competitive sound.

Perhaps someone can confirm or deny the veracity of what I was told about the Digit. The story was, the digital clipper and processing code in the box was purchased from an outside vendor. Continental's, OTOH, was done in-house with the idea in mind of platying with the rest of the unit as a whole. Like many stories, it's probably 'sort of' correct.

 

[fm-engineer]

It was not the idea to try to hear anything above 15 or 16 kHz, but the shape of the low pass filter. Steep brick wall filters created anomalies in the digital domain that would affect the entire audible spectrum. Again, the two major FM processor companies do not filter to 15 kHz anymore. Now common LPF are out to the 16-17 kHz range.

The Omnia FM, and later the Omnia 6, were not strictly limited to 15 kHz. The filters were designed based upon the "sound."

It is understood that "sound" is a very subjective topic, but how much of that sound is produced by audio content above 15-16 kHz -- that not too many adults can hear very well, and not many analog FM receivers can reproduce very well?

A Digit/Digit CD using an AES-3 program source even with a 48 kHz sample rate has no "sound" of its own, and extremely low modulation overshoot as long as the audio bandwidth does not exceed ~ 15.5 kHz.

 

[fm-engineer]

From memory, the Digit used the "Waves" limiter.

As tlc points out, the initial Digit stereo composite limiter (may we say 'clipper' here?) was not designed to enhance the signal, and would in fact do Very Bad Things to it. The Continental produict, OTOH, was designed with enhancement in mind, and would in fact to a credible job of loudness before it became unhinged. It still does. Tlc and I disagree on the relative merits of the two boxes. In the instances where I've had both, the Digit was quickly relegated to the backup chain, and was >always< fed composite. It simply ain't got the poon in the digital input section... and the initial digital input was in a funky plastic box hung on the back of the thing - hardly confidence inspiring. The 802D and D1 still put out a competitive sound.

Perhaps someone can confirm or deny the veracity of what I was told about the Digit. The story was, the digital clipper and processing code in the box was purchased from an outside vendor. Continental's, OTOH, was done in-house with the idea in mind of platying with the rest of the unit as a whole. Like many stories, it's probably 'sort of' correct.

 

[littlejohn]

Yeh, Waves. That's what I had heard. I guess if two of us heard the same, makes it fact, no? It was underwhelming in that service at any rate.

Interestingly,somebody told me you could drive the composite module on a Digit with a Breakaway and diddle the linearity adjustment, and eventually get frighteningly good sound out of it. About 2015, when I get a free minute, I'm gonna try.

 

[jesseg]

DDS excitation directly to channel is superior to everything previous though, and may not ever be improved upon. A technology quite similar to how SACD/DSD/DXD works... except instead of audio, you get 88-108 mHz *directly* from digital. Its accuracy is so high that I haven't seen any measurement equipment capable of measuring its distortion. 8) But that would still be cool to see some day.

It's right around the corner too.

 

[rfry]

DDS excitation directly to channel is superior to everything previous though, and may not ever be improved upon. ... Its accuracy is so high that I haven't seen any measurement equipment capable of measuring its distortion. But that would still be cool to see some day.

Maybe so, but depending on the quality of the program source it was fed, the measured L&R audio channel performance demodulated from an analog r-f transmission line sample produced by the original Harris DIGIT FM exciter of 15+ years ago already was/is well below the threshold of human audibility.

 

[jesseg]

That is true, but when transparency is the name of the game, the required 15kHz limitation DOES make an audible difference to many people. Not just a random golden ear. A whole series of studies have even shown that despite us not being able to hear specific pitches past ~20kHz (some more, some less), that sound can still effect how we hear things up to 100kHz.

Experiments were conducted to assess the human discriminability of temporal convolution. The experiments employed either lowpass filtering or delays due to spatial misalignment. By using special ultrahigh-fidelity equipment, both experiments demonstrated discernment at a ~ 5 microsecond timescale, which is much shorter than found previously. While the signal manipulations affect both the spectrum and temporal definition of the signal, the spectral changes fall below the known just noticeable differences. The discrimination may therefore involve mechanisms additional to the auditory system’s ability to distinguish spectral amplitude differences. Furthermore the present work shows that typical instrumentation used in psychoacoustic research may, for some purposes, have insufficient temporal speed and bandwidth. Also this work proves that that digital sampling rates used in consumer audio are insufficient for fully preserving transparency.

So... what is the core sampling rate of the Flexstar?

 

[rrsounds]

If any among you were to do a laboratory test to determine human hearing extremes, a task that has been repeated many times over the last eighty or so years always with similar results, your finding might possibly be that, absent all other sounds, some humans may on some occasions in some environments hear beyond the typical human hearing envelope. (see http://www.extron.com/technology/img/loudnesscontrol_ts_1-lg.jpg)

Since these extreme signals are on the very verge of sensation even when isolated 'solo' from all other sounds and even when under laboratory conditions, in real programming where much stronger audio in much more ear-sensitive ranges dominates and psycho-acoustically masks the extremes by many tens of decibels, I doubt any supersonic or infrasonic material makes much difference.

But, as they say, knock yourself out!! LOL!

Kind Regards,
David Reaves

 

[fm-engineer]

Well, since Harris never answered the question, the FlexStar appears to have a base sample rate of 48 kHz. For whatever reason, I also have station that is looking at one for an HD upgrade, and to replace a damaged Digit (trying to talk them out of it)! I read their online manual and gathered that the analog FM side has a 48 kHz base SRC chip when using the AES input.

That is true, but when transparency is the name of the game, the required 15kHz limitation DOES make an audible difference to many people. Not just a random golden ear. A whole series of studies have even shown that despite us not being able to hear specific pitches past ~20kHz (some more, some less), that sound can still effect how we hear things up to 100kHz.

Experiments were conducted to assess the human discriminability of temporal convolution. The experiments employed either lowpass filtering or delays due to spatial misalignment. By using special ultrahigh-fidelity equipment, both experiments demonstrated discernment at a ~ 5 microsecond timescale, which is much shorter than found previously. While the signal manipulations affect both the spectrum and temporal definition of the signal, the spectral changes fall below the known just noticeable differences. The discrimination may therefore involve mechanisms additional to the auditory system’s ability to distinguish spectral amplitude differences. Furthermore the present work shows that typical instrumentation used in psychoacoustic research may, for some purposes, have insufficient temporal speed and bandwidth. Also this work proves that that digital sampling rates used in consumer audio are insufficient for fully preserving transparency.

So... what is the core sampling rate of the Flexstar?