Multi-band Audio Processor Users - What freq band divisions do YOU use?

[K1JJ]

If you have a multi-band audio processor, or the knowledge of using one, I would be curious how you recommend I divide up the various audio bands.  

For example, let's say it's a 6-band processor and you have a choice of setting band #1  from say, 40hz-200hz, band #2 from 200 hz to 800hz....... to band #6  4kc to 10KC....  or whatever.

And why these choices?

I'm using a Behringer 6-band 9024 and set the bands by seat-of-the-pants guesswork with no rhyme or reason, so wondered if there is a proper way, like using octaves or something else.

BTW, my 31 band graphic EQ goes into the 6-band processor and is  +3 db at 50hz, drops down to -8db at 125 hz, stays at -8db until 800 hz, then gradually rises up to +6db at 3kc to 5kc. It then drops off quickly to -12db at 6-7kc and beyond.

Thanks.

Tom, K1JJ

 

[KC2IFR]

Tom,
That is a loaded question. U will have to decide yourself how to set up the different frequencies. If everyone had to use the same setup, they wouldnt make the parameters variable!
There are starting points.....for example the 200-250 htz range is where most of the mud frequencys are. If u can, put the rig into a dummy load and then record yourself using different settings. Setting up an audio chain can be daunting but its a good learning experience! The other way  to do it is on the air having someone record u and then send the recording back to u.

Bill 

[N2DTS]

6 bands seems like way to many for voice on the ham bands.
Say you set it up to do 20 to 6000 Hz, that is a lot of bands over 6000 Hz!

If you squeeze 6 bands in (every 1000 Hz) you are doing to have a VERY dense signal!

Brett

[KD6VXI]

If you have a multi-band audio processor, or the knowledge of using one, I would be curious how you recommend I divide up the various audio bands.  

Thanks.
Tom, K1JJ

Tom,

Best way I've found is to watch your voice, so to speak, on a spec-an or Flex type display.  Then you can dial it up for the individual way your voice responds / the iron in your TX, etc.  Backing that up with 'on air' or mod monitor testing is necessary, as well.

Mine is virtual, but I run a 4 band.  Centered on 103 hz, 1084 hz and 3418 hz.  Those are the cutoff freq's, btw.  Mine doesn't have a 'center freq'.....  The parametric EQ does, tho.  I'm sure it's a nomenclature thing.


--Shane

[KC2IFR]

Tom,
Please checkout this web page. It is one of the best Ive ever seen on the web that the average person can understand. They cover many topics that also relate to ham radio.
Let me know what u think!
http://www.rane.com/library.html

Bill

[KC2IFR]

BTW.....
This is the Rane EQ I use with my Johnson 500.
http://www.rane.com/peq55.html

Bill

[K1JJ]

Thanks for the suggestions, guys.

I read some of the split-band processor material, Bill. There certainly are infinite setting combinations.

It said:
"You can decide to process different ranges of an instrument [vocals] differently. You could use no compression on the low end of a bass, with heavy compression on the top end to put the string slaps in balance with the bottom. Or you could tighten a boomy bottom with compression but leave the top less controlled for an open feeling."


They mention compressing the deep low end by a 10:1 ratio with fast attack, and leaving the highs lightly compressed - and doing the opposite for other instruments.

So bottom line there are an infinite amount of settings to experiment with.  But for vocals, I was hoping someone who has already been thru this experimentation would suggest something like, "I compress the 30-125 range at 3:1 ratio with fast attack... the 125-500 gets 6:1, etc.  A starting point would be good to know with reasons why they feel certain bands need emphasis on this or that. But I figure multi-band processing is used by a few, so not a lot of real-whirl info around yet.

I'll play around with settings and recordings tmw if I don't get any specific suggestions by then.  I'll post my final results that sound the best to me as a starting point for others.

T

[Rob K2CU]

HI Tom,

most say that octave slices are best. if hard clipping is used, then there is a lot of odd order harmonic products produced. sharp cut off low pass filters after each bandpass/clipper should do well. you would presumably require 7 bands, but might squeeze it into 6. starting at 40 Hz, you would have:
40 - 80, 80 - 160, 160 - 320, 320 - 640, 1280 - 2560, and 2560 to 5120.
besides the bandpass filters, you will need sharp cut off low pass filters just above the top end of each.

