Elecraft

RX Mush, why the fuss?

Classic

List

Threaded

17 messages Options

Wes Stewart-2

RX Mush, why the fuss?

I must confess to some bewilderment about the seemingly endless discussion about
the adjustment, or mis-adjustment, of AGC slope, threshold or "RF" gain.

Claims are made that one's favorite settings cause signals within the passband
to retain their relative amplitudes thus allowing the discrimination between
them, while less favorable settings compress them into "mush."

I will confess that my experience with the design of AGC systems is limited to
analog receivers and perhaps there is some digital magic that makes DSP radios
act differently from analog ones in this case. But in my experience, AGC
control is derived from the stronger signal received.

After the SNR is adequate (delayed AGC in 1960s terms, above threshold today),
the overall gain is reduced by some amount to maintain a desired output or to
prevent overload, and any other signals present suffer the same gain reduction.
Hence a signal 30 dB stronger than another is still 30 dB stronger even after
the application of AGC. If it isn't then we have a very nonlinear receiver,
which is desirable if we're receiving FM but highly undesirable otherwise.

My reading between the lines suggests that the "mush" proponents think that
after achieving threshold, changing the slope somehow changes the ratio between
signals, i.e. there is less gain for strong signals than there is for weaker ones.

Frankly, after 60 years of listening to shortwave noise and in my youth working
in a machine shop and hanging around too many alcohol and nitro burning race
cars, my tinnitus practically drives me nuts at times; I welcome a flat AGC slope.

If I'm all wet with this, I'd like to be enlightened.

Wes N7WS

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

k6dgw

Re: RX Mush, why the fuss?

I too have come to that same surmise, Wes, and I hope whoever explains
it [someone always does] does so on the reflector so I can benefit too.
One of my K3's two AGC systems is highly configurable. I finally found
a threshold and fairly flat slope that, for my compromised hearing works
very well. I operate very little SSB, and I notice nothing I'd call
"mush" on CW at narrow DSP bandwidths. If there are multiple signals in
that BW, the strongest one sets the gain, and all the others respond
accordingly.

I normally use AGC-F on CW. Occasionally, with a large difference in
signal strengths, the gain reduction from the code elements of the
stronger signal make copy of the weaker signal at a different pitch
difficult. AGC-S almost always cures that issue.

73,

Fred ["Skip"] K6DGW
Sparks NV DM09dn
Washoe County

On 2/28/2017 12:53 PM, Wes Stewart wrote:

> My reading between the lines suggests that the "mush" proponents think
> that after achieving threshold, changing the slope somehow changes the
> ratio between signals, i.e. there is less gain for strong signals than
> there is for weaker ones.
>
> Frankly, after 60 years of listening to shortwave noise and in my
> youth working in a machine shop and hanging around too many alcohol
> and nitro burning race cars, my tinnitus practically drives me nuts at
> times; I welcome a flat AGC slope.
>
> If I'm all wet with this, I'd like to be enlightened.
>
> Wes N7WS

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

Dave-7

Re: RX Mush, why the fuss?

In reply to this post by Wes Stewart-2

My thoughts on this are that those who are concerned about the slope
and threshold settings are barking up the wrong tree. The mush would
result if you have your hold time or hang time or decay set too short.
With a brief hold time the weaker signal pops up to the level of the
stronger on as soon as the stronger one disappears. A longer hold time
keeps the relative level of the two signals at the correct
relationship. The answer is hold time, not threshold or slope.

Those of us in the southeastern US may have a problem with all the
lighting we get. The lightning spikes tend to drive the desired signal
too low. So we kinda have to keep hold time short if we are to hear
anything. But there should be some level of decent compromise in there
somewhere. IIRC the K3 has an AGC setting that helps with this but
does not eliminate it.

72 de dave
ab9ca/4

On 2/28/17 2:53 PM, Wes Stewart wrote:

> I must confess to some bewilderment about the seemingly endless
> discussion about the adjustment, or mis-adjustment, of AGC slope,
> threshold or "RF" gain.
>
> Claims are made that one's favorite settings cause signals within the
> passband to retain their relative amplitudes thus allowing the
> discrimination between them, while less favorable settings compress
> them into "mush."
>
> I will confess that my experience with the design of AGC systems is
> limited to analog receivers and perhaps there is some digital magic
> that makes DSP radios act differently from analog ones in this case.
> But in my experience, AGC control is derived from the stronger signal
> received.
>
> After the SNR is adequate (delayed AGC in 1960s terms, above threshold
> today), the overall gain is reduced by some amount to maintain a
> desired output or to prevent overload, and any other signals present
> suffer the same gain reduction. Hence a signal 30 dB stronger than
> another is still 30 dB stronger even after the application of AGC. If
> it isn't then we have a very nonlinear receiver, which is desirable if
> we're receiving FM but highly undesirable otherwise.
>
> My reading between the lines suggests that the "mush" proponents think
> that after achieving threshold, changing the slope somehow changes the
> ratio between signals, i.e. there is less gain for strong signals than
> there is for weaker ones.
>
> Frankly, after 60 years of listening to shortwave noise and in my
> youth working in a machine shop and hanging around too many alcohol
> and nitro burning race cars, my tinnitus practically drives me nuts at
> times; I welcome a flat AGC slope.
>
> If I'm all wet with this, I'd like to be enlightened.
>
> Wes N7WS
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to [hidden email]
> .
>

Don Wilhelm

Re: RX Mush, why the fuss?

I am not certain what it has to do with "mush", but it does seem to help.

What is certain is that if the AGC Threshold is set too low, the AGC
will be activated on band noise. Whether that 'band noise' is the
result of ambient noise level, or the ambient 'noise' of a multitude of
signals in close frequency proximity is for the user to determine.

