Noise in receivers

[ianbatty311]

Hi, folks.

I'm looking for an answer to a mystery.

When I get a radio on the bench and turn it on at maximum volume, I get a certain amount of noise output, let's say 1 milliwatt.

Then, I connect a sign gen and bring the signal up from zero.

I can hear the 400 Hz modulation tone, but I also hear (and measure) an *increase* in the noise level - it can easily get to 10x the noise level with zero signal input.

As I increase the signal level, the tone output increases, and the noise level eventually subsides so that I can get the standard of 20 dB Signal-to-Noise.

This is the puzzle: why does the noise level increase initially? I get why it decreases with stronger signal input, but why is it highest with the weakest signals?

Which radios?

All superhets with good sensitivity, whether it's valve sets (wire antenna or loopstick) or transistor sets (wire antenna or loopstick).

... and I first noticed this back in 1963 testing VHF aircraft receivers in a military-grade screened room.

Yep, all sig gens, too: HP, Advance, Fluke, homebrew, whatever.

Outside interference?

Nup, got my own screened room now that is electrically quiet (really quiet, not a peep from broadcast stations and no electrical interference) for testing.

Any ideas?

Ian

[Radio Wrangler]

All radios
All signal generators
Everything with an oscillator inside it.

Why?

Well, the answer is text-book sized. Look up 'phase noise' and 'reciprocal mixing'

I put my short version take on it in the ARRL handbook. It's in the Oscillators and synthesisers chapter in any year after 1995. Much too painful to retype here.

David

[SteveCG]

What you observe is absolutely typical of AM radios. I explain it to myself in a rather hand waving way: When there is no carrier present (your signal generator output turned down to zero) then the noise you hear from the loudspeaker is just one part of the IF noise spectrum beating with another part producing an audible difference signal. These noise parts are independent of each other - ie a random spectrum. When you have a carrier present then what you hear is the sum of the parts still beating with each other AND the carrier beating with the parts. The signals are multiplied together - the detecting diode is a non-linear device - so the audible sound gets louder as you increase the carrier level.

I am certain a text book will explain it all in a proper manner - and probably far more accurately than me!

[kellymarie]

I think the reason the levels appear to decrease after a certain input is reached is possibly due to the action of receiver AGC as the signal gets bigger the rx reduces its gain sounds reasonable to me

[Wendymott]

How strange !!!!! I was thinking of posting the identical "mostly" post by Ian.
I have.. somehow..got 4 X FRG 7's in my workshop, that I agreed to look at, as part of buying one for myself.
They all have issues, but mainly the noise increase as you increase the RF level from say 2 uV, again at say... 10 Mhz, as a reference.
I concluded that. either.. as they are 40 years old ish..... it is a common degradation, possibly , alignment...... or. Thats how they are...
I have not had use of these "Wardley loop" types before...so I didnt know what to expect, but as they are all the same, I guess the replies from David and Steve fairly cover it.
Incidentally, one of my ceiling lights (Led) causes RFI, so that was one source. Fortunately the Main Led light is RF quiet, the smaller one is for "boost" if I am tired.
Going back to receivers...is it pointless having a receiver look at < 2uV.. as the background noise... at mine.. is S7, anyway.

[G6Tanuki]

If you're talking 'classic' AM receivers with diode detectors then one such 'noise rise' effect can be because in the absence of a signal the receiver's internally-generated noise is only able to drive the diode detector part-way up its initial 'knee' of conduction, meaning that the detection process is inefficient and the detected noise-level is low.

When you introduce a carrier the diode is driven further up-the-line from its conduction-threshold and into its linear-conduction portion; the diode is now more-able to detect the receiver's intrinsic noise-component so it appears noisier.

This effect can actually be useful - some old receivers had 'quiet tuning' in which a negative bias was applied to the detector so it remained entirely cut-off in the absence of a carrier (or only a very weak carrier) so effectively muting inter-station noise.

