EMRFD Message Archive 9015
Message Date From Subject 9015 2013-08-27 18:24:52 AD7ZU AGC questions
Now for a couple questions.. since most of the historical AGC systems have used an RC timing mechanism for the hold state we are all used to hearing an exponential decay in the AGC over some fraction of a second or what ever is deemed useful for accommodating pauses in speech.
Is a exponential response ideal for a SSB AGC?
Since the control function is implemented in software the response can be anything and multiple responses can be designed and selected depending on the mode of operation. Most of the DC QRP radios have not implemented AGC. Some just use diode limiting to prevent ear injury. I have used radios with AGC for SSB (fast attack slow decay). but what AGC is most appropriate for reasonable speed CW .. say 20wpm??
The ADC and control computation could easily monitor the signal level at 500us intervals and the digital pot response is very fast. ..so are there any thoughts on an ideal AGC response for various common amateur modes .. SSB, CW, digital (psk31 or other). It should also be viable to implement a noise blanker using the same mechanism.
Any thoughts??
[Non-text portions of this message have been removed]9016 2013-08-28 16:30:34 Dean Blake Re: AGC questions
audio in a very unpredictable and rapidly changing HF propagation environment.. so why
do we tie the attack and decay of the AGC circuit to some time constant based on the
demodulated baseband.. and not more on ionospheric fade times? after all isn't it the
average signal level over mostly seconds of time that we are trying to average...
My point... listen to the HF bands on a good receiver that has the ability to shut off AGC..
I've done this many times (as a shortwave listener even before I was a ham) the signal
level and hence mostly audio level do long ebb and peak cycles of seconds to tens of seconds
that tells me the attack time can still be fairly fast (to limit 'blasting') and the decay time could
successfully be over 100's of milliseconds. Even long enough to cover whole words and keep the
gain of the receiver more constant... hence a smoother audio response..
Same with CW... working contests and general rag chews I see that a similar AGC response
here would work fine... in fact maybe better...
The signal level at the receiver ( regardless of modulation type) is based on the ionospheric
propagation, not the type of modulation.. so why does the AGC circuit get all hung up on the
type of received modulation...
I always thought the best approach is when receivers convert the decoded modulation to
digital... (DSP) here lies the real control over the audio (or receiver gain) which can be keep
very constant using a digital processor.. that's a no brainer! But what is the best approach (or compromise)
for a good old fashion HF analog receiver??
K4DSB
Dean
9017 2013-08-28 18:42:54 Ford Peterson Re: AGC questions
I'm watching this thread with great interest because I'm in search of
solutions for AGC. The N3ZI receiver I've been playing with has no AGC and
I've been playing with different methods of deriving AGC for it. None that
I like enough to brag about yet for some of the very observations you cite.
http://n0fp.com/2011/01/29/n3zi-dds-2-vfo/
The lack of AGC will just about throw you out of the chair when listening to
weak signals. You crank up the audio gain to listen and then a local will
key up and blast your ears. The fading can be several seconds whereas the
attack must be immediate for typical use. I've always preferred fast AGC
when I have an instrument that makes it variable. Slow can be nice in
certain circumstances but fast is universally usable when the attack time
will deal with the blasting. Dealing with the AGC 100% in the audio stages
is not exactly high performance. With a simple rig like the N3ZI, quite a
bit of the total gain in the signal path happens at audio. It's better to
scatter the gain control throughout the IF chain. Most of the devices used
to do that have become expensive. At least that's what my reading and
experimentation is telling me.
73
Ford-N0FP
-----Original Message-----
9018 2013-08-29 09:47:34 William Re: AGC questions
I love the #12 black wire insulation you use for the home made bezel surrounding the readout.
Cool.
Bill N7EU9019 2013-08-29 10:08:01 Paul Daulton Re: AGC questions
http://www.onsemi.com/PowerSolutions/product.do?id=NCP2824
interesting device maybe someone could come up with a "surfboard" to
experiment with this device/
Paul k5wms
Reply via web post Reply to sender Reply to group Start a New
Topic Messages in this topic (4)
Recent Activity:
* New Members 4
* New Files 1
Visit Your Group Switch to: Text-Only, Daily Digest • Unsubscribe
• Terms of Use • Send us Feedback .