Or,....consider an RF clipper. Procur a pair of 6 KHZ AM IF filters.  They can be cheap ceramnic types, just need stop band of say 50 dB or more. You generate DSB at  a frequency just off the bottom edge of the filters, amplify and clip the resulting wideband SSB signal to the max, pass the result through the second filter, and then demod it back to baseband audio. You could use a pair of 602's of 1496's or whatever for the balanced modulators, both driven by the same L.O.  There is minimal requirements on pre and post audio bandpass filtering as the filters do that. The deal is that all the harmonic products or intermod baseband audio products from the clipping don't get past the second filter.

look at Murata AM ceramic filter part number CFWLB455KHFA-B0 at Mouser for $4.25 each.


just an idea.

[W8IXY]

I have experimented with crossover frequency settings in multiband processors for many years, including in designing Omnia audio processors.  I have found that the settings of a 4 band processor that please my ears are (all numbers are in 'cycles per second'):  <130, 130-800, 800-3200, >3200.  Much of the "nasality" builds up from a peak about 1600, so I centered the 2 octave band at 1600.  This works for speech that is sharply low pass filtered at 4500 as well as for music that has components to beyond 15000.  I found using an 18db/octave slope works well.  FYI, the old Dorrough DAP310 used slopes <6db/octave. 

Using 3 bands, my favorite was: <130, 130-800, >800 for AM work.  Using 5 bands, I liked: <130, 130-800, 800-3200, 3200-7500, >7500 (unless you are using extremely wideband audio, thats not too practical for AM).  And using 6 bands: <120, 120-360, 360-800, 800-3200, 3200-6400, >6400.

If you use pre-emphasis (recommended), I found the greatest increase in intelligibility was by using a shelving high frequency equalizer @ 6db/octave using the standard 75 microsecond time constant up to 4500.  That gives you about a 6 - 8 db rise starting just below 1000 and levels off around 4500, but doesn't keep rising to +18db @ 15000.

I'd be interested in hearing from anyone who tries those crossover frequencies, different slopes (for different folks), and what, if any, pre-emphasis is used, as well as results from using other crossover frequencies as well. 

Saturday, I heard some very sibilant audio from a couple of stations that sounded like they were using some extreme high frequency boost, and were driving the audio into clipping and distortion that also affected much of their mid-band audio, as well as causing splatter 20 kHz from the carrier. 


73
Ted  W8IXY

[K1JJ]

HI Tom,

most say that octave slices are best. if hard clipping is used, then there is a lot of odd order harmonic products produced. sharp cut off low pass filters after each bandpass/clipper should do well. you would presumably require 7 bands, but might squeeze it into 6. starting at 40 Hz, you would have:
40 - 80, 80 - 160, 160 - 320, 320 - 640, 1280 - 2560, and 2560 to 5120.
besides the bandpass filters, you will need sharp cut off low pass filters just above the top end of each.

Thanks Rob -

Yes, that's what I was looking for. I will try these bands as a starting point. I don't know what kind of LP filters they have programmed in-between the bands, but there sure is a lot going on in there being a recent digital design.

Can you suggest compression ratios and rough attack and decay times starting points for these various bands? Or a general philosphy of why the highs/mids or lows should be compressed more than others in a relative way?

I'm looking for dense audio (that is adjustable on the fly with a knob) and the big boss-jock sound, caw mawn...
T

[K1JJ]

And using 6 bands: <120, 120-360, 360-800, 800-3200, 3200-6400, >6400.

If you use pre-emphasis (recommended), I found the greatest increase in intelligibility was by using a shelving high frequency equalizer @ 6db/octave using the standard 75 microsecond time constant up to 4500.  That gives you about a 6 - 8 db rise starting just below 1000 and levels off around 4500...

Very interesting, Ted.  Thanks for adding this.

You mentioned 75 us time constant. How does this relate to the attack and decay time settings in the processor? Could you suggest a starting attack and decay time for each band you listed above? Also, what compression ratios (3:1, etc) would you suggest for each band as a starting point?


You said: "Much of the "nasality" builds up from a peak about 1600, so I centered the 2 octave band at 1600."

Could you elaborate on the statement above - and to why centering the band will help this situation? How would this same idea apply to getting rid of the muddy mid-bass many of us have in our voices between ~150-800hz?

I'll admit I am at the edge of my understanding of these multi-band processor concepts, but will learn fast...  