As far as the slope is concerned, yes, the AGC will respond to the
strongest signal in the passband, that is a fact.
The reason for setting the slope to something other than flat is not a
consideration for reducing the RX Mush, but it allows the user to
evaluate by ear the relative strength of signals.

In summary, signals below the threshold will be received the same as
with the AGC off. Signals above the threshold will be reduced according
to the slope of the AGC response, but the strongest one in the passband
will take control of the AGC.

It is nice to be able to report an S6 signal is weaker than an S8 or S9
signal without depending on the S-meter. Of course, all signals in a
contest or for reports to a DX station are always 59 or 599.

On 2/28/2017 6:37 PM, dave wrote:
>
> My thoughts on this are that those who are concerned about the slope and
> threshold settings are barking up the wrong tree. The mush would result
> if you have your hold time or hang time or decay set too short. With a
> brief hold time the weaker signal pops up to the level of the stronger
> on as soon as the stronger one disappears. A longer hold time keeps the
> relative level of the two signals at the correct relationship. The
> answer is hold time, not threshold or slope.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

David Gilbert

Re: RX Mush, why the fuss?

In reply to this post by Dave-7

That isn't actually true. Threshold and slope combine to form a point
of non-linearity that can cause all sorts of in-band mixing products
when multiple signals exist at roughly the same levels. I and others
have experienced that first hand in the past. I can running stations in
a contest with a fairly narrow passband and if I get more than a few
callers at roughly the same level, and if that level is in the vicinity
of the knee in the gain curve formed by the onset of the AGC, the A-2B
and 2A-B (etc) products generated by the nonlinearity create all sorts
of mush that muddles the copy. Noise at that point in the curve can
perform similar dirty deeds.

I use as little AGC slope as possible with a fairly high threshold as a
result ... tough on the ears sometimes but helps the rate dramatically.

If I remember correctly, even without the AGC there is a knee at the low
end of the response curve (but still above the noise limit) in the
original synths. I recently purchased the new synths for my K3 and
supposedly they help significantly on that score, but I haven't had the
opportunity to install them yet.

73,
Dave AB7E

On 2/28/2017 4:37 PM, dave wrote:

>
> My thoughts on this are that those who are concerned about the slope
> and threshold settings are barking up the wrong tree. The mush would
> result if you have your hold time or hang time or decay set too short.
> With a brief hold time the weaker signal pops up to the level of the
> stronger on as soon as the stronger one disappears. A longer hold time
> keeps the relative level of the two signals at the correct
> relationship. The answer is hold time, not threshold or slope.
>
> Those of us in the southeastern US may have a problem with all the
> lighting we get. The lightning spikes tend to drive the desired signal
> too low. So we kinda have to keep hold time short if we are to hear
> anything. But there should be some level of decent compromise in there
> somewhere. IIRC the K3 has an AGC setting that helps with this but
> does not eliminate it.
>
> 72 de dave
> ab9ca/4
>
>
>
> On 2/28/17 2:53 PM, Wes Stewart wrote:
>> I must confess to some bewilderment about the seemingly endless
>> discussion about the adjustment, or mis-adjustment, of AGC slope,
>> threshold or "RF" gain.
>>
>> Claims are made that one's favorite settings cause signals within the
>> passband to retain their relative amplitudes thus allowing the
>> discrimination between them, while less favorable settings compress
>> them into "mush."
>>
>> I will confess that my experience with the design of AGC systems is
>> limited to analog receivers and perhaps there is some digital magic
>> that makes DSP radios act differently from analog ones in this case.
>> But in my experience, AGC control is derived from the stronger signal
>> received.
>>
>> After the SNR is adequate (delayed AGC in 1960s terms, above threshold
>> today), the overall gain is reduced by some amount to maintain a
>> desired output or to prevent overload, and any other signals present
>> suffer the same gain reduction. Hence a signal 30 dB stronger than
>> another is still 30 dB stronger even after the application of AGC. If
>> it isn't then we have a very nonlinear receiver, which is desirable if
>> we're receiving FM but highly undesirable otherwise.
>>
>> My reading between the lines suggests that the "mush" proponents think
>> that after achieving threshold, changing the slope somehow changes the
>> ratio between signals, i.e. there is less gain for strong signals than
>> there is for weaker ones.
>>
>> Frankly, after 60 years of listening to shortwave noise and in my
>> youth working in a machine shop and hanging around too many alcohol
>> and nitro burning race cars, my tinnitus practically drives me nuts at
>> times; I welcome a flat AGC slope.
>>
>> If I'm all wet with this, I'd like to be enlightened.
>>
>> Wes N7WS
>>
>> ______________________________________________________________
>> Elecraft mailing list
>> Home: http://mailman.qth.net/mailman/listinfo/elecraft
>> Help: http://mailman.qth.net/mmfaq.htm
>> Post: mailto:[hidden email]
>>
>> This list hosted by: http://www.qsl.net
>> Please help support this email list: http://www.qsl.net/donate.html
>> Message delivered to [hidden email]
>> .
>>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to [hidden email]
>

Jim Brown-10

Re: RX Mush, why the fuss?

On Wed,3/1/2017 10:29 PM, David Gilbert wrote:
> I recently purchased the new synths for my K3 and supposedly they help
> significantly on that score, but I haven't had the opportunity to
> install them yet.

Don't put it off -- it's an easy 30 minute job.

73, Jim K9YC

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

alorona

Re: RX Mush, why the fuss?

