Analog vs Digital FM Exciters

[speakerman]

Would anyone care to comment on the audible differences (if any) between the best varactor exciters and the new DSP digitally-derived exciters? Do they offer potentially better low end tightness and do they suffer from any audible high-end artifacts when operated stereo only?

I wonder if any of the NPR affiliates have any input on this as they used to be really tuned into it in the past. They have traditionally not run their audio through an electronic Play-Doh extruder http://www.stupid.com/giftshop/playdoh.jpgto get the last dB in loudness.

 

[littlejohn]

Some the digital ones which generate on frequency modulated carrier kick a$$ and take names. Some don't. Probably more important than the modulation scheme is the baseband processing. Which, in an exciter being fed AES3 digital audio, must of necessity be a part of the exciter. Currently Continental leads, BE is close, Harris isn't. This is subject to change as new boxes hit the market. I haven't played with the Naitel, but it gets good press from those who have. The digital ones seem to be a bit less tweaky and more noise immune, which you'd expect.

 

[Kelly]

I really like the BE FXI-60 exciter. Noise is low, and the AES inputs are easy to set up with very little overshoot or nasty byproducts. Just don't bother with the built-in wide band compression nor limiting. Assuming you're using one of the newer digital audio processors and AES into the STL, you won't need it.

 

[Sgeirk]

I dunno. What's the point of a digital exciter REALLY? Anyone have a really good answer to that?

If your plant is all AES, then oh...ok...just avoid the Harris Digit...otherwise, there is no shame in slapping an old Continental 802 on the air...that's a great sounding exciter...and as odd as it sounds...a re-capped Bext...cheap as they are sound pretty good too! ymmv.

 

[konbaasiang]

Littlejohn, I've used a Nautel M50. Flawless performance - whatever your feed either the AES input or the Composite, basically gets put on the air without modification (other than the fact that it gets frequency modulated). No low frequency tilt, no high frequency overshoots, no filtering of the incoming audio signals - and the on-screen modulation meter is spot-on (if a little over-sensitive).

Speakerman, a good DSP-design is likely to be even cleaner than an analog exciter - digital is inherently perfectly linear, and doing direct-to-channel FM modulation in software (as long as it's properly written) essentially means the signal is bulletproof -- any non-linear distortion to the FM-modulated signal would only add noise and interference (which can be filtered out) and would as far as I understand it not affect either frequency response or modulation overshoots.

Sgeirk, reproducability and perfectability! Once the design is done, it won't drift, it won't degrade over time (other than completely failing at some point in time). With analog gear you sometimes get a warning (performance degrades gracefully), sometimes you don't (sudden catastrophic failure), just like digital. Overall, seems like an improvement to me .

Kelly, the BE FXI-60 exciter is excellent - other than the input section which has some serious flaws.
The AES inputs have 16 KHz low pass filters -- even with compression/limiting disabled! Why they implemented this, I have no idea, but it causes overshoots with both Optimod 8500 (some) and Omnia 6 (lots). With 8200 or 8400, there's no overshoots, because they're strictly bandlimited to 16 KHz already - nothing for the filter to remove.

The other major problem with the BE FXI-60 (and this particular thing may be inherent only to the unit I was testing, but worrying nonetheless) is that some high-end from the composite input leaks through into the exciter even when you've selected the AES inputs - causing 10% overshoot! I spent *hours* trying to find out where they were coming from, until I randomly disconnected the BNC cable feeding composite -- and they disappeared.

I have absolutely no clue how a supposedly all-digital exciter could have this problem when you've selected a digital input. Where is the analog signal supposed to leak in? Something doesn't add up.

Ignoring those two problems (using composite input only), the BE FXI-60 performs flawlessly -- no low frequency tilt, no high frequency rolloff, and I measured a noise floor of -70dB using the MPX output of a Belar Wizard, and MpxTool.