[John_BS]

Originally Posted by kellymarie

I think the reason the levels appear to decrease after a certain input is reached is possibly due to the action of receiver AGC as the signal gets bigger the rx reduces its gain sounds reasonable to me

I agree. Sort of inter-station muting.

[Refugee]

You can demonstrate the AGC effect quite well by recording with a germanium transistor shoe box tape recorder.
If you leave long pauses in speech in a quiet room you can hear the noise come up as the AGC cranks the gain up.

[Radio Wrangler]

With most receivers using a diode detector for AM you get a threshold effect so if things work out well, noise tends to get suppressed as it doesn't even begin to turn the diode on. Put in a signal that does get the diode going and the noise appears. Select SSB or CW and turn a BFO on and that will fix the detector threshold effect.

Put some more signal in and the agc starts to confuse things. The dominant stage for noise contribution will go walkies as the signal level climbs and gain adjustments are made in different areas.

Close to carriers, phase noise sidebands will appear and they appear on all signals through reciprocal mixing. Everybody went off hunting mixer and amplifier linearity in the 80's, but phase noise was a worse limitation in the better receivers.

David

[Wendymott]

I have more observations to add, with the FRG 7 in mind. I decided to go through the re alignment "hoops" as per the manual, with what I thought was the best sample,
I then made a sensitivity run at 1 Mhz intervals noting all the settings.

I then tested one of the particularly noisy receivers, the results were very disappointing , both in sensitivity and noise.
So sample 3 was tested, again poor sensitivity and noise, however I conducted a re alignment. I found the 52.5 Mhz strip was considerably mis aligned, possibly due to component ageing, but who knows. Once re aligned the sample was nearly as good as the First sample, less noise, better sensitivity.... etc.
I was fortunate to have not just one but four receivers for comparison, and I think I am winning this one.

[ianbatty311]

Hi, folks,

Yes, I have read of the phase noise phenomenon - that it is due to phase noise in the local oscillator interacting with phase noise in the signal with low-amplitude inputs.

...and that it can be greatly reduced by using a coherent/low phase noise local oscillator.

OK, so I got my HP synthesised sig gen and a transistor set using a separate local oscillator (i.e., not an autodyne).

The set uses emitter injection, so I disabled the LO and injected a volt of LO frequency (the injection level implied by the circuit schematic) from the sig gen to the emitter of the converter.

I could not get much/any improvement in the noise-level-rise-with-weak-signal effect.

So....

1 - is the LO phase noise thesis just a Furphy? (Australian English - a fanciful or false tale)

2 - is it likely that the HP generator's output contains just as much phase noise as my old-fashioned, valve-equipped, Advance sig gen?

3 - Has *anyone* actually tried to solve this mystery OR had/built a radio that does not exhibit the effect?

(really wann'a nail this one down before I shuffle off my mortal coil, folks)

p.s. Furphy - said to derive from soldiers' chatter while standing around a Furphy water barrel - see Scuttlebutt for the Naval derivation.

[ianbatty311]

To GXtanuki

Um.....

Some years ago, at an HRSA Farm Meeting (thanks to Laurie Harris), I ran a receiver competition for really old sets.

Most had neither frequency nor station calibrations, so I brought along my trusty HP as a 'spotter' to help tune to known local stations.

Jim Easson (hi, Jim) had an older set that he thought would work pretty well, and it was OK.

...until I turned off my HP.

Then the radio's output fell noticeably - you could easily hear the drop in gain.

Ummm... so I turned the HP back on, and we both listened. Got the 400 Hz modulation, so I turned that off.

Well, Dave and Bill had given me a sig gen accurate to a few hertz of dial setting (you could just hear a low-frequency 'swoosh' kind of sound behind the station program.

Turn the sign gen off, station sound drops.

It's heterodyne detection, where - within reason - adding in a very-near or actually-coherent signal pushes a diode demodulator up its non-linear curve to give more output.