In reply to this post by David Gilbert

It isn't so much that 'the strongest signal in the passband determines the gain of the receiver', it's that once that strong signal sends the
receiver into AGC, additional signals in the passband do not increase the audio output power when the Slope is set at or near its extreme. This is a form of gain compression, which is distortion, strictly speaking.

The 'landmark' paper on this was written by Jack Smith, K8ZOA, available at: http://www.cliftonlaboratories.com/elecraft_k3_agc_and_s-meter.htm . In it, Jack showed that the K3's curve of Audio Output vs. RF Input at Slope = 15 acts as a hard limiter. He follows that with a table of measurements where a 10 dB change in RF signal level results in virtually no increase in audio output. This 10 dB change in RF signal level could come from a single signal that increases by 10 dB, or additional signals in the passband that add 10 dB of RF input to an existing 'strong' signal -- the receiver doesn't care which.

For many listeners, more signals added to the passband that don't result in any more audio is a condition that confuses and fatigues the brain and can make it difficult to decode the relationship between signals. I believe this is the 'mush' that has been reported here. The effect occurs on CW, too, though I believe that the effect is made even worse by the heavily compressed phone signals that many contesters generate.

I completely agree with Dave AB7E that the use of as little slope as possible (lower values of the AGC Slope parameter), coupled with higher values of AGC Threshhold allows the receiver to sound very natural, or open, or clean, and preserves as much as possible the relationship between multiple signals, which makes it easier for your brain to copy them. (I'm sorry for using such non-scientific terms, but it's the best I can do to describe it.)

Al W6LX

>>
>> After the SNR is adequate (delayed AGC in 1960s terms, above threshold
>> today), the overall gain is reduced by some amount to maintain a
>> desired output or to prevent overload, and any other signals present
>> suffer the same gain reduction. Hence a signal 30 dB stronger than
>> another is still 30 dB stronger even after the application of AGC.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

ab2tc

Re: RX Mush, why the fuss?

Hi,

Where in Smith's article does it say that AGC with the slope set for 15 acts as a hard limiter? There is a huge difference between AGC action (which is simply a reduction in gain with linearity retained) and hard limiting. I have read the article and agree that his measurements are very helpful. The "waviness" he observes he correctly attributes to the log function in the DSP being less than accurate. This has since been corrected as far as I know. Whatever some people at the receiving end of a CW pileup hear is not caused by hard limiting. Why can't we ever see some hard evidence of this phenomenon (like a video with quality audio or at least an audio only recording)?

AB2TC - Knut

alorona wrote

It isn't so much that 'the strongest signal in the passband determines the gain of the receiver', it's that once that strong signal sends the
receiver into AGC, additional signals in the passband do not increase the audio output power when the Slope is set at or near its extreme. This is a form of gain compression, which is distortion, strictly speaking.

The 'landmark' paper on this was written by Jack Smith, K8ZOA, available at: http://www.cliftonlaboratories.com/elecraft_k3_agc_and_s-meter.htm . In it, Jack showed that the K3's curve of Audio Output vs. RF Input at Slope = 15 acts as a hard limiter. He follows that with a table of measurements where a 10 dB change in RF signal level results in virtually no increase in audio output. This 10 dB change in RF signal level could come from a single signal that increases by 10 dB, or additional signals in the passband that add 10 dB of RF input to an existing 'strong' signal -- the receiver doesn't care which.

For many listeners, more signals added to the passband that don't result in any more audio is a condition that confuses and fatigues the brain and can make it difficult to decode the relationship between signals. I believe this is the 'mush' that has been reported here. The effect occurs on CW, too, though I believe that the effect is made even worse by the heavily compressed phone signals that many contesters generate.

I completely agree with Dave AB7E that the use of as little slope as possible (lower values of the AGC Slope parameter), coupled with higher values of AGC Threshhold allows the receiver to sound very natural, or open, or clean, and preserves as much as possible the relationship between multiple signals, which makes it easier for your brain to copy them. (I'm sorry for using such non-scientific terms, but it's the best I can do to describe it.)

Al W6LX

<snip>

Don Wilhelm

Re: RX Mush, why the fuss?

The AGC will not cause "hard limiting" - in other words, it should
maintain linearity.

Even though the strongest signal in the passband will control how much
AGC is applied, weaker signals in the passband should sound
proportionally weaker.

Now that you mention hard limiting, there is a limiter in the K3 that if
turned on will protect your ears. I am wondering if some instances of
reported receiver mush did have limiting set on - that would be
particularly true for those who chose to ride the RF Gain and/or run
with AGC off.

73,
Don W3FPR

On 3/2/2017 3:37 PM, ab2tc wrote:
> Hi,
>
> Where in Smith's article does it say that AGC with the slope set for 15 acts
> as a hard limiter? There is a huge difference between AGC action (which is
> simply a reduction in gain with linearity retained) and hard limiting.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

alorona

Re: RX Mush, why the fuss?

In reply to this post by ab2tc

Oh no! I fear this is going to get bogged down in definitions. From Wikipedia:

"Limiting can refer to non-linear clipping, in which a signal is passed through normally but 'sheared off' when it would normally exceed a certain threshold. It can also refer to a type of variable-gain audio level compression, in which the gain of an amplifier is changed very quickly to prevent the signal from going over a certain amplitude.
* Hard limiting ("clipping") is a limiting action in which there is
* (a) over the permitted dynamic range, negligible variation in the expected characteristic of the output signal, and
* (b) a steady-state signal, at the maximum permitted level, for the duration of each period when the output would otherwise be required to exceed the permitted dynamic range in order to correspond to the transfer function of the device."

