ALC and IMD - Why compression is better than peak limiting on ssb

[K1JJ]

A good article for the archives - thanks to Craig / VK3HE for bringing it to my attention.


This excellent article concerns the bad effects of ALC on IMD using ssb transmitters. It goes into actual spec analyzer examples using an FT-1000D.


"The Abominable ALC":
http://www.sm5bsz.com/dynrange/alc.htm

The problem with limiting amplifier power using ALC is the overshoot on the attack. (slow attack times)  This can generate splatter on the first syllable as shown in the pictures. The better way is to adjust the radio for NO ALC at all and limit the RF using a compressor or built in "RF speech processor." Just a touch, maybe 5db of processing will do it.

Some of the later riceboxes (after the FT-1000D, but before DSP) are truly poor in this area.

I ran some tests with my FT-1000D. After setting it up for no ALC action and about 5db of compression/ speech processing, I saw the side products at 3.5kc away go down about 2 S-units. That is significant. In the past I was running ALC and no compression at all - just the opposite.

I plan to run more system wide tests on my rig and amplifier chain once I set up the SoftRock spec analyzer.  The bottom line is the FT-1000D was the limiting factor to better IMD numbers. All this time I thought it was the linear amplifer.

The later DSP rigs possibly do not have this issue due to pre-processing delays to isolate the leading edge problem, but I do not know this for fact.

*BTW, as we know, most AMers recommend no ALC action when running a ricebox on AM. Evidently we already practiced this, but for other reasons..  

T

[Bill, KD0HG]

Tom, good article and posting.

I gotta ask why not use a fast, broadcast style peak limiter OR fast compressor for use on SSB, as we often do on AM?

If you set the attack time, to say, 1 mS, that's how long it would take to drop one's audio by, say, 6 db. That's 1 cycle of audio at 1 KHz, which should be plenty fast enough. Set the output of the peak limiter to coincide with the max PEP output of the sideband rig and let it brick wall there.

It used to be that simple audio clippers (couple of diodes) would themselves create unpleasant sounding odd-order distortion products, but that's not true any more.

ALC is only a very slow form of compression. Not fast enough, as you have found.

Ideally, you would use a moderately fast audio compressor at maybe 5 db worth to give you a wall of sound, followed by a bloody fast peak limiter.

I think for sideband, the best approach is still the RF clipper technology of the 1960s, where a second IF filter scrubs off any and all clipping and IMD products outside your normal speech bandwidth. A second 2.X KHz filter just wouldn't pass the crud you observed at 3.5 KHz out.





.

[K1JJ]

Hi Bill,

Yes, prelimiting in the audio chain would do it FB.

I have audio limiting here too, but find the compressor in the rig more cornvenient to use. 

Either way, both techniques bring up the background noise. This is my problem when using Dr. Love, so I use just a touch of processing.  I really need a "massage room screen" in front of the rig to block the noise...

You are right about the outboard "RF speech clipping" technique. I must look at the schematic, but I was under the impression the 1000D used this I.F. clipping technique with a second filter, but now I think it was just audio compression.  It wud be cool to get one of those old units, though it wud have to be matched to the IF freq. Maybe one cud be homebrewed. I'll check and see.

T

[WA1GFZ]

Drake made an RF clipper for the TR7. Converts audio to 455 KHz compresses then down converts back to audio. I usually run about 6 dB.

[K1JJ]

Drake made an RF clipper for the TR7. Converts audio to 455 KHz compresses then down converts back to audio. I usually run about 6 dB.

Sounds like one of those Waters units that was used on Collins in the 60's.

T

[K1JJ]

Well, this posting from another site answer the question. Looks like the FT-1000D already has a very FB "RF speech processor" built in.

T


------------


Quote:

    "Dear Rick,

    With all due respect, I'm interested in finding out why you are using
    the Ten Tec RF processor with a 1000D when the 1000D has one of the
    best RF speech processors ever to be included with a ham rig, built
    right into one of the rig's IFs where an RF processor belongs."

     

There is another issue at work here when comparing the Yaesu design to the
Ten-Tec 717....

A check of the FT-1000D User Manual confirms that Yaesu included an RF Speech
Compressor circuit... whereas the Ten-Tec Model 715 is an RF Clipper.... not a
compressor. There are rather significant differences to how compressor and
clipper circuits work, with clippers being more complex in design therefore
generally more costly.... which is why most transceivers use either audio
compression, audio clipping or RF compression for speech processing circuits.
Then of course all audio compressors are not alike.... all RF compressors are
not alike, etc.