It's the radio equivalent of the low-intensity bias light used to push camera tubes (ever since the Iconoscope) up out of their low-sensitivity 'inertial' part of the response curve.

Where was I?

Oh, yes -

so I get that there is a native noise level - mixer noise, usually - and that adding the signal can push the demodulator up off its low-sensitivity curve and effectively give more amplification by virtue of moving to the linear part of operation.

This is distinct from the phase noise argument, however...

So what do we think?

Is the phenomenon due to pushing the demodulator up its response curve (as was clearly demonstrated with Jim's radio), or to phase noise?

If it *is* phase noise (which the textbooks assert), where did I go wrong using my HP generator to substitute for the LO in my transistor radio experiment

Your thoughts?

[Synchrodyne]

Some items that may help or hinder – I’m not sure which.

Firstly, a generalized quieting curve, which shows the expected initial increase in noise as signal level increases, followed by decline.




Secondly, the quieting curve for the Marconi HR92, HR93 series of point-to-point SSB/ISB receivers.




Here the noise curve shows no initial increase, just a steady decline. My initial thought that was this might have been because these receivers had only coherent demodulation, not an envelope demodulator of any kind.

That was confirmed by the quieting curve for the GPO SSB-ISB receiver (I think Radio Receiver #22, or a derivative thereof), which was designed to a similar target to the Marconi HR92, HR93.




But then this from the same source as the first, generalized curve.




Here the shape of the noise curve depends on the number of RF amplifiers. With two, there is no hump at all. The same chapter of the book, written by a Marconi staffer, goes on to describe what is surely a Marconi communications receiver, which I think might be the CR150/5. I imagine that the curves were obtained via the AM demodulator, a diode. The described receiver has a typical Marconi RF amplifier section, with a Z77 (EF91) followed by a W77 (EF92). As best I can work out, the HR92,93 would have been similar. Anyway, it suggests that the type of demodulator might not be the key factor as I first thought.

That all leaves me as confused as ever, although with one thought that the mixer might be the guilty party as far as causing the “noise hump” is concerned, and that where it is preceded by RF amplification, the hump is pushed down the scale relative to input signal strength until perhaps like the Oozlum bird, it just disappears…

Quieting curves for FM receivers seem not to show a noise hump.





Cheers,

[ianbatty311]

Synchrodyne said, in part:

'Quieting curves for FM receivers seem not to show a noise hump.'

Now, THAT is interesting.

An FM demodulator responds to changes in phase/frequency, right?

So, if the increase in phase noise in an AM receiver is caused by LO-signal interaction (due to a noisy LO), then why do FM receivers not show this phenomenon in their lowest-signal/below limiting threshold region?

Curiouser and... whatever.

Any ideas?

Ian.

[David Simpson]

Ian, from your reminiscences of military aircraft equipment back in the 60's(working in a wireless bay ?), I'll stab a guess that you've done an RAAF, or RAF "Fitters Course" ? Just like me. Get out your old training notes & have a shufti - its all explained. "Aircraft VHF Receivers" eh ? I'll again stab - at the 1985 series 10 channel VHF transceivers. Or possibly - ARC52. AP's for both can be accessed via the VMARS huge online archive, or bought from Birketts in Lincoln over here. Again, these "2nd line/3rd line AP's fully explain the noise reduction circuitry. Stretching my memory to the limit - there was a whole chapter in the ARC52's AP on their efficient "squelching" circuitry. ( Set at 5uV, using a CT394B) - I've done quite a few back in those far off days.
The whole plethora of "Radio noise" sources/avoidance would take up reams of Forum thread pages. (More than phooked AVO VCM's of late, eh).

Regards, David

[Synchrodyne]

There is some discussion of the phenomenon at interest in Langford-Smith. See pages 1229-1234 of the Classic edition.

I have excerpted a graph that is the subject of discussion therein. One receiver configuration had a “noise hump”, the other did not.