T

[w1vtp]

Tom

This is good stuff on EQ.  Here is a representation of what I used to do on my analog system.  The EQ has 27 bands and I like using the term "lazy S" for where to set things.  Bill's comment about the "mud frequencies" is commented on a lot by AMers.  Because of my age-related hearing degradation above 6 KC, I like a bright "sparkly" emphasis.  Listen some time to WBZ and how they do things.  In my view they go too far in the presence settings.  It gets on my nerves after a while but it does have a lot of punch.

Anyway, for what it's worth. Here's what I used to do on the FT-301.  I like the idea of either recording yourself using a broad band detection system (not a receiver with an IF filter that's going to color what you hear).  Or better yet -- have someone record you and email it to you.

Al

[N2DTS]

Mind if I ask why you pass out past 16 KHz on the eq?
The eq also has a low pass filter of 8KHz it looks like, but its not on?

Brett

[KD6VXI]

Tom,

Here's the settings on mine, YMMV.  I use an Astatic D-104 (Silver Eagle model) as the microphone.  It's fed straight into the sound card, and then through the processor and back out via another 1/8 inch jack.

Band 1
Threshold = 0dB
Gain = 13dB
Ratio = 2:1
Attack = 10 ms
Release = 100 ms
Cutoff = 103 hz

Band 2
Threshold = -21.6
Gain = 0.0
Ratio = 6:1
Attack = 10 ms
Release = 100 ms
Cutoff = 1084

Band 3
Threshold = -16.3
Gain = 0
Ratio = 2:1
Attack = 10 ms
Release = 100 ms
Cutoff = 3418

Band 4
Threshold = 0
Gain = 9
Ratio = 2:1
Attack = 10ms
Release = 100ms
Cutoff = >30khz


Hope it's of some use...


--Shane

[W8IXY]

Very interesting, Ted.  Thanks for adding this.

You mentioned 75 us time constant. How does this relate to the attack and decay time settings in the processor? Could you suggest a starting attack and decay time for each band you listed above? Also, what compression ratios (3:1, etc) would you suggest for each band as a starting point?


You said: "Much of the "nasality" builds up from a peak about 1600, so I centered the 2 octave band at 1600."

Could you elaborate on the statement above - and to why centering the band will help this situation? How would this same idea apply to getting rid of the muddy mid-bass many of us have in our voices between ~150-800hz?

I'll admit I am at the edge of my understanding of these multi-band processor concepts, but will learn fast...  Wink

Elaborating on a couple of things:

The 75 us time constant is the same used in FM broadcast transmission.  However, adding additional pre-emphasis beyond about 5kHz will make sideband splatter VERY pronounced on our ham bands.  Even the AM broadcasters don't sound all that good if they continue to use a rising characteristic beyond 5 kHz.  Thats why I prefer a shelving curve after 5 kHz or so.

Pre-emphasis and time constants as well as compression ratios are all different things, but knowing how to set those characteristics and combine them into a system is how the "magic" works for stations to have great audio.  I have found a 2.5:1 or 3:1 wideband AGC is the best sounding first step in processing.  A brick wall 100:1 AGC causes a "congested" sound at that point in the processing.  Check out the "over easy" compression scheme that DBX used (and other components as well) that results in a "curve" of input vs output levels, that compresses at a higher ratio when input level is higher.

I had been experimenting on both the ham AM and broadcast AM some years ago and wondering how I could minimize the "nasality" of the station's audio, and was trying some band limited pink noise through the audio chain.  I swept the frequency up and down and, by ear, heard the harshness of the pink noise on a typical narrowband consumer AM radio, accentuated around 1600 Hz.  I decided to try a crossover that placed the two octaves centered around 1600 Hz in one of the processing bands.  I discovered that using that bandpass greatly diminished any "nasality" in the programming, voice and music.  If the EQ of the source material accentuated the 2 octaves ceneterd around 1600 Hz, the dynamic EQ from the multiband AGC flattened it out.  After all, a multiband AGC can be considered as somewhat like a dynamic tone control.

Depending on the voice of the old man at the mike, if you can look at a spectrum analysis of your voice's characteristics (that characteristic word again...) try centering the passband of the AGC section, one to two octaves wide on those frequencies, to dynamically control the muddiness, and increase punch and intelligibility.