AGC of the type we're discussing falls under this definition, and not the non-linear "back-to-back diode" clipping you might be thinking of. Look at the first graph under the section titled "Adjusting AGC SLP" and observe that for Slope=15, above about -104 dBm, the output follows a horizontal line. Moreover, the very first table that follows that graph shows that with Slope=15 if the input increases by 10 dB the output increase is virtually zero. That transfer characteristic is what audio engineers call "hard limiting".

I also fear we may be talking past each other when we say 'linear' and 'nonlinear'. The AGC curve is 'nonlinear' in the sense that if we pour more RF input into the receiver, the output doesn't get any bigger. It's like that by design. All I was saying is that it doesn't sound good to me. I'm very happy there is Slope = 5.

Call it whatever you want; the curve speaks for itself.

Al W6LX

>>> There is a huge difference between AGC action (which is

>>> simply a reduction in gain with linearity retained) and hard limiting.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

Guy Olinger K2AV

Re: RX Mush, why the fuss?

One problem here is that a lot of the discussion has been in analog terms.

We have to remember that in the K3, the digital AGC controlled by the AGC
parms is just that, a digital algorithm. What it does has nothing to do
with diodes. It can do anything weird and completely non analog resembling,
it only cares about the program code.

The graphs I have seen are entirely based on steady or very slow moving
signal states, not a photo of an audio rate amplitude varying signal
traversing the AGC knee. They seem intended, well-enough done, just to
convey the rudimentary function variants.

You need to have the program code to estimate exactly what is happening to
the AGC at audio rates. Good luck with that.

The only control you have over the non parameterized hardware AGC is to
reduce the gain in front of it so it isn't engaged. You can't turn it off,
it's always potential if the signal coming through the roofing filter is
getting up to around 20 over 9.

If you have a pile-up of 20 over signals, it's time to turn off PRE, or
turn on ATT, or back off the RF gain. Otherwise you are engaging the
hardware AGC, not sophisticated, which is only there to properly range
input to the ADC chip.

This ain't your grand-daddy's analog radio.

73, Guy

On Thu, Mar 2, 2017 at 11:16 PM, Al Lorona <[hidden email]> wrote:

> Oh no! I fear this is going to get bogged down in definitions. From
> Wikipedia:
>
>
> "Limiting can refer to non-linear clipping, in which a signal is passed
> through normally but 'sheared off' when it would normally exceed a certain
> threshold. It can also refer to a type of variable-gain audio level
> compression, in which the gain of an amplifier is changed very quickly to
> prevent the signal from going over a certain amplitude.
> * Hard limiting ("clipping") is a limiting action in which there is
> * (a) over the permitted dynamic range, negligible variation in the
> expected characteristic of the output signal, and
> * (b) a steady-state signal, at the maximum permitted level, for the
> duration of each period when the output would otherwise be required to
> exceed the permitted dynamic range in order to correspond to the transfer
> function of the device."
>
>
>
> AGC of the type we're discussing falls under this definition, and not the
> non-linear "back-to-back diode" clipping you might be thinking of. Look at
> the first graph under the section titled "Adjusting AGC SLP" and observe
> that for Slope=15, above about -104 dBm, the output follows a horizontal
> line. Moreover, the very first table that follows that graph shows that
> with Slope=15 if the input increases by 10 dB the output increase is
> virtually zero. That transfer characteristic is what audio engineers call
> "hard limiting".
>
> I also fear we may be talking past each other when we say 'linear' and
> 'nonlinear'. The AGC curve is 'nonlinear' in the sense that if we pour more
> RF input into the receiver, the output doesn't get any bigger. It's like
> that by design. All I was saying is that it doesn't sound good to me. I'm
> very happy there is Slope = 5.
>
> Call it whatever you want; the curve speaks for itself.
>
> Al W6LX
>
>
>
>
> >>> There is a huge difference between AGC action (which is
>
> >>> simply a reduction in gain with linearity retained) and hard limiting.
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to [hidden email]
>

David Gilbert

Re: RX Mush, why the fuss?

In reply to this post by Don Wilhelm

I've had my K3 since 2008 or so, and over the years I've seen people
describe different forms of "mush". One set of comments indeed involved
complaints about the hard limit at the upper end that has nothing to do
with AGC. It is, as you say, simply a hard limit ... pretty much a
clipper to protect the ears (and maybe also to help protect the output
stage in the speaker driver before that issue got addressed). That
creates a distortion, but it's not really what I would describe as "mush."

The nonlinearity I described in my earlier post was at the opposite end
of the curve ... down where the AGC just begins to kick in. As W6LX
says, it's a nonlinearity in the curve, and no matter what you call it
that contributes to the generation of mixing products from multiple
signals that happen to be at roughly the same level within the
passband. The low end of Jack Smith's plots showed that pretty
clearly. During some of my contest runs, individual signals were
perfectly clear and distinguishable, two not terrible, but even three
signals could generate enough mixing products to cause problems if they
were low enough in volume and close enough in frequency. Since I
typically operate with a very narrow passband (about 150 HZ on CW), the
mixing products end up very close to the real signals. For example,
2x500Hz - 510 Hz gives another phantom signal at 490 Hz. Things get
really messy with three or more signals.

It is also, possible, of course, to get mixing anywhere there is a knee
in the AGC curve, but if you put the knee up higher there is less
likelihood that multiple signals will be of the same amplitude to cause
a problem (one will dominate), and their amplitude swings will range
further afield of the knee ... meaning that a lower percentage of the
energy will be mixed. At the low end, you're pretty much screwed ...
any signal you hear will be at that nonlinearity and the amplitude
swings will be small enough that they spend all their time in the
nonlinearity. As I said before, reputedly the new synths greatly
improve this.