There have been more than a few studies over the years comparing the maximum
possible 'clean(ish)' increase in average voice for a RF compressor versus a RF
clipper circuit. Ten-Tec provides a copy of one of those studies with each 715
and I believe it can be downloaded from their website as well. I suggest a good
place to begin is to allow each person to do their own investigation and
afterwards make up their own minds as to what is right for their needs.

73,
Jerry, KG6TT

[WA1GFZ]

Tom,
It is a stand alone box. I think called SP75. it needs 12 volts from the TR7 but current is pretty low. I try to stay out of the ALC when I talk. The TR7 flashes a LED when you have ALC action.

[Bill, KD0HG]

Drake made an RF clipper for the TR7. Converts audio to 455 KHz compresses then down converts back to audio. I usually run about 6 dB.

Sounds like one of those Waters units that was used on Collins in the 60's.

T

I had one of those RF speech clippers in my 32S-1. IIRC, it plugged into the rig right after the factory 2.1 KHz mechanical filter...A little box contained the clipper which acted on the SSB signal, then fed an added second 2.1 KHz mechanical filter. So regardless of how much processing or clipping you dialed in, your bandwidth was gonna be 2.1 KHz. It was almost impossible to screw things up so that you got 'wide'. It all worked at 455 KHz.

As far as bringing up background noise, a pro mic processor is going to have adjustable gating, which works like a bass-ackwards compressor. When there's little or no audio input below an adjustable threshold, it rolls it's own gain back so as not to pick up background noise.

The cheapo DBX-286 mic processor I use has gating built in, which completely cures the audio feedback I used to get from the singing modulation transformer. I can use a shirtload of compression and a bunch of clipping with the Volumax, and the acoustic feedback is eliminated. When there's dead air during moments of silence, there's dead air. That's how highly processed radio stations still manage not to pick up studio background noise, the DJs breathing, papers shuffling and etc.

[K1JJ]

I'm still not cornvinced the 1000D is using true RF clipping in the IF. I say not.  I'll bet it's some quasi-compression scheme, even if done at IF level. 

Yes, using the output of the ssb filter to drive a clipper that then goes into a second filter is the way to go.  It is more expensive cuz of the 2nd filter, so manufacturers probably don't use it.

DSP rigs may have a cheaper scheme to do the filtering after clipping.

I may still look into building one for the IF as described and see what happens.

BTW, Bill, I do have a noise gate I use on ssb. However, the blower noise is still evident between syllables as I speak. It cuts off quickly, but the noise still rides on my speech and is really a form of intermod, actually.  I really need to find a way to put an acoustical screen in front of the rig to drop the noise at least another 10db. I already have the mike turned away and nulled out and a Variac on the blower to slow it down a little. But I'm a big believer in adequate cooling and will not drop the air down below a certain level.

T

[WA1GFZ]

Tom,
I have a pile of Collins LSB filters if you need any.

[K5UJ]

I saw that article a few years ago and it started the chain of events that got me on AM.   I was (and still am) running a FT1000MP mk V and after reading that article, determined that I needed to limit the output without employing the ALC.  I studied the way it works.   Basically, a forward power coupler generates a control voltage sampling some of the PA output.  That v. is fed back to a VCA at a very early stage, I think right after the 455 KHz IF.   the problem with ALC is that by the time a high peak has gotten to the coupler, the herd of cattle are off the reservation and out to the antenna.  But with time constants and VCA gain reduction slowly returning to baseline the next peak can get trapped if it comes soon enough.  I am in a hurry so I am not descibing this very well but I hope you get the idea.   IOW ALC is a poor way to do peak limiting.  
But it is cheap because the method and circitry are mostly there for high vswr protection (in that case the reflected power is sampled and a control voltage fed back and for that the method is pretty effective).

I eventually determined that I needed the kind of tight audio peak limiter that is used in AM broadcasting to control my audio peaks so I could limit the SSB PA output consistently with no overshoot and not get into any ALC involvement.  That led to my purchasing an Inovonics 222 and one day I realized I could use it for AM.  I checked into the DX 60 net one Sunday morning and never operated SSB again, except for one SSB net I check into occasionally on 20 m.  

RF speech processing gets your baseband audio up to a USB signal (typically 455 KHz) clips the peaks and puts the sig. through a filter to remove the IMD products before converting it back to audio.   If you con't care about audio quality it works but  multi-band compression at audio frequencies such as what was found with the old ETO Vomax (they occasionally show up used for around $70) is much more effective in my opinion.   With that, the audio is split up and clipped and filtered to remove the distortion products and recombined.  That is okay for SSB.  About the only thing that isn't is a single band audio clipper that cuts the RF so abruptly that you are transmitting distortion like if you went over 100% negative on AM and clipped the carrier.  It's that fast cutoff whether it's a fast cw rise time or some other instant RF change of state that transmits broad spurious RF.  I'd resort to something like the Vomax,  or DSP compression where you have a knee or soft compression lead-in at audio as opposed to a diode clip, that gets you the desired result without the "key clicks" if you will,  that diode clipping produces.    But the problem of getting by without the ALC intervention will still be there because these devices usually are not fast acting enough to limit your audio peaks so that the ALC will never be triggered.  For that you need something like a 222.  

