How much 455KHz gain is too much?

[Radio Wrangler]

That's the one. There are at least two editions.

David

[G6Tanuki]

My "Rolls Royce" solution would be for 3 IF stages using valves such as the 6BA6 or later improved versions.

This will of course give you _lots_ of IF gain - but rather than 'damping' the IFTs with resistors [which potentially introduce noise as well as reducing the 'Q' of the IFTs] I would either see if it was possible to reduce the coupling between the IFT tuned-circuits [this is easy with some IFTs as used by Pye/Philips in the WWII days] or alternatively use pairs of IFTs back-to-back with a low-value [5pF] coupling capacitor between the back-to-back windings so you essentially have a quadruple-tuned IFT - that way you can achieve something less than 'critical' coupling, maximising Q and sharpness of tuning at the expense of a few dB of gain.

Then feed the resulting IF into a _low impedance_ double-diode product-detector; the point about being low-impedance is that it means the thing is working largely on current rather than voltage so capacitive leakage of the CIO back into the front of the IF strip is reduced.

There are potentially advantages in using a push-pull balanced product-detector; I remember seeing something along these lines, using "Gated Beam" pentodes, in a commercial/military RTTY demodulator [where the budget allowed for two independent demodulators, one for the 'high' tone and one for the 'low' tone of the FSK]

As to AGC, the big problem with this for _any_ CW receiver is that, averaged over half a second or so, the received and integrated 'carrier' level varies both with the quality of the received signal _and_ the composite of the onff ratio of the Morse elements [manual CW being a non-return-to-zero form of digital modulation]. Which makes using 'average' received carrier-level a bit of a fools errand which falls apart if the 'CW' includes lots of strings of repeated dots or dashes.

What about forgetting the idea of AGC, and instead running the IF-strip flat-out [crossmodulation permitting], then interposing a FM-style limiter between the output of the IF strip and the product-detector. Slice off all the amplitude-variations and present a square-wave to the product-detector! You then only need to differentiate between 'carrier present' and 'carrier absent'.

[Alan_G3XAQ]

Originally Posted by G6Tanuki

Which makes using 'average' received carrier-level a bit of a fools errand which falls apart if the 'CW' includes lots of strings of repeated dots or dashes.

Yes, I agree. The trick is to use the peak received carrier level to rapidly charge the AGC capacitor but then discharge the capacitor with a long time constant. Collins suggested 50ms attack and 500ms or more decay. I suspect the attack time constant mustn't be too short for fear of the whole gain loop oscillating. Someone who is deep into control theory could doubtless design this from scratch but my inclination would be to use a pot and twiddle it for the best compromise between rapid attack and AGC stability.

Three 6BA6/EF93 IF stages has been used before so I think it's at least worth a try. If I go for IF-derived AGC I'll need all the 120dB gain on offer (and I'll need to copy the Collins approach of a potential divider on the screen supply instead of just a series resistor if I am to see the 30dB AGC range per stage that I will need).

I remain unsure about the product detector and BFO leakage. The proliferation of designs suggests there is no one perfect solution. The classic 75S3 receiver just uses a single triode with 100:1 attenuation to give 75mV on the IF input to the grid and about 4v p-p BFO injection to the cathode. A simplistic simulation suggests most of the BFO signal appears at the grid but if there is adequate screening maybe the 4/100 = 40mV that escapes into the IF isn't a problem? I think experimentation is needed.

73, Alan G3XAQ

[Radio Wrangler]

An oscillator needs an amplitude stabilising mechanism to control its amplitude. This means that short of a detector based control loop, they all rely on active devices being driven either into serious non-linearity or cutoff. Consequently, all intentional oscillators give big oscillations. If their frequency is a problem, then you need attenuation....but harder still, you need serious screening to go with the attenuation.

David

[G6Tanuki]

Not to denigrate the Collins-guys bur i guess they were still thinking very much from the dual-sideband-and-full-carrier AM days; why 'do' lots of IF gain if you only go to attenuate the resultant signal before feeding it into the product-detector?