The bottom line is that if you have two or more signals within a
passband that traverse a nonlinearity, you get mixing products within
the same passband that blur the individual signals ... i.e., "mush."
And since the mixing products on CW only occur when both (or more) of
the signals are keyed, the mixing products aren't even intelligible. ;)

At least this is how I understand the situation. I'd be happy to get
corrected if my comments are flawed.

73,
Dave AB7E

On 3/2/2017 3:19 PM, Don Wilhelm wrote:

>
> Now that you mention hard limiting, there is a limiter in the K3 that
> if turned on will protect your ears. I am wondering if some instances
> of reported receiver mush did have limiting set on - that would be
> particularly true for those who chose to ride the RF Gain and/or run
> with AGC off.
>
> 73,
> Don W3FPR
>
> On 3/2/2017 3:37 PM, ab2tc wrote:
>> Hi,
>>
>> Where in Smith's article does it say that AGC with the slope set for
>> 15 acts
>> as a hard limiter? There is a huge difference between AGC action
>> (which is
>> simply a reduction in gain with linearity retained) and hard limiting.
>

Guy Olinger K2AV

Re: RX Mush, why the fuss?

There is some conflation of two quite different concepts going on here.

The first thing you need to know about an AGC response graph is the speed
that the incoming signal was varied to produce the curve. In many cases,
the input signal was steady state from a signal generator, set to a list of
input levels, observing output levels, both recorded in Excel, and the
resultant data pairs used to create a graph line. In this case the input
variation speed is zero. This is a static analysis.

If the input signal was **amplitude**-swept at audio rates, and together
with the output signal, used to provide the X, Y values to drive an
oscilloscope, then you have a dynamic analysis.

At this point it is good to make a note of what test equipment you are
familiar with that will provide an **amplitude**-swept, steady frequency
signal.

Inferences from a static AGC analysis and AGC induced IMD are apples and
oranges.

The second thing that bears heavily is the attack and decay speeds.

Attack speeds are usually quick. If the attack and decay are **BOTH**
quick, and that actual attack/decay is at an audio rate, then there is a
case for distortion, because the variable gain can actually work at an
audio rate.

The question is whether the attack/decay cycle can continuously recur
because the decay goes down as fast as the attack goes up, then intermod is
indeed possible on a grand scale. However if the decay effectively holds
the AGC gain level at a point set by the attack, delaying even as little as
100 milliseconds, then the AGC cannot create audio distortion products
except very short low frequency distortion products only at AGC attacks.

Since well before the significant AGC changes in firmware 4.7x (or whatever
that one was), I have been running my slow AGC (CONFIG: AGC-S) at maximum
fast, and my fast AGC (CONFIG: AGC-F) at maximum slow. In retrospect, that
was probably why I never heard the stuff that a lot of people were
complaining about.

In contests I always use my max fast setting slow AGC, and back off the RF
gain when I have primarily very loud signals in pile-ups to get the signals
out of hardware AGC range, which has zero intelligent tweaks available.

[And yes I have just about guaranteed pile-ups in contests with for-credit
USA to USA QSO's, because of RBN spots which pick up everyone. Those are
"spotting pile-ups" and assisted or unlimited class folks using point and
click on the band map or control characters to move to the next unworked
station.]

My exception to using max fast setting slow AGC is when I'm trying to copy
through lightning static, and need to hear weaker stations down in between
the crashes. Then I use my max slow setting fast AGC.

To summarize, in order for AGC to create audio distortion products strictly
from the AGC, the AGC must be responding at an audio rate. Frankly, why
would anyone want to set it that way escapes me.

To Wayne, I would like to be able to set a minimum hold for fast AGC as
well. That with a fast decay, would be better than what we have.

Decay rate is something left over from analog days, when the way you
decayed AGC was letting a capacitor discharge.

73, Guy K2AV

On Fri, Mar 3, 2017 at 3:02 AM, David Gilbert <[hidden email]>
wrote:

>
> I've had my K3 since 2008 or so, and over the years I've seen people
> describe different forms of "mush". One set of comments indeed involved
> complaints about the hard limit at the upper end that has nothing to do
> with AGC. It is, as you say, simply a hard limit ... pretty much a clipper
> to protect the ears (and maybe also to help protect the output stage in the
> speaker driver before that issue got addressed). That creates a
> distortion, but it's not really what I would describe as "mush."
>
> The nonlinearity I described in my earlier post was at the opposite end of
> the curve ... down where the AGC just begins to kick in. As W6LX says,
> it's a nonlinearity in the curve, and no matter what you call it that
> contributes to the generation of mixing products from multiple signals that
> happen to be at roughly the same level within the passband. The low end of
> Jack Smith's plots showed that pretty clearly. During some of my contest
> runs, individual signals were perfectly clear and distinguishable, two not
> terrible, but even three signals could generate enough mixing products to
> cause problems if they were low enough in volume and close enough in
> frequency. Since I typically operate with a very narrow passband (about
> 150 HZ on CW), the mixing products end up very close to the real signals.
> For example, 2x500Hz - 510 Hz gives another phantom signal at 490 Hz.
> Things get really messy with three or more signals.
>
> It is also, possible, of course, to get mixing anywhere there is a knee in
> the AGC curve, but if you put the knee up higher there is less likelihood
> that multiple signals will be of the same amplitude to cause a problem (one
> will dominate), and their amplitude swings will range further afield of the
> knee ... meaning that a lower percentage of the energy will be mixed. At
> the low end, you're pretty much screwed ... any signal you hear will be at
> that nonlinearity and the amplitude swings will be small enough that they
> spend all their time in the nonlinearity. As I said before, reputedly the
> new synths greatly improve this.
>
> The bottom line is that if you have two or more signals within a passband
> that traverse a nonlinearity, you get mixing products within the same
> passband that blur the individual signals ... i.e., "mush." And since the
> mixing products on CW only occur when both (or more) of the signals are
> keyed, the mixing products aren't even intelligible. ;)
>
> At least this is how I understand the situation. I'd be happy to get
> corrected if my comments are flawed.
>
> 73,
> Dave AB7E
>
>
> On 3/2/2017 3:19 PM, Don Wilhelm wrote:
>
>>
>> Now that you mention hard limiting, there is a limiter in the K3 that if
>> turned on will protect your ears. I am wondering if some instances of
>> reported receiver mush did have limiting set on - that would be
>> particularly true for those who chose to ride the RF Gain and/or run with
>> AGC off.
>>
>> 73,
>> Don W3FPR
>>
>> On 3/2/2017 3:37 PM, ab2tc wrote:
>>
>>> Hi,
>>>
>>> Where in Smith's article does it say that AGC with the slope set for 15
>>> acts
>>> as a hard limiter? There is a huge difference between AGC action (which
>>> is
>>> simply a reduction in gain with linearity retained) and hard limiting.
>>>
>>
>>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to [hidden email]
>