Paul Daulton K5WMS
beacon WMS 185.302 khz qrss30/slow 24/7
Jacksonville,Ar 72076
em34wu
[Non-text portions of this message have been removed]9020 2013-08-29 10:27:52 William Carver Re: AGC questions
> http://www.onsemi.com/PowerSolutions/product.do?id=NCP2824
>
> interesting device maybe someone could come up with a "surfboard" to
> experiment with this device/
>
> Paul k5wms
device in typical application".9021 2013-08-29 14:56:31 Ford Peterson Re: AGC questions
This is yet another method of AGC at the audio stage. I've never seen AGC done solely at audio that has worked well. Most of the old designs use RF attenuation throughout the IF chain. I've been looking for a device that does the AGC action at RF rather than audio. The devices used in past designs are almost universally past final run and on vapors. The proverbial unobtainium. While there are devices that have a good dynamic range, they are also big bucks. Good dynamic range, high IP3, and cheap. Pick any two appears to be the name of the game.
Maybe the device is the holy grail, and maybe there is a better way to do it.
Ford-N0FP
-----Original Message-----
9022 2013-08-29 15:06:30 William Carver Re: AGC questions
are available, or were available, from Roger Hayward. It used J310
JFETs, so obtainable and inexpensive.
W7AAZ
On Thu, 2013-08-29 at 16:43 -0500, Ford Peterson wrote:
>
> While I should never say 'never...'
>
> This is yet another method of AGC at the audio stage. I've never seen
> AGC done solely at audio that has worked well. Most of the old designs
> use RF attenuation throughout the IF chain. I've been looking for a
> device that does the AGC action at RF rather than audio. The devices
> used in past designs are almost universally past final run and on
> vapors. The proverbial unobtainium. While there are devices that have
> a good dynamic range, they are also big bucks. Good dynamic range,
> high IP3, and cheap. Pick any two appears to be the name of the game.
>
> Maybe the device is the holy grail, and maybe there is a better way to
> do it.
>
> Ford-N0FP
>
> -----Original Message-----
>9023 2013-08-29 15:13:01 Dean Blake Re: AGC questions
the R.F amplifier stage (if it had one, many started out going into the first converter)...
Your right, you need a fairly good dynamic range as signals can very over 50db
PS...I built a Bitx20a transceiver kit and I added 'audio AGC' not fabulous, but works
pretty good compared to the 'no AGC' in the original kit...
K4DSB
Dean
To: emrfd@yahoogroups.com
From: ford@highmarks.com
Date: Thu, 29 Aug 2013 16:43:56 -0500
Subject: RE: [emrfd] Re: AGC questions
While I should never say 'never...'
This is yet another method of AGC at the audio stage. I've never seen AGC done solely at audio that has worked well. Most of the old designs use RF attenuation throughout the IF chain. I've been looking for a device that does the AGC action at RF rather than audio. The devices used in past designs are almost universally past final run and on vapors. The proverbial unobtainium. While there are devices that have a good dynamic range, they are also big bucks. Good dynamic range, high IP3, and cheap. Pick any two appears to be the name of the game.
Maybe the device is the holy grail, and maybe there is a better way to do it.
Ford-N0FP
-----Original Message-----
9024 2013-08-29 18:42:07 Ford Peterson Re: AGC questions
Audio Derived AGC
"...But there is a major difficulty with audio derived AGC. This relates to the sampling nature of the detection process. The detectors we have examined obtain one sample for each peak of the waveform being detected. Audio waveforms have fewer peaks, especially if the signal is a low-pitched CW carrier. This allows the receiver to be overwhelmed in the period between peaks."
They then proceed to discuss some compromises or work-arounds.
I think I found a sample circuit at Fig 6.50 I'll have to ponder. The double jfet eliminates the need for the dual-gate variety, which are quickly becoming unobtainium as well. I may just have to build one of these around my crystal filter after C9 and b4 Q3 to see how it works.
Ford-N0FP
-----Original Message-----
9025 2013-08-29 19:06:20 Dana Myers Re: AGC questions 9026 2013-08-29 19:28:04 AD7ZU Re: AGC questions
Some observations referring to IF based AGC systems:
1. sensing the signal level must be done post filtering, else the out of band signals will result in an errant AGC control level.
