Horrible Detector/AGC Circuits in Receivers *full docs added 10/26* see 1st post

[steve_qix]

Added 10/26/07 - Complete documentation on the low distortion detector/AGC,  including background information, overview, a detailed circuit description and receiver connection information (with schematics) can be found at www.classeradio.com/detector.pdf

Hi !

I recently dragged an old RME-45 receiver from my basement to use with the special events station at the recent NearFest.  The receiver works OK, so I hooked it up to an external hi-fi amplifier (tapped across the audio level control), and thought all would be OK.

Well, much to my dismay (at NearFest), I noticed what appeared to be distortion on otherwise good sounding signals.  Well, we muddled through with the receiver at NearFest, but when I got back, I hooked up the calibration transmitter, fed a 100% modulated triangle wave it, and tuned in the signal on the RME.

Well, the output was, as suspected, clipped at about 80% negative modulation.  And, on very low frequencies, the modulation could be seen on the AGC line.

After fooling around with various "fixes" to the original circuitry, I finally gave up and built an external detector/AGC amplifier for the thing.  BLISS !  No distortion at least to 99% negative and 200% positive modulation.  No modulation in the AGC, and the AGC works much better (more gain).

I have seen this sort of distortion on EVERY tube receiver I've ever worked with.  75A1, 2, 3 and 4, NC-303s, HQ170s, SP-600s, HRO50s and 60s - I've owned them all, and there was distortion on every one of them, caused by the detectors.

Anyway, here's the circuit I used.  Feel free to ask any questions.  It works VERY nicely.

[WA1GFZ]

The detector K2CU posted here a few years ago is the cleanest one I have ever seen.

[steve_qix]

The detector K2CU posted here a few years ago is the cleanest one I have ever seen.

Yes, I've see that circuit also.  Looks good, certainly for the 50kHz IF  for which it was designed.  I've tried the LM318 as a precision rectifier at 455kHz with varying results.  445kHz is nearing the useful edge of the LM318 gain/bandwidth product for LARGE signals.  It will work sufficiently as a unity gain amplifier at these frequencies.

The distinct advantage of the circuit I posted here is the ability to handle very large signals, highly modulated and with low distortion, and of course, to generate a suitable (and properly filtered) AGC for a tube receiver [this is VERY important].  If one uses the existing AGC - PARTICULARLY if the AGC is derrived from the audio detector (which many are) and/or if both the AGC and audio detectors are connected to the same IF, the IF stage in question will be loaded by the internal detector, causing distortion. 

This was the case in the RME-45, which necessitated a complete outboard of all detectors for both the AGC and audio.  But, even if there were a separate AGC amp and detector, it would most likely need modification to get the audio out of the AGC lines.   Some AGC systems (such as the one in the SP600) reverse bias the detector slightly, meaning that the rectified AGC is higher with modulation than at carrier.  VERY BAD - gives a compression effect within the receiver, as the signal is modulated.  Also, there is audio on the AGC line.  I saw the same thing (back biased AGC rectifier) on a receiver I worked on for Wayne, WA1SSJ, only I can't remember what the receiver was.

It absolutely amazes me that the original designers of these old receivers would USE such a bad detector/AGC system.  They certainly COULD have got it right, but they didn't even try.   

Oh well, at least we can fix their mistakes 

Regards,

Steve

[steve_qix]

Both look like LM318 to me?

Mack

Absolutely !  The 318 is a good IC in this application.   I have another circuit using the 318 as a precision half-wave rectifier.  I'll dig it up and post that one as well.  Noteably, this one performs identically and is slightly simpler.  I used the other one in a receiver MANY years ago - and it's still working somewhere - although I no longer own ther receiver 

[w3jn]

This is generally caused by too much capacitance in the detector and/or AVC inputs.  That RME has a pretty funky detector and AVC rectifier, and looking at the factory manual and the Sams schematic there are some differences between the two skizmatics.  I would say that the 1uF AVC filter cap is WAY too large, and the 47K diode load resistor (which is actually just prior to the volume control) is way too small.