David Gilbert

Re: RX Mush, why the fuss?

I agree with your comments. Thanks for the clarifications!

And I did indeed forget to mention the attack/decay speed influences. I
even posted comments here about that myself back shortly after I bought
my K3 ... that the time rate of change in gain is itself a
non-linearity. I think my settings are similar to yours (too lazy to
check right now).

73,
Dave AB7E

On 3/3/2017 11:28 AM, Guy Olinger K2AV wrote:

> There is some conflation of two quite different concepts going on here.
>
> The first thing you need to know about an AGC response graph is the
> speed that the incoming signal was varied to produce the curve. In
> many cases, the input signal was steady state from a signal generator,
> set to a list of input levels, observing output levels, both recorded
> in Excel, and the resultant data pairs used to create a graph line. In
> this case the input variation speed is zero. This is a static analysis.
>
> If the input signal was **amplitude**-swept at audio rates, and
> together with the output signal, used to provide the X, Y values to
> drive an oscilloscope, then you have a dynamic analysis.
>
> At this point it is good to make a note of what test equipment you are
> familiar with that will provide an **amplitude**-swept, steady
> frequency signal.
>
> Inferences from a static AGC analysis and AGC induced IMD are apples
> and oranges.
>
> The second thing that bears heavily is the attack and decay speeds.
>
> Attack speeds are usually quick. If the attack and decay are **BOTH**
> quick, and that actual attack/decay is at an audio rate, then there is
> a case for distortion, because the variable gain can actually work at
> an audio rate.
>
> The question is whether the attack/decay cycle can continuously recur
> because the decay goes down as fast as the attack goes up, then
> intermod is indeed possible on a grand scale. However if the decay
> effectively holds the AGC gain level at a point set by the attack,
> delaying even as little as 100 milliseconds, then the AGC cannot
> create audio distortion products except very short low frequency
> distortion products only at AGC attacks.
>
> Since well before the significant AGC changes in firmware 4.7x (or
> whatever that one was), I have been running my slow AGC (CONFIG:
> AGC-S) at maximum fast, and my fast AGC (CONFIG: AGC-F) at maximum
> slow. In retrospect, that was probably why I never heard the stuff
> that a lot of people were complaining about.
>
> In contests I always use my max fast setting slow AGC, and back off
> the RF gain when I have primarily very loud signals in pile-ups to get
> the signals out of hardware AGC range, which has zero intelligent
> tweaks available.
>
> [And yes I have just about guaranteed pile-ups in contests with
> for-credit USA to USA QSO's, because of RBN spots which pick up
> everyone. Those are "spotting pile-ups" and assisted or unlimited
> class folks using point and click on the band map or control
> characters to move to the next unworked station.]
>
> My exception to using max fast setting slow AGC is when I'm trying to
> copy through lightning static, and need to hear weaker stations down
> in between the crashes. Then I use my max slow setting fast AGC.
>
> To summarize, in order for AGC to create audio distortion products
> strictly from the AGC, the AGC must be responding at an audio rate.
> Frankly, why would anyone want to set it that way escapes me.
>
> To Wayne, I would like to be able to set a minimum hold for fast AGC
> as well. That with a fast decay, would be better than what we have.
>
> Decay rate is something left over from analog days, when the way you
> decayed AGC was letting a capacitor discharge.
>
> 73, Guy K2AV
>
> On Fri, Mar 3, 2017 at 3:02 AM, David Gilbert
> <[hidden email] <mailto:[hidden email]>> wrote:
>
>
> I've had my K3 since 2008 or so, and over the years I've seen
> people describe different forms of "mush". One set of comments
> indeed involved complaints about the hard limit at the upper end
> that has nothing to do with AGC. It is, as you say, simply a hard
> limit ... pretty much a clipper to protect the ears (and maybe
> also to help protect the output stage in the speaker driver before
> that issue got addressed). That creates a distortion, but it's
> not really what I would describe as "mush."
>
> The nonlinearity I described in my earlier post was at the
> opposite end of the curve ... down where the AGC just begins to
> kick in. As W6LX says, it's a nonlinearity in the curve, and no
> matter what you call it that contributes to the generation of
> mixing products from multiple signals that happen to be at roughly
> the same level within the passband. The low end of Jack Smith's
> plots showed that pretty clearly. During some of my contest runs,
> individual signals were perfectly clear and distinguishable, two
> not terrible, but even three signals could generate enough mixing
> products to cause problems if they were low enough in volume and
> close enough in frequency. Since I typically operate with a very
> narrow passband (about 150 HZ on CW), the mixing products end up
> very close to the real signals. For example, 2x500Hz - 510 Hz
> gives another phantom signal at 490 Hz. Things get really messy
> with three or more signals.
>
> It is also, possible, of course, to get mixing anywhere there is a
> knee in the AGC curve, but if you put the knee up higher there is
> less likelihood that multiple signals will be of the same
> amplitude to cause a problem (one will dominate), and their
> amplitude swings will range further afield of the knee ... meaning
> that a lower percentage of the energy will be mixed. At the low
> end, you're pretty much screwed ... any signal you hear will be at
> that nonlinearity and the amplitude swings will be small enough
> that they spend all their time in the nonlinearity. As I said
> before, reputedly the new synths greatly improve this.
>
> The bottom line is that if you have two or more signals within a
> passband that traverse a nonlinearity, you get mixing products
> within the same passband that blur the individual signals ...
> i.e., "mush." And since the mixing products on CW only occur when
> both (or more) of the signals are keyed, the mixing products
> aren't even intelligible. ;)
>
> At least this is how I understand the situation. I'd be happy to
> get corrected if my comments are flawed.
>
> 73,
> Dave AB7E
>
>
> On 3/2/2017 3:19 PM, Don Wilhelm wrote:
>
>
> Now that you mention hard limiting, there is a limiter in the
> K3 that if turned on will protect your ears. I am wondering
> if some instances of reported receiver mush did have limiting
> set on - that would be particularly true for those who chose
> to ride the RF Gain and/or run with AGC off.
>
> 73,
> Don W3FPR
>
> On 3/2/2017 3:37 PM, ab2tc wrote:
>
> Hi,
>
> Where in Smith's article does it say that AGC with the
> slope set for 15 acts
> as a hard limiter? There is a huge difference between AGC
> action (which is
> simply a reduction in gain with linearity retained) and
> hard limiting.
>
>
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> <http://mailman.qth.net/mailman/listinfo/elecraft>
> Help: http://mailman.qth.net/mmfaq.htm
> <http://mailman.qth.net/mmfaq.htm>
> Post: mailto:[hidden email]
> <mailto:[hidden email]>
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to [hidden email] <mailto:[hidden email]>
>
>