Not the only factor, but the incidence of AGC bias does appear to be a contributor here. How it is applied to the various stages (delay and proportion) could have an effect on the shape of the noise curve. That might help explain the apparent “no noise hump” FM curves. Where RF AGC was used on FM equipment, it was usually seriously delayed, so noise increase due to the early gain reduction of an RF stage, both reducing its own SNR and allowing the following stage noise to make a greater contribution, was unlikely.




Cheers,

[Radio Wrangler]

The diode in your AM detector is a non-linear element.

Your incoming signal is a carrier along with a pair of symmetrical sidebands. The idea is that the carrier is stronger than the sidebands and so dominates the action of the diode. Seen in the limit of a very strong signal, the carrier component turns the diode hard on and hard off. So the diode operates as a gate. This on/off gating modulates the passage of the sideband signals into the output of the detector. It's the same as mixing the DSB components in the IF with the carrier, and the result is the wanted audio. THe carrier also mixes with itself. It is, of course, perfectly in phase with itself, and so it also produces a DC output... the rectified carrier, seen another way. With the diode turning hard on and hard off, the conversion of IF sideband power into audio power is quite efficient.

So how about a very small signal in the IF reaching the detector diode?

There is less voltage in the sidebands, so you get less audio. But there is also less carrier, so the diode is not being turned on very hard. So its efficiency at converting sideband voltage into audio voltage is degraded and gets worse the smaller the signal.

So these two effects combine, as your signal gets smaller the signal gets smaller and the detector gets less efficient, so the amount of audio recovered falls disproportionately.

So now you keep the low level signal, but you add a signal generator operating dead on the carrier frequency and at a level bigger than the signal. This works like the signal's own carrier, but being bigger, it operates the detector harder, making it switch more strongly, making it more efficient! so the amount of audio you perceive increases.

The added carrier could be a little off-tune if the beat frequency is below the cut-off of the audio stages. This is called Exalted Carrier reception. It's more sensitive than a plain old diode AM detector. The added carrier could be phase locked to the carrier of the incoming signal if there is enough of it for a narrow band PLL to acquire it. This in the Synchronous detector found in some receivers - particularly ones intended for HF broadcast DXing.

Anyway, that's an explanation of why the audio improves when you add an artificial carrier. You can add it at RF or IF frequencies.

Phase noise effects only come into serious play when there are multiple signals or whn you tune close to DC.

If your LO has bad phase noise sidebands, the principle of reciprocal mixing makes things work as if your LO was clean, but every signal in the band had sidebands on it just like your LO. In bad cases these can overlap and add. So your receiver experiences a worsened noise floor. Thke the signals away and try to measure the floor, and it's OK. Drives people round in circles!

David

[Wendymott]

Let me throw a virtual "hand grenade" into the mix...... There is a lot of theory being spoken about the diode and its actions....... I now know where my noise is coming from. The early variants of my FRG 7's as in the Somercamp versions. Its an AGC problem. The RF section is being shut down too early, as in < 10uV..... any increase in RF level above that the noise increases but the audio level of the tone does not...either its a design error that was rectified later, I know not... yet. I will let you guys get back to the Theory

[Radio Wrangler]

It's amazing just how badly AGC is implemented in some receivers.

The AGC detector is right at the end of the IF, benefitting from the best selectivity of the set. Unfortunately, earlier stages can get tenderised by larger unwanted signals, just offset out of the channel, or further away.

Few sets have detectors sprinkled around the earlier stages to tell when they are hurting, so at best, the planning of an AGC system is guesswork. Somewhere between blind and rabidly optimistic is usual.

To get good signal to noise ratio (no interference) it is usual for the AGC system as it comes progressively into action with increasing signal level to first start reducing gain in the later IF stages where damage to overall noise levels is least. Only later, at higher levels, is the gain wound back in the earlier stages. This is confusingly called "Delayed AGC" There is no time delay. It is the onset of gain reduction in early stages being offset in level. On the whole, early stages are left to sink or swim if there are large out-