Hallicrafters receivers have this problem in their ANL circuits, and the waveform is clipped regardless of whether the ANL is on or off.  Just clipping the .1 bias filter cap in the ANL will reduce the clipping threshold from 70% to about 90% on the negative peaks.  Of course this renders the ANL ineffective... but my point is that this issue can generally be vastly improved without resorting to adding many extra components.

[WA1GFZ]

My last HB hot rod receiver I played with a number of configurations and the best one I found was the 1/2 wave from rf design. That is where Rob started his design and came up with the new one. I think he states the 318 might have problems at high frequency IFs. You need the gain to compensate for diode knee.

[AB2EZ]

Steve

Excellent!

As a suggestion for consideration:

I built one of these  (external AM demodulator and AGC controller) for use with my 75A-4 some time ago. 

To derive the AGC control voltage, when receiving AM signals, I use a carrier level detector instead of a peak (envelope) detector. I.e., I rectified the i.f. output signal from the receiver with a rectifier that was fast enough to follow the i.f. frequency... and then I averaged (low pass filtered) the output of the rectfier. [Basically the same circuit as used in "precision rectifiers" / AM demodulators, except a much lower frequency cutoff on the low pass filter]


That way, the AGC responds to the carrier level... even if the speaker remains silent for a long time. This approach also produces a better low frequency response from the AM detector... without requiring the time constant of the AGC to be inordinately long.

Of course, I switch back to using a peak detector for AGC if I am trying to listen to an SSB or a CW signal.

Of note: on the 75A-4, there is a test point on the top of the chassis that provides a convenient place to insert the external AGC control signal. I just turn the "AVC" switch on the front panel to "off", and the external control signal takes over.

Best regards
Stu

[Steve - WB3HUZ]

Stu, you approach avoids the problem with many AGCs where low frequency modulation causes the AGC to react, thus creating IMD. It's amazing how nice the bass on some people's signals can sound on a good receiver or if you turn the AGC off on many receivers.

Good stuff.

[steve_qix]

Steve

Excellent!

As a suggestion for consideration:

I built one of these  (external AM demodulator and AGC controller) for use with my 75A-4 some time ago. 

To derive the AGC control voltage, when receiving AM signals, I use a carrier level detector instead of a peak (envelope) detector. I.e., I rectified the i.f. output signal from the receiver with a rectifier that was fast enough to follow the i.f. frequency... and then I averaged (low pass filtered) the output of the rectfier. [Basically the same circuit as used in "precision rectifiers" / AM demodulators, except a much lower frequency cutoff on the low pass filter]


That way, the AGC responds to the carrier level... even if the speaker remains silent for a long time. This approach also produces a better low frequency response from the AM detector... without requiring the time constant of the AGC to be inordinately long.

Of course, I switch back to using a peak detector for AGC if I am trying to listen to an SSB or a CW signal.

Of note: on the 75A-4, there is a test point on the top of the chassis that provides a convenient place to insert the external AGC control signal. I just turn the "AVC" switch on the front panel to "off", and the external control signal takes over.

Best regards
Stu

Hi Stu,

That is exactly what I'm doing in the circuit    The circuit on the AGC side takes the detector output (follows the IF, but does not store the audip peaks) and passes the output through a low pass filter, so there is no peak storage (or detection) happening.  If I wanted to add a product detector, I would have to put in a peak detector circuit, so the AGC would follow the audio and not the carrier.

This detector circuit sounds absolutely superb !  By the way, I put up a copy of the circuit using a precision rectifier, in case someone wants to use that instead of the original.  So far, it's the best detector I've used in any tube receiver.

I suppose with the addition of a product detector, it might be something to lay out on a PC board.

Regards,

Steve

[steve_qix]

This is generally caused by too much capacitance in the detector and/or AVC inputs.  That RME has a pretty funky detector and AVC rectifier, and looking at the factory manual and the Sams schematic there are some differences between the two skizmatics.  I would say that the 1uF AVC filter cap is WAY too large, and the 47K diode load resistor (which is actually just prior to the volume control) is way too small.

Hallicrafters receivers have this problem in their ANL circuits, and the waveform is clipped regardless of whether the ANL is on or off.  Just clipping the .1 bias filter cap in the ANL will reduce the clipping threshold from 70% to about 90% on the negative peaks.  Of course this renders the ANL ineffective... but my point is that this issue can generally be vastly improved without resorting to adding many extra components.