2. Optimal attack, hold, and release times may not all be easily implemented without separate control loops. The example in EMRFD is a good design, however it is complex when compared to sensing the AGC level in a ucontroller then calculating the ideal response based9027 2013-08-29 19:51:20 Ashhar Farhan Re: AGC questions
I tried using arduino to control gain. I guess like in most otber
things, I didn't try hard enough. It didn't work well enough. Here is
what I learnt :
1. The agc detection had to be done outside the uP. That is, the
envelope detector and the filter had to be in analog circuitry. The uP
didn't have juice enough to sample the envelope quickly enough to
figure out when to attack.
2. Arduino lacked analog output. Using PWM put out a buzz into the audio line.
I guess it can still be done. But the lowly 8-bit controllers like pic
and avr may not be the candidates.
I am excited by the potential of raspberry pi though. A 1 GHz cpu with
DSP co-processor, running full Linux, the size of a credit card and
costing less than 25 dollars is full of potential for radio builder.
Watch this space closely...
- farhan
On 8/30/13, AD7ZU <ad7zu@yahoo.com> wrote:
> My original questions were concerning the use of an ADC into a ucontroller
> to measure the signal level then apply a correction to a digital
> potentiometer as a means to control the signal level. My design is a DC
> receiver so there is no IF and the filtering and phase shift are done
> digitally at baseband without any gain in those sections.
>
> Some observations referring to IF based AGC systems:
> 1. sensing the signal level must be done post filtering, else the out of
> band signals will result in an errant AGC control level.
>
> 2. Optimal attack, hold, and release times may not all be easily
> implemented without separate control loops. The example in EMRFD is a good
> design, however it is complex when compared to sensing the AGC level in a
> ucontroller then calculating the ideal response based9028 2013-08-29 20:43:05 kb1gmx Re: AGC questions
IT can be done digitally, some cpu power is needed but not as much as some speculated. DSPIC32 can do it and well enough. However it's still a servo system and the dynamics are now program controlled
to a point.
Dual gate fets are still around and in production. BF998 is an excellent part. The only problem is it's surface mount though more than few have made ways to mount and bring it out to leads.
They are cheap parts to the old leaded parts.
However the AGC system can be done and well with anything.
It's just as effective using MMICs and PIN diodes!
The HYCAS board and project was one of the best examples of a
wide application IF with AGC. Its biggest and often missed
features were no exotica or unobtainium and enough gain with
enough gain control range.
One of the biggest problems with Audio AGC is not enough samples
due to the audio itself. IF AGC system have the problem of BFO
getting into them and making pushing the AGC. Elecraft soved that
in the K2 by converting the IF to about 150khz, high enough to get plenty of samples for rectification and not on the IF frequency to
get BFO into it and hand it up. It also put a bit more gain in the AGC loop to help with detection.
Allison
9029 2013-08-29 21:02:58 Ford Peterson Re: AGC questions
The raspberry pi is almost a cripple by comparison to some of the newer
stuff. Take a look at the BeagleBone Black by TI. Texas Instruments is
pushing development tools for free and will likely surpass the Pi's
installed base soon if it hasn't already. Considering much of the Pi's
installed base is school kids, it makes me wonder how long it will survive
outside of that educational environment. The BeagleBone Black has got 3
complete 8 bit ports, SPI and I2C ports including several analog in/out
ports. Everything is accessible using their "capes," which are
daughter-boards. $45 complete.
Ford-N0FP
-----Original Message-----
9030 2013-08-29 21:03:59 AD7ZU Re: AGC questions
thanks .. good comments
that is almost exactly the implementation i have in mind .. albiet im a bit bummed you didnt find it a good solution.
I am using an 8bit AVR 16mhz with on-chip ADC but am using the SPI to control a digital pot as an attenuator in lieu of an analog output. There are other digital pots that use an up and down count to control the tap position that may operate faster than the SPI, however they only move one position per rising edge .. so the controller must take the pot through each intermediate postion...and also keep track of the last position thats why i initially chose the SPI version.
.. yes i have seen the raspberry PI ...and some other boards out there now the processing horsepower is astounding for the size, cost, and power consumption.