///Leif

 

[FFoti1]

The other major problem with the BE FXI-60 (and this particular thing may be inherent only to the unit I was testing, but worrying nonetheless) is that some high-end from the composite input leaks through into the exciter even when you've selected the AES inputs - causing 10% overshoot! I spent *hours* trying to find out where they were coming from, until I randomly disconnected the BNC cable feeding composite -- and they disappeared.

I have absolutely no clue how a supposedly all-digital exciter could have this problem when you've selected a digital input. Where is the analog signal supposed to leak in? Something doesn't add up.

Leif,

The problem occurs due to mixing the MPX/BNC sampled and AES signals inside the exciter. I had this problem with the digit too. Aparently, even though AES is selected, the sampled MPX signal appears to be 'filtered' off as compared to shut off, and leakage happens. Not sure why this would be designed/configured like this.

-Frank Foti

 

[Sgeirk]

konbaasiang: I've never been able to detect any sonic improvement....not enough to justify the price tags to my ears and my boss.

 

[littlejohn]

The old 802D sounds as good now as when we first put it in. The 802 analog wanted twesking and component replacement (caps) in the same period to maintain performance. Also, as I understand it (p'raps Foti and Orban will chime in here) if I iterate a filter in software, the components are exact, and thus the generated nasties from that filter are exact and can be better taken care of than in an analog design where component drift and value comes into play.
My current concern is the FlexStar exciter, whose analog 19KHz subcarrier has substantially more jumping around with modulation than I like. I took the monitor and compared the 802D subcarrier, it's substantially less frisky, so I don' think the monitor is at fault. Any of the gurus present have thoughts on what might be going on? Monitor is the whizbang Audemat box with the analyzer option, backed by an FMM-2/FMS-2.

 

[FFoti1]

The old 802D sounds as good now as when we first put it in. The 802 analog wanted twesking and component replacement (caps) in the same period to maintain performance. Also, as I understand it (p'raps Foti and Orban will chime in here) if I iterate a filter in software, the components are exact, and thus the generated nasties from that filter are exact and can be better taken care of than in an analog design where component drift and value comes into play.

Yes, in the digital domain, filters do not change or drift due to component aging. The gotcha becomes how does the filter behaves under non-linear conditions? FIR filters, by design, are linear but they can overshoot depending on the design characteristics of the filter. This was the achilles heel of the digit exciter. Poorly designed filters, and they do overshoot.

Digital filters come very close to theoretical performance...BUT...Unless all of the filters parameters are understood with regards to behavior with all signal content, anomalies can occur, just as their analog counterparts.

-Frank Foti

 

[speakerman]

It has been a while since this topic was started. I'm wondering if if there is new information to share about digital FM exciters and new low power integrated one-piece transmitters (2.5k or less). Are we reaching the price point now with digital electronics that direct-digital not only sounds better but also is not as much a hit on the budget as it was? Do any of the integrated exciter/amplifier packages offer true direct to carrier digital signal generation/modulation like you get using a standalone digital exciter?

How has the reliability been so far for the early adopters? I have been burned in the past by jumping on the latest technology only to find out there were software bugs that needed to be squashed by a re-flash to fix or the design turned out to be lacking due to chip compromises to save cost.

I see Nautel now offers Livewire direct connection. What would be the advantage if your audio processor is located at the transmitter? I'm not aware of anybody building an audio processor with Livewire output yet. That would make me also wonder if there is any advantage using the stereo generator part of the exciter when most of the DSP audio processors already put out pristine composite.

 

[dspxscott]

Leif, et al.

Some of the points are not strictly true. Some digital exciters have to use analog filters to remove sideband and other spurious responses from the DDS circuitry. These analog filters can create phase shifts that negate the numerical accuracy of the DDS.
Most competent DDS exciters won't suffer but I know some that do. I've even seen a few that don't need CE harmonised standard emission masks, although on sale openly.