The RME detector is a bit odd, isn't it !  I was able to make some changes (the same ones you saw, actually !).  I changed the 47k diode to a higher volue, and reduced the 1uF to a .1.  This moved the distortion point up to around 90% or thereabouts, but it was still there - and there was some amount of non-linearity of the triangle wave, when this was detected.  I just could not get it to be distortion free without the external circuit.  I'm sure it could be done without the external circuit, but I didn't want to get into the complexities of figuring it out.  Also, this circuit leaves the receiver intact, for those who like original circuitry.  All I'd have to do would be to unsolder the external detector/AGC, and go back to the one in the receiver, and all would be original.

[WA1GFZ]

Yup, Stu's approach is the best. My problem was making the S meter respond to modulation peaks while taming the agc to not move on modulation as a peak hold. I have a variable hang on the AGC so didn't want to compress the modulation. I wanted to pump a little peak detected audio back into the s meter circuit but ran out of room in the module.

Drentea has his first article on his latest master piece in this issue of QEX. You know you have a strapping RX when you need fans to cool your internal modules.  Talk about raising the bar in RX performance
here you go.

[steve_qix]

Yup, Stu's approach is the best. My problem was making the S meter respond to modulation peaks while taming the agc to not move on modulation as a peak hold. I have a variable hang on the AGC so didn't want to compress the modulation. I wanted to pump a little peak detected audio back into the s meter circuit but ran out of room in the module.

Drentea has his first article on his latest master piece in this issue of QEX. You know you have a strapping RX when you need fans to cool your internal modules.  Talk about raising the bar in RX performance
here you go.

Hi Frank,

Yes - it's the best approach.  Stu and I are doing the same thing, by the way [in case there was any confusion].  For AM, under no circumstances would one want the AGC to follow or be affected by the audio.

Hey, why do you want the S meter to follow the audio?  I've found it helpful to *not* have the S meter follow the audio, so I can readily observe carrier shift on the received signal.   But, I suppose it would be an interesting feature to be able to select audio into the S meter - sort of a general, average modulation indicator...  Is that what you're doing?

Regards,

Steve

[WU2D]

The good old LM318 is a great building block because it has good DC specs, yet it handles high frequencies. You might try the nice op-amp offerings from ADI or Linear Technology for an upgrade. Example: LT1028
You can get these from Digikey for a few bucks.

Mike

[WA1GFZ]

Good question Steve, I never added the hardware to see the modulation peak because I ran out of board space but never considered carrier shift. I just assumed it would happen with high levels of positive modulation.
After looking at Drentea's article I could dust off the HB rx and go at it again. That would keep me from building a modulator though. Bad enough the HPSDR has my attention.
There are a number of super op amps out there. AD810 is another good one.

[AB2EZ]

Steve

Thanks for the follow-on comments.

When I looked at the original schematic that you posted, I must have missed the fact that the other half of the initial diode pair was rectifying the i.f., rather than peak detecting it.

Good stuff!

Best regards
Stu

[W1DAN]

Hi Steve and all:

Good posts and thanks for the schematic!

Some questions:

On the AC line in, there is a 1a fuse. I would think a 1/4a would work here?

On the input to the LM318, there is no DC blocking cap, would this be an advantage? Also would the input series resistor be low for a tube circuit?

On the detected output to the audio buffer transistor..would there be any advantage to AC coupling the signal there?

Any thoughts to doing full wave detection?

Years ago I solid stated an HRO receiver and used an infinite impedance detector using a single MPF-102 (in the ARRL HB). This is with lower RF voltages though...it was the best detector I have used and really enjoyed the quality of the audio. Did not load the previous circuit.

Your circuit is probably better though.

And I still use an LM-318 for my phono preamp!

Thanks for the thread!

Dan
W1DAN

[steve_qix]

Hi Dan,

Great questions (thoughtful).

Hi Steve and all:

Good posts and thanks for the schematic!

Some questions:

On the AC line in, there is a 1a fuse. I would think a 1/4a would work here?

Probably a 1/2 amp fuse would work - maybe even a 1/4 amp, although it would have to be a slow blow.  That would probably be the way to go.