________________________________
9031 2013-08-29 21:48:34 William Carver Re: AGC questions
that doesn't overshoot, oscillate, or have other aberrations is not a
trivial design task. I don't know about today's education, but EEs had
to take at least one class in "control system theory" and learn how to
design stable servo (AGC) systems.
DSP filtering inherently delays the filtered output signal. A faster CPU
doesn't make the DSP filter run faster: you need to sample the signal
over a period of time. AGC computations based on that filtered output
are therefore also delayed and"delay inherently makes feedback
systems less stable". DSP AGC is tricky.
AGC is amplitude modulation. So residual ripple on a PWM-generated
analog AGC signal will modulate the incoming signal(s) and add AM
sidebands.
The ARRL Laboratory tests receivers for IMD by applying a two tone
signal. I don't remember the exact spacing, I think it was 100 Hz. The
AGC is specified as "set to the fastest position".
Lets say your receiver has a +60 dBm intercept, super linear analog
circuits, etc. The circuits don't generate detectable IMD with those two
tones.
The envelope of the two tone signal is varying at a 100 Hz rate. The AGC
circuits or DSP calculations have to suppress that 100 Hz component or
it will modulate the incoming signals and produce 100 Hz AM sidebands on
the signals. Those sidebands at PRECISELY the same frequencies where IMD
is: AGC produces the same result in the audio as IMD in early receiver
stages.
A 100 Hz envelope has a period of 10 milliseconds. You need to filter
the AGC so 100Hz is down 40 dB for the ARRL two-tone test. You want to
achieve a 3mS "attack" time on the leading edge of a CW signal so your
AGC doesn't make perfectly good CW sound "clicky". Those are conflicting
requirements. Agh! AGC is NOT like falling off a log. Don't get
discouraged too quickly: like spotting PWM buzz-modulation, you have to
track down how every aberration is generated and squash them just like
any other RX bug.
Bill W7AAZ9032 2013-08-30 09:30:10 Dan Mills Re: AGC questions
envelope is taken to be sqrt (I^2 + Q^2), you still have a loop dynamics
headache but now at least your notion of signal level is no longer
dependent on exact timing relative to the input signal.
This is especially nice in a DC radio using the common quadrature switching
mixer, as you have an I/Q pair right out the gate, sample both and do the
work in analytic space, otherwise google 'Hilbert transform'.
AGC is inherently amplitude modulation, but there is a trick in the digital
domain... Delay your audio after the AGC detector such that the control
signal after a gaussian filter set low enough to remove the ripple still
arrives at the control stage in time to do the job. Delay in the digital
domain is easy to implement, and being able to see a few ms into the
'future' turns out to be all sorts of useful.
I also commend the notion of 'hang' mode agc to folks attention, again easy
on a fairly modest micro, and trivial on any modern DSP.
One thing I have found useful is to not do AGC as such, but rather to have
the volume control adjust both the volume and a post volume control limiter
threshold, that way someone 40dB or 50dB up on what you are listening for
does not cause hearing damage as they obliterate your dx.
Finally, I would note that different detection strategies are needed for
SSB, CW and AM, the same systems are not optimal for all of them.
73 Dan M0HCN.
[Non-text portions of this message have been removed]9033 2013-08-30 09:54:55 Ashhar Farhan Re: AGC questions
thanks for pointing out the qaudrature method for agc. i am continually
amazed at how much I/Q can achieve. i must chew on this for a bit now...
- farhan
9034 2013-08-30 10:07:12 blumu Re: AGC questions
by N6WA, published in 73 Magazine, September 1979 ?
I breadboarded this in the 1980s. Although an audio-only approach,
the constant-level output could doubtless also be used to control RF/IF gain.
I didn't try that, not least as I was seriously impressed with
its audio performance.
Michael
----- Original Message -----
9035 2013-08-30 10:53:33 William Carver Re: AGC questions
>Good point: you want to know signal power as soon as possible and as
> The answer to the sampling thing is to do quadrature sampling then the
> envelope is taken to be sqrt (I^2 + Q^2), you still have a loop
> dynamics headache but now at least your notion of signal level is no
> longer dependent on exact timing relative to the input signal.
early as possible.