Rob

[KM1H]

One of the earliest RF clippers was the Comdel of the early 60's. It worked at a very low RF freqency of around 25KHz, iffn I remember, then discrete component filtering and back to baseband. I ran one for decades and always got good reports except when the batteries ran down.

Used it on a HT-37, CE-100V, T4XC and a bunch of little CE and other phasing exciters that drove VHF-UHF gear. Sold it when I got TS-930's.

The Split Band Speech Processor was initially in Ham Radio Magazine. Radiokit offered a kit version and ETO soon copied it as a finished product. I might still have a RK version around but its only usefull on old BA's, even the old Kenwood hybrids had a great sounding RF processor built in.

Ive been experimenting with the TS-950SDX Ive only had less than a month and using the processor on AM and backing off the ALC. Its getting good reports with the LK-500 on 75 but its a royal pita to setup with the way the AM monitoring gear is in another area. I just received a SM-220 and will start using that once I fix the dumbass packing damage, cant blame USPS on this one and its just a 1M Intensity pot.  The 950 has a 6KHz AM filter in both IF's and I have to see if I can do some diode steering to transmit thru the 455 one also. I did similar years ago to a 930 and 940; use the sharp 2100KHz Inrad filter on receive and transmit thru the broad 2.8 ceramic filter that is used normally. Helped give some shoulder room in contests   Some bitched as they thought it "wasnt fair"

Carl

[ke7trp]

I ran a quick test on the Icom 756 pro. Its clean either way.   

c

[Opcom]

So far it seems like compression is the way to go as long as it can be assured the later stages are not ever overdriven. That would depend on the gain and operating conditions of those stages. It may not matter inside a plastic brick but anything with a loading and tuning control may present a varying result.

[K1JJ]

Ive been experimenting with the TS-950SDX Ive only had less than a month Carl

Carl,

Do you suppose the FT-1000D is using a REAL "RF speech processing CLIPPER"? Or is it a homo-chip diode clipper instead of using a second IF filter?   One post on another site mentioned some kind of FM chip in the I.F. doing it for the Mk V.  Check out the 1000D ad below too.

T

Here's what an old FT-1000D ad says:
"On transmit, the 30-Volt power amplifier system puts out a solid 200 Watts, with an RF clipping-type Speech Processor providing a useful boost in talk power. Four separate drive chain controls (MIC GAIN, PROC, DRIVE, and RF PWR) afforded the user precise adjustment capability within the ALC loop."


** Below - Here's another part of that RF clipper thread that refers to the FT-1000 MK V, which is different from the 1000D:

Not sure about FT1000D, but FT1000MP/MkV actually performs RF clipping,
and it's very cheap, using TA7302P FM IF Amplifier.

Users of "external RF Clippers" must be aware of the fact that the processed RF
is converted back to audio, which is then put through the entire TX chain.

There is a very real danger that artificial RF peaks will be reintroduced,
which will trigger ALC (otherwise PA will be overdriven).
This will reduces clipping gain typically by 3-10 dB.
Please consult the Collins SSB book.

Repeaking effect can be minimized on FT1000MP MkV, Orion, etc
by selecting the widest possible SSB TX bandwidth.

Turning OFF the TX equalization is mandatory!

[KM1H]

I owned a 1000D for a few months a few years after they came out and quickly sold it as it offered a step backwards in receiver performance to my hot rodded TS-940's. I never even looked at the TX except to run a few IMD tests.

The new FT-5000 "looks" decent in the plain vanilla QST review but Im not up to spending $6K on the first release of anything.

Carl

[n4wc]

Walter  K1YZW was part of the original design team of the VOMAX.  Walter is still operating 40 and 20 ssb.  He is a retired engineer from the old National Radio Co.  I don't think he worked in the amateur design group.  He used to have friendly arguments with the owner the the company that made the DATONG processors.  Split band processors were the only way to go ;-).
Walter wrote quite a few articles for Ham Radio Magazine.  A really good guy, and still going strong in his 80's.

Bill   N4WC

[KM1H]

I know Walter well and worked with him at National. He was in the commercial (ham, etc) R&D group for awhile and Comdel was a side investment for him and another did the day to day work. When National started their layoffs he went early and was always bitter about that.

Carl