Eric Swartz - WA6HHQ

Re: RX Mush, why the fuss?

Administrator

This has been an interesting discussion. We've certainly passed the posting
overload limit on it though, so lets wind it down in the next hour to let others
recuperate from email overload ;-)

73,
Eric
List mooderator..
/elecraft.com/

On 3/3/2017 10:59 AM, David Gilbert wrote:

>
> I agree with your comments. Thanks for the clarifications!
>
> And I did indeed forget to mention the attack/decay speed influences. I even
> posted comments here about that myself back shortly after I bought my K3 ...
> that the time rate of change in gain is itself a non-linearity. I think my
> settings are similar to yours (too lazy to check right now).
>
> 73,
> Dave AB7E

wc1m

Re: RX Mush, why the fuss?

In reply to this post by Wes Stewart-2

Guy K2AV and Dave AB7E!

Your comments are incredibly helpful. I and many others would be very grateful if you would post all your Config setting related to coping with RX mush.

73, Dick WC1M

73, Dick WC1M

> On Mar 3, 2017, at 1:59 PM, David Gilbert <[hidden email]> wrote:
>
>
> I agree with your comments. Thanks for the clarifications!
>
> And I did indeed forget to mention the attack/decay speed influences. I even posted comments here about that myself back shortly after I bought my K3 ... that the time rate of change in gain is itself a non-linearity. I think my settings are similar to yours (too lazy to check right now).
>
> 73,
> Dave AB7E
>
>
>> On 3/3/2017 11:28 AM, Guy Olinger K2AV wrote:
>> There is some conflation of two quite different concepts going on here.
>>
>> The first thing you need to know about an AGC response graph is the speed that the incoming signal was varied to produce the curve. In many cases, the input signal was steady state from a signal generator, set to a list of input levels, observing output levels, both recorded in Excel, and the resultant data pairs used to create a graph line. In this case the input variation speed is zero. This is a static analysis.
>>
>> If the input signal was **amplitude**-swept at audio rates, and together with the output signal, used to provide the X, Y values to drive an oscilloscope, then you have a dynamic analysis.
>>
>> At this point it is good to make a note of what test equipment you are familiar with that will provide an **amplitude**-swept, steady frequency signal.
>>
>> Inferences from a static AGC analysis and AGC induced IMD are apples and oranges.
>>
>> The second thing that bears heavily is the attack and decay speeds.
>>
>> Attack speeds are usually quick. If the attack and decay are **BOTH** quick, and that actual attack/decay is at an audio rate, then there is a case for distortion, because the variable gain can actually work at an audio rate.
>>
>> The question is whether the attack/decay cycle can continuously recur because the decay goes down as fast as the attack goes up, then intermod is indeed possible on a grand scale. However if the decay effectively holds the AGC gain level at a point set by the attack, delaying even as little as 100 milliseconds, then the AGC cannot create audio distortion products except very short low frequency distortion products only at AGC attacks.
>>
>> Since well before the significant AGC changes in firmware 4.7x (or whatever that one was), I have been running my slow AGC (CONFIG: AGC-S) at maximum fast, and my fast AGC (CONFIG: AGC-F) at maximum slow. In retrospect, that was probably why I never heard the stuff that a lot of people were complaining about.
>>
>> In contests I always use my max fast setting slow AGC, and back off the RF gain when I have primarily very loud signals in pile-ups to get the signals out of hardware AGC range, which has zero intelligent tweaks available.
>>
>> [And yes I have just about guaranteed pile-ups in contests with for-credit USA to USA QSO's, because of RBN spots which pick up everyone. Those are "spotting pile-ups" and assisted or unlimited class folks using point and click on the band map or control characters to move to the next unworked station.]
>>
>> My exception to using max fast setting slow AGC is when I'm trying to copy through lightning static, and need to hear weaker stations down in between the crashes. Then I use my max slow setting fast AGC.
>>
>> To summarize, in order for AGC to create audio distortion products strictly from the AGC, the AGC must be responding at an audio rate. Frankly, why would anyone want to set it that way escapes me.
>>
>> To Wayne, I would like to be able to set a minimum hold for fast AGC as well. That with a fast decay, would be better than what we have.
>>
>> Decay rate is something left over from analog days, when the way you decayed AGC was letting a capacitor discharge.