As Bob Orban often says, "the devil is in the details"

 

[gtomas]

I see Nautel now offers Livewire direct connection. What would be the advantage if your audio processor is located at the transmitter? I'm not aware of anybody building an audio processor with Livewire output yet.

Omnia One and Omnia 11 both have Livewire inputs and outputs.

That would make me also wonder if there is any advantage using the stereo generator part of the exciter when most of the DSP audio processors already put out pristine composite.

A good question. For analog FM, I don't see any advantage. Actually, there are potential disadvantages unless the stereo generator in the transmitter is complementary to the processor used.


Regards,
Goran Tomas

 

[dspxscott]

Having pristine MPX analog outputs from a processor is great but that is not enough if the DDS exciters analog mpx input is poorly designed. Sample rate convertors, undersampled A/Ds, poorly designed analog buffering and poorly designed digital audio to on channel DDS conversion can introduce a whole range of performance affecting issues that reduce the performance of the audio processor's 'pristine' output and the 'apparent' textbook performance of the DDS modulator.

Mr Foti once proposed a digital mpx format and that would be the way to go, although i have not seen it implemented commercially.

IMHO, a digital audio output from a processor to a stereo generator in the TX with in built phase locked SCA/RDS subcarriers would offer superior performance, but only when composite processing is not employed. The processor manufacturers have invested significantly more R&D into composite processing research than the TX manufacturers and the benefits obtained probably do mitigate the noise floor improvements and other small advantages obtained from a direct digital audio path between the processor and the TX.

In real world terms, with analog FM, there isn't much noticeable or measurable difference between a digital or analog composite connection between the processor and the TX. On the assumption that some form of composite processing (even light overshoot protection) is employed, it's probably worthwhile using the processors analog MPX output. With the recent introduction of digital transmitters with in-built competent processors this may not always be the case.

 

[TomT]

I'm wary of digital for digital's sake. Adds complexity. Difficult to fix in the field. In the end, I'm not convinced there is enough difference to be noticeable--especially compared to a good analog exciter such as the Continental 802B. Nor do I see any particular advantage to going to an all digital plant over a modern analog plant with judiciously selected digital components. For the most part the current generation of digital processors. Even then, I prefer using the analog composite output to an analog exciter. One less digital conversion.

However, both the new BE and Nautel low power transmitters have built in exciters. The Nautel can be patched around, the BE, apparently not so easily.

 

[speakerman]

So...this makes me wonder if there is really an advantage to a DDS type exciter as compared to a really good varactor analog one. The ultimate goal is to have the best dynamic transfer function that really sounds good to the ears. With the amount of dynamic range reduction everyone uses to get the average R.M.S. level up, ultimate signal to noise ratio seems to be less important in choosing an exciter. I have looked at varactor type schematics and noticed some use piece-wise linear-approximation distortion cancellation circuits to pre-distort the non-linear transfer function of the modulator (with measurable success). How clean does it really need to be and in the end and does it really mater since there are so many different detector transfer functions in the receivers out there?

Can anybody say that the DDS designs definitely sound better than the varactor/oscillator types? I know from the work I did in the 70s the introduction of ICs to audio was not an audible improvement. It took a while before we realized there was much more to what the ear detected as distortion than harmonic and intermod. We later discovered (which I suspected for quite a while) SID and TIM was a real design problem with early OP-amp designs that made them sound harsh. Faster slew rates and better filter designs solved those problems.

I will be quick to say I love Livewire and find it to sound like a wire connected from point A to point B. They are using a high enough sampling that latency is really low and it is apparently easy to implement filters that have no audible affect to the audio.

 

[TomT]

A well-maintained analog FM system--into a good quality receiver with a strong multipath-free signal, should be capable of at least 65 db signal to noise ratio.

But...

Remember that most listening is done in the car nowadays. Go look for a table radio at Wally World or K Mart. Even for a pure classical format, you need to restrict dynamic range to around 20 db, otherwise the low passages vanish in the ambient noise (e.g. wind, road noise) that exists in such a listening environment.