> This is especially nice in a DC radio using the common quadratureYes, you get knowledge of amplitude at the earliest possible time, but
> switching mixer, as you have an I/Q pair right out the gate, sample
> both and do the work in analytic space, otherwise google 'Hilbert
> transform'.
there is NO filtering at that point. You will get the power in very
broad band. Signal, QRM, and wideband noises like lightning strokes or
light switch transients would most likely dominate the AGC scheme.
> AGC is inherently amplitude modulation, but there is a trick in theTo the extent you can perform ALL AGC in the digital domain you do have
> digital domain... Delay your audio after the AGC detector such that
> the control signal after a gaussian filter set low enough to remove
> the ripple still arrives at the control stage in time to do the job.
> Delay in the digital domain is easy to implement, and being able to
> see a few ms into the 'future' turns out to be all sorts of useful.
time to do nice things like that. But all digital radios are still not
up to performance of mixed analog/digital radios. As the analog gain
variation becomes a smaller piece of the pie its getting better, but the
dynamics even of all digital radios isn't impressive.
And....NOTHING is free. Being a QSK CW guy, the DSP delay is very
irritating. The K3 audio has DSP delay, but the much bigger delay in my
QS1R makes it essentially useless for making CW QSOs.
> I also commend the notion of 'hang' mode agc to folks attention, againAnd the release time is under your control. Although a perfectly flat
> easy on a fairly modest micro, and trivial on any modern DSP.
hang isn't very good, the decay rate while in the hang state can be made
low enough that IMD generation is almost eliminated. PA3AKE has inband
IMD 60 dB below the two tone test signals.
> One thing I have found useful is to not do AGC as such, but rather toYES!!!
> have the volume control adjust both the volume and a post volume
> control limiter threshold, that way someone 40dB or 50dB up on what
> you are listening for does not cause hearing damage as they obliterate
> your dx.
> Finally, I would note that different detection strategies are needed73 Bill W7AAZ
> for SSB, CW and AM, the same systems are not optimal for all of them.
>
> 73 Dan M0HCN.
>
> [Non-text portions of this message have been removed]
>
>
>
>
>9037 2013-08-30 12:16:31 Dan Mills Re: AGC questions 9038 2013-08-30 13:09:56 blumu Re: AGC questions
. . .
And....NOTHING is free. Being a QSK CW guy, the DSP delay is very
irritating. The K3 audio has DSP delay, but the much bigger delay in my
QS1R makes it essentially useless for making CW QSOs.
. . .
> One thing I have found useful is to not do AGC as such, but rather toYES!!!
> have the volume control adjust both the volume and a post volume
> control limiter threshold, that way someone 40dB or 50dB up on what
> you are listening for does not cause hearing damage as they obliterate
> your dx.
> Finally, I would note that different detection strategies are needed73 Bill W7AAZ
> for SSB, CW and AM, the same systems are not optimal for all of them.
>
> 73 Dan M0HCN.
>
---------------------------------------------
This kinda cosies up with the KISS thoughts at the back of my mind :-
"How do these options compare with the "audio elixir"
"by N6WA, published in 73 Magazine, September 1979 ?
"I breadboarded this in the 1980s. Although an audio-only approach,
"the constant-level output could doubtless also be used to control RF/IF gain.
"I didn't try that, not least as I was seriously impressed with its audio performance."
Any observations ?
Michael
UK9039 2013-08-30 16:14:11 Ford Peterson Re: AGC questions
This kinda cosies up with the KISS thoughts at the back of my mind :-
"How do these options compare with the "audio elixir"
"by N6WA, published in 73 Magazine, September 1979 ?
"I breadboarded this in the 1980s. Although an audio-only approach,
"the constant-level output could doubtless also be used to control RF/IF gain.
"I didn't try that, not least as I was seriously impressed with its audio performance."
Any observations ?
Michael
UK
...unsnip...
For those of us that were children in 1979, and not subscribers to 73 at the time or forever thereafter, can you provide a link to the article? Or a scan of same?
Thanks
Ford-N0FP9042 2013-08-30 16:47:58 NeilDouglas Re: AGC questions
http://ia700807.us.archive.org/18/items/73-magazine-1979-09/09_September_197
9.pdf
NeilD
G4SHJ
_____
9043 2013-08-30 16:54:39 NeilDouglas Re: AGC questions
http://ia700807.us.archive.org/18/items/73-magazine-1979-09/09_September_197
9.pdf
Looks like there is an error in Fig 1, the feedback to the gate of Q1 should
not be connected to the 10K / 10K bias supply.