I also preferred about a 3:1 compression curve on the multiband AGC sections.  I have also found that using about a 3:1 ratio on the lowest and highest bands, and a 2:1 or 2.5:1 ratio on the middle bands resulted in a nice enhancement and minimal "congestion".  AGC usually is run with relatively moderate attack and slow release times.  Depending on how dense (or compressed) you want to sound, you can set the attack and release times accordingly.  For a starting point for AGC attack times, try a 100 millisecond attack time, and a 2 second release time, for the lowest bands, and up to a 10 millisecond attack time and 600 millisecond release time for the highest bands.

Then try about a 2 millisecond attack time for the limiter, and about a 125 millisecond release time.  The limiter ratio should be at least 10:1.

Also, don't dismiss a nice controlled amount of clipping (1-4db), followed by a flat bandpass rolloff filter, to add the final "average to peak" ratio reduction, and be the ultimate brick wall to the audio applied to the modulator.  Remember, any response anomalies after the final limiter/clipper will result in loss of loudness and ultimate modulation percentage.

These are just starting points.  If you have a fairly complex software program, such as Adobe Audition, you can play around with AGC, compression ratios, limiting ratios, and cascading all those effects to see what it will sound like.  You can dicker around with each section of the processing and hear the final effect when you combine all of them.  Then you can build the "ultimate processor" for your station based on what you found worked for you in software.

When I adjust for the best compromise between quality, intelligibility, punch, and loudness, I first figure out what each stage (AGC, EQ, Limiting, clipping, etc.) that I want to incorporate sounds like, then "line 'em up", and crank the gain processing up until the loudness no longer is increasing, and the tweak each setting to make the final adjustment do what I want it to sound like.  Setting everything to "11" usually results in something sounding horrible. (Think cascaded power mikes on CB!)

Then, getting deeper into the subject, you can consider an "all pass filter" that removes the asymmetry from our voices.   Thats for another discussion.

So, there is some elaboration as to how we construct and adjust our processors at Omnia.  Hope it gives anyone interested some starting points on keeping an ever improving AM presence on all of our bands!

73
Ted W8IXY

[W2XR]

Hi Tom,

I just thought I would chime in with a potential issue pertaining to EQing an AM transmitter. Very probably this is known to many if not most on this forum, but for those who may not be aware; when you change the EQ boost or cut in a given frequency range or band, you also shift the phase of the signal in that band.

This deserves some consideration, as it can have a significant impact on the asymetrical modulation characteristics of the transmitter. I suspect the effect is more profound in plate-modulated transmitters, or those rigs where there are a number of audio xfmrs or complex impedances in general in the signal path.

In my rig, from microphone input to modulation xfmr output, the audio takes a trip through five (5) transformers. Each of these has an impact to the audio phase, and the cummulative phase shift of so many xfmrs in cascade in the signal path may be significant.

At any rate, I have found in my rig at least, that the greatest impact to the relationship between positive and negative-going modulation peaks from EQ boost is in the region below about 125 hz. If I run the EQ settings flat on my equalizer in this range, the transmitter can easily modulate in excess of 110% on nearly every syllable, with peaks of frequent reoccurrence hitting 125% or greater, without the negative-going peaks ever hitting the baseline. However, once I start increasing the boost below about 100 hz, the positive-going peaks are reduced in amplitude, and the negative-going peaks start increasing.

If I run the EQ flat, (or the EQ in bypass mode), the transmitter will consistently hit positive mod peaks in excess of 133% (the limit of my modulation monitor), but I don't like the sound of the rig in this case. I can easily bury the needle of the positive modulation meter.

I encountered this a number of years ago when I re-EQ'd the rig after adding the NFB loop around the modulator and driver stages. It resurfaced this past Saturday, when you and Tron suggested I boost the spectrum from 40 to 100 hz about 3-4 dB to add the proverbial 3rd BA. The extreme low-end came up alright, but the positive-going modulation peaks only rarely hit 125% at this point. The difference in perceived loudness is minimal, and the trade-off is of course the increase in deep bass weight.

I look forward to replacing the existing audio driver with the forthcoming WA1QIX FET driver circuit and seeing how this impacts the asymetrical modulation characteristics of my rig vs. low-end EQ boost. Doing so will eliminate two xfmrs; the audio input/phase-splitting xfmr at the input to the driver, and the always problematic audio driver xfmr. Perhaps I will then be able to have my cake and eat it too, in terms of positive-going peaks of high amplitude with less impact to the magnitude of these peaks from the EQ settings.

Everyone's transmitter is different, of course; as such, your mileage in this regard may vary.

73,

Bruce

[WA1GFZ]

I ran a signal generator through an analog EQ once and saw the same phase distortion Bruce. Imagine what you could do with a DSP to make up for transformer phase issues. With the right software...

[w1vtp]

Mind if I ask why you pass out past 16 KHz on the eq?
The eq also has a low pass filter of 8KHz it looks like, but its not on?

Brett

Didn't really matter Brett.  I was stuffing things into a Yaesu FT-301.

Al

[K5UJ]

<<<It resurfaced this past Saturday when you and Tron suggested I boost the spectrum from 40 to 100 hz about 3-4 dB to add the proverbial 3rd BA.>>>

What is the "proverbial 3rd BA?"

[W2XR]

<<<It resurfaced this past Saturday when you and Tron suggested I boost the spectrum from 40 to 100 hz about 3-4 dB to add the proverbial 3rd BA.>>>

What is the "proverbial 3rd BA?"

Hi Rob,

Let's just say it injected some additional testosterone into my audio.

73,

Bruce

[K1JJ]

Thanks for the additional info, everyone!

Terry - That's a lot of good info to digest. I plan to dive into my processor probably tmw and make a new saved set of parameters using your suggestions. Thanks for the limiter parameters too. There is a limiter in the 9024 that will accept your settings.

For now, I'll just hook the headphones directly to the audio amp and see what I can accomplish - then later do some off-air listening.

Bruce - Yes, I also see the polarity (phase) effects of boosting the lows. I'm cornvinced when the TX can pass deep lows flawlessly, (like a balanced modulator or Class E TX) then the correct polarized voice makes shark fins of the lows. But when the lows are not perfect, then the best polarity is the opposite of shark fins. Right now my FT-1000D with bal mod WILL do shark fins with highest positive peaks, whereas, the 4X1 rig must be run opposite polarity. I'm hoping when the forthcoming WA1QIX audio driver is available, this will improve the 4X1 deep lows transparency. We'll have to see. The mod iron may be the bottleneck, however, even though it is quality 1KW BC iron..


My present processor settings are probably "OK" as-is, but might as well play around and do some fine tuning to make it even better.  These things usually improve once understood and intelligent choices are made...

I'll post the results.

Thanks again for the background info and starting points, guys.

T

[K5UJ]

<<<It resurfaced this past Saturday when you and Tron suggested I boost the spectrum from 40 to 100 hz about 3-4 dB to add the proverbial 3rd BA.>>>

What is the "proverbial 3rd BA?"

Hi Rob,

Let's just say it injected some additional testosterone into my audio.

73,

Bruce

Ah, that's what I thought.  Okay.  I eventually figured out that for me eqs work best for inserting a low and high cut and making up for attenuated or boosted ranges of frequencies in the audio chain due to weird mic responses etc.   I deliver to the compressor and limiter a flat sig. from around 300 hz up to 4 KHz (5 on an open band) then a drop off of > = 30 dB at 6 KHz.  Going the other way there's about 10 dB drop from 300 to 100 Hz then down to nothing below 80 Hz.    I used to want that deep late night DJ voice but I just don't have the pipes -- if you got 'em you may as well use 'em    But anyway, back to multi-band compressors...

Rob


I decided when hams try to use them to alter the voices they got from mom and dad they can wind up being worse off than if they just ran stock audio with an old D104.  Back when I was mainly SSB operating and messing around with audio gear I heard hams with so much bass boosted they sounded kind of nauseating.   Anyway, before we digress too much back to multi-band compressors.  

[K1JJ]

"Back when I was mainly SSB operating and messing around with audio gear I heard hams with so much bass boosted they sounded kind of nauseating."

Rob, ever see a chick who looks good from a distance, but then opens her mouth and lets out a bassy, low "Hello there..."   Some of the Extended-SSB guys with overly-boosted lows remind of that.    

The Huzman and I call it "Drag Queen audio."

T

[K5UJ]

"Back when I was mainly SSB operating and messing around with audio gear I heard hams with so much bass boosted they sounded kind of nauseating."

Rob, ever see a chick who looks good from a distance, but then opens her mouth and let out a bassy, low "Hello there..."   Some of the Extended-SSB guys with overly-boosted lows remind of that.    

The Huzman and I call it "Drag Queen audio."

T

That's pretty good    I'll have to keep it in mind.   What I can't get enough of is that muffled bass being the only thing coming through in marginal conditions.  40 to 100 Hz is the strongest part of their signal by about 15 dB.   

Rob