[The Trio-Kenwood 9R-59 series receivers popular in the 60s and 70s also had a capacitive-potential-divider at the input to their product-detectors - I guess they were still stuck in the broadcast-band-listening mindset]

Fast-attack, slow-decay was the thing both for CW and SSB AGC in times-past; the big failing being that a noise-spike [whether from a lightning-strike, a thermostat-switching or the pulses of ther "Russian Woodpecker' or the "Commie Kookaburra" as nicknnamed by those in the Southern Hemisphere whose SW receiving-experience was blighted by it] could 'stuff' enough charge into your AGC that your receiver was desensitized-to-the-point-of-uselessness for over half a second.

Clipping, rather than desperately trying to follow the supposed recovered- signal-envelope of a CW signal-plus-noise, always seemed much more sensible to me.

Have you considered another frequency-shiift before the product-detector, separate from the AGC-loop? An ECH42 or ECH81 downconverting to a seriously-low IF [anything from 25 to 100KHz] would sidestep the whole BFO-leakage-into-the-IF-strip issue.

Whatever you do, avoid the likes of the 6AT6 dual-diode-and-triode; the design of these introduces lots of nasty couplng-at-RF between the two diode-plates and the triode anode/grid - meaning you're going to get 'inadvertent' diode-action going on where you really don't want it!

[Alan_G3XAQ]

Originally Posted by G6Tanuki

Not to denigrate the Collins-guys bur i guess they were still thinking very much from the dual-sideband-and-full-carrier AM days; why 'do' lots of IF gain if you only go to attenuate the resultant signal before feeding it into the product-detector?

Lots of gain is needed somewhere to produce the -20V AGC needed to control variable mu valves. The techniques for generating this directly from the IF output was/is "straightforward", but yes I agree it's somewhat a sledgehammer approach. The DC amplifier you might use today with an opamp wasn't in the Collins toolkit.

Originally Posted by G6Tanuki

Fast-attack, slow-decay was the thing both for CW and SSB AGC in times-past

Isn't it still in use today, albeit in software with SDR receivers?

Originally Posted by G6Tanuki

the big failing being that a noise-spike [whether from a lightning-strike, a thermostat-switching or the pulses of ther "Russian Woodpecker' or the "Commie Kookaburra" as nicknnamed by those in the Southern Hemisphere whose SW receiving-experience was blighted by it] could 'stuff' enough charge into your AGC that your receiver was desensitized-to-the-point-of-uselessness for over half a second.

Woodpecker type over the horizon radars have long (modulated) pulses that are hard for a simple receiver to distinguish from Morse so I'll just have to live with them. Their pulse trains go on for ages, much longer than the half a second it takes for the receiver to recover when they stop.

More sophisticated receivers than I want to build use a "noise blanker" ahead of the narrow IF filter to reduce the impact of short noise spikes. My Elecraft K3 has one but I rarely use it because the spikes from the local horsey fences don't pump up the AGC very much (hence the Collins 50ms attack time?).

But yes, it's all a series of compromises to try to arrive at something simple enough for me to build that works well enough for me to make a few CW QSOs on 80m with my EF80/5763/6146 3-valve homebrew transmitter.

Cheers, Alan G3XAQ

[Radio Wrangler]

Originally Posted by Alan_G3XAQ

The DC amplifier you might use today with an opamp wasn't in the Collins toolkit.

DC AGC amplifiers were used in the valve era of receivers. The GEC BRT400 as an example. Collins were likely aware. I'm afraid it's too late to ask Bill or Luke. Having some DC gain to help with the volts needed for good AGC range means you need less gain at IF, so it helps with the IF stability issues. This means it comes down to the level at which a reasonable detector can be made to work.

David

[FrankB]

I remember servicing a common AM radio with 3 stages of IFT's. really scratched my head on that one trying to figure out why 3 stages. This was back about 50 years ago, so can't remember any details.