NeilD
G4SHJ
_____
9044 2013-08-30 17:23:01 William Carver Re: AGC questions
seeming failure of an AGC system may well be aberrations from feedback
rather than attributable to the detector or gain-controlled stages
themselves, and one might not need to throw out the baby, just the loop
filtering.
Of course there is delay in analog filters, but nothing close to what
I'm seeing with the direct sampling QS1R. It decimates samples with an
FPGA before going to the PC: can most of the delay blame can be placed
on the PC software, or is it the PC environment? I should try running
one of the Linux programs to see if there's a substantial difference.
One could do a feed-forward AGC with analog circuits or digital
sampling, eliminating feedback and all its compromises. But they only
work if a signal-bandwidth filter precedes them which clobbers the
wideband seduction of the sampling receiver.
For my purposes the direct sampling receiver as a panadaptor, operating
a huge LCD screen, and a conventional narrowband hybrid analog/digital
receiver are a good combination...although perhaps a bit overboard.
The QS1R noise figure is a bit too high, and the delay intolerable, so I
can't discard the narrowband RX. The spectral display is habit forming,
making it impossible to discard the QS1R.
Bill W7AAZ9045 2013-08-30 17:48:44 William Carver Re: AGC questions
Looks like there is an error in Fig 1, the feedback to the gate of Q1
should not be connected to the 10K / 10K bias supply.
Yes, the draftsman put a "dot" connecting the gate of Q1, tsk tsk. Just
erase the dot on the gate lead.
W7AAZ9047 2013-08-30 20:51:32 AD7ZU Re: AGC questions
Thanks everyone for the continuing education.
The concept of a feedforward I/Q implementation ala Dan’s concept is appealing on several levels. In my current design this could be implemented but I would need to add a DAC…I think? I looked at this idea for a few minutes over lunch today. The thought occurred to me that if I divided the FIR/Phase shift filter into 2 cascaded parts say a 75%/25% split. The phase shift and stop band of the first portion of the filters would be adequate for the AGC. The delay through the remaining 25% of the FIR allows sufficient time for the AGC acti9048 2013-08-30 21:31:37 William Carver Re: AGC questions
always kept ahead of the output. I don't think you'd need parallel
AGC/signal filters, just a matter of doing the AGC multiplication on
data that slightly precedes the data which generated the AGC itself.
W7AAZ
On Fri, 2013-08-30 at 18:33 -0700, AD7ZU wrote:
>
> I can only say I’ve learned quite a lot reading a few posts from this
> group on the AGC topic.
> Thanks everyone for the continuing education.
>
> The concept of a feedforward I/Q implementation ala Dan’s concept is
> appealing on several levels. In my current design this could be
> implemented but I would need to add a DAC…I think? I looked at this
> idea for a few minutes over lunch today. The thought occurred to me
> that if I divided the FIR/Phase shift filter into 2 cascaded parts say
> a 75%/25% split. The phase shift and stop band of the first portion
> of the filters would be adequate for the AGC. The delay through the
> remaining 25% of the FIR allows sufficient time for the AGC acti9050 2013-08-31 09:31:56 AD7ZU Re: AGC questions
Bill,
Its not a parallel filter just a tap in the total fir..
the feedforward idea cleans up a lot of issues
its a relatively simple approach and relieves the tight latency required for a fast attack time.
This thread is very enlightening.
-- Randy
9053 2013-08-31 09:40:49 William Carver Re: AGC questions
Sounds good. I can do microcontrollers in my sleep, but need to get
better at software.
Bill
On Fri, 2013-08-30 at 22:04 -0700, AD7ZU wrote:
>
>
> Bill,
>
> Its not a parallel filter just a tap in the total fir..
> the feedforward idea cleans up a lot of issues
> its a relatively simple approach and relieves the tight latency
> required for a fast attack time.
>
> This thread is very enlightening.
>
> -- Randy
>
>9055 2013-08-31 13:32:02 Alberto I2PHD Re: AGC questions