On the input to the LM318, there is no DC blocking cap, would this be an advantage? Also would the input series resistor be low for a tube circuit?

In the case of the RME-45, no blocking cap was needed because the IF output is ground referenced (I grounded one side of the secondary of the IF transformer).  However, if DC is involved, a blocking cap would be a must !!!  If a blocking cap is used, an RF choke should be employed to maintain a DC ground at the op-amp input.  A resistor would load the IF too much.

The impedance of the non-inverting input of the LM318 is really high (multi megohms), so there is virtually no load on the IF.  that's nice !!  The 47k series resistor is just there to prevent the LM318 from taking off, which it will do if steps are not taken to prevent it.

On the detected output to the audio buffer transistor..would there be any advantage to AC coupling the signal there?

Hmm  I can't think of any, and by AC coupling to the buffer, additional resistors would be necessary to set the buffer's operating point.  It just seemed easier to DC couple the buffer, and do the DC blocking after buffering.  Also, this method reduces the component count.

Any thoughts to doing full wave detection?

It would make the output easier to filter, but beyond that, there is no advantage.  At 445kHz, it's easy to filter the IF out and still have a really good high end (over 20kHz), but at 50kHz (the IF of some receivers), a full-wave system would be an advantage, and I would use it.

Years ago I solid stated an HRO receiver and used an infinite impedance detector using a single MPF-102 (in the ARRL HB). This is with lower RF voltages though...it was the best detector I have used and really enjoyed the quality of the audio. Did not load the previous circuit.

Your circuit is probably better though.

You'd have to test it with a triangle wave to verify the linearity.  It might be OK !  I have no experience with infinite impedance detectors, so can't make any suggestions here.  It would be worth checking out with a good modulated source.

And I still use an LM-318 for my phono preamp!

Thanks for the thread!

Dan
W1DAN

Well, hopefully it will be useful to someone who has the same problem I did - a good receiver with a BAD detector/AGC :-)

Talk later and Regards,


Steve

[N3DRB The Derb]

so after all the class e and d tx stuff, which one of youse guys is gonna get around to making an entire SS rx?

[steve_qix]

so after all the class e and d tx stuff, which one of youse guys is gonna get around to making an entire SS rx?

Done it (4 times).  I think Frank 'GFZ has also.  I solid-stated 2 BC-1004 receivers (one of them is my main station receiver, the other is in the kitchen), a National (I forget the model) and a TransOceanic.  The BC-1004s came out VERY well.

Solid-stating a tube receiver is basically a complete redesign.  It's a major project - harder in many ways than any transmitter.

Frank is the RX guru here.  I'm more of a hacker

Regards,

Steve

[WA1GFZ]

The all time best way to modify an RX, Design an RX or solid state one is start with a good speaker and work your way to the antenna. This way you can test each stage and integrate new ones as you go.
I've built 3 upconversion receivers in the past 20 years and getting the itch to do it again after looking at Drentea's beautiful design.
Funny thing I'm always collecting parts and have a lot of the stuff he used.
There is no cure for Receiver disease

[W1DAN]

Hi:

My SS HRO is what I considered a true SS receiver. Learned alot by doing it. The detector was the best part(can't think of anything missing in it's performance), but I had problems getting my LO to track well.

73,
Dan

[WA1GFZ]

I would love to see someone take a clean R390A and yank out the 6c4 mixers and install 7360s. I bet it would turn nice times in a quarter.

[Rob K2CU]

Almost all of the old tune receivers relied on the logarithmic I/V curve of a diode, either hollow or solid state to act as a "square law" detector, that being the major component of the series expansion of a log function. This basically worked by multiplying the signal by itself. Problem was that even though you got detection, the upper and lower sidebands would produce high levels of second order harmonic distortion. Single sideband AM detection would always sound clearer when you slid the signal to one side.  Poorly filtered AGC voltages result in power supply hum, or low frequency audio components getting on the AGC line and modulating the incoming signal. YIKES!

Steve's deign is fine and will make almost any radio sing much clearer. Just be sure to match the AGC voltage out vs. IF input level in to that of the original radio to not mess up the AGC characteristics of the radio.