>>
>> 73, Guy K2AV
>>
>> On Fri, Mar 3, 2017 at 3:02 AM, David Gilbert <[hidden email] <mailto:[hidden email]>> wrote:
>>
>>
>> I've had my K3 since 2008 or so, and over the years I've seen
>> people describe different forms of "mush". One set of comments
>> indeed involved complaints about the hard limit at the upper end
>> that has nothing to do with AGC. It is, as you say, simply a hard
>> limit ... pretty much a clipper to protect the ears (and maybe
>> also to help protect the output stage in the speaker driver before
>> that issue got addressed). That creates a distortion, but it's
>> not really what I would describe as "mush."
>>
>> The nonlinearity I described in my earlier post was at the
>> opposite end of the curve ... down where the AGC just begins to
>> kick in. As W6LX says, it's a nonlinearity in the curve, and no
>> matter what you call it that contributes to the generation of
>> mixing products from multiple signals that happen to be at roughly
>> the same level within the passband. The low end of Jack Smith's
>> plots showed that pretty clearly. During some of my contest runs,
>> individual signals were perfectly clear and distinguishable, two
>> not terrible, but even three signals could generate enough mixing
>> products to cause problems if they were low enough in volume and
>> close enough in frequency. Since I typically operate with a very
>> narrow passband (about 150 HZ on CW), the mixing products end up
>> very close to the real signals. For example, 2x500Hz - 510 Hz
>> gives another phantom signal at 490 Hz. Things get really messy
>> with three or more signals.
>>
>> It is also, possible, of course, to get mixing anywhere there is a
>> knee in the AGC curve, but if you put the knee up higher there is
>> less likelihood that multiple signals will be of the same
>> amplitude to cause a problem (one will dominate), and their
>> amplitude swings will range further afield of the knee ... meaning
>> that a lower percentage of the energy will be mixed. At the low
>> end, you're pretty much screwed ... any signal you hear will be at
>> that nonlinearity and the amplitude swings will be small enough
>> that they spend all their time in the nonlinearity. As I said
>> before, reputedly the new synths greatly improve this.
>>
>> The bottom line is that if you have two or more signals within a
>> passband that traverse a nonlinearity, you get mixing products
>> within the same passband that blur the individual signals ...
>> i.e., "mush." And since the mixing products on CW only occur when
>> both (or more) of the signals are keyed, the mixing products
>> aren't even intelligible. ;)
>>
>> At least this is how I understand the situation. I'd be happy to
>> get corrected if my comments are flawed.
>>
>> 73,
>> Dave AB7E
>>
>>
>> On 3/2/2017 3:19 PM, Don Wilhelm wrote:
>>
>>
>> Now that you mention hard limiting, there is a limiter in the
>> K3 that if turned on will protect your ears. I am wondering
>> if some instances of reported receiver mush did have limiting
>> set on - that would be particularly true for those who chose
>> to ride the RF Gain and/or run with AGC off.
>>
>> 73,
>> Don W3FPR
>>
>> On 3/2/2017 3:37 PM, ab2tc wrote:
>>
>> Hi,
>>
>> Where in Smith's article does it say that AGC with the
>> slope set for 15 acts
>> as a hard limiter? There is a huge difference between AGC
>> action (which is
>> simply a reduction in gain with linearity retained) and
>> hard limiting.
>>
>>
>>
>> ______________________________________________________________
>> Elecraft mailing list
>> Home: http://mailman.qth.net/mailman/listinfo/elecraft
>> <http://mailman.qth.net/mailman/listinfo/elecraft>
>> Help: http://mailman.qth.net/mmfaq.htm
>> <http://mailman.qth.net/mmfaq.htm>
>> Post: mailto:[hidden email]
>> <mailto:[hidden email]>
>>
>> This list hosted by: http://www.qsl.net
>> Please help support this email list: http://www.qsl.net/donate.html
>> Message delivered to [hidden email] <mailto:[hidden email]>
>>
>>
>
>

Don Wilhelm

Re: RX Mush, why the fuss?

In reply to this post by ab2tc

The data for Jack Smith's (SK) graphs represents the THR and SLP before
the range of those parameters was increased. Fred Cady has graphs with
the more recent values. See
http://www.ke7x.com/home/miscellaneous-k3-information/agc-4-51.

73,
Don W3FPR

On 3/2/2017 3:37 PM, ab2tc wrote:
> Hi,
>
> Where in Smith's article does it say that AGC with the slope set for 15 acts
> as a hard limiter? There is a huge difference between AGC action (which is
> simply a reduction in gain with linearity retained) and hard limiting. I
> have read the article and agree that his measurements are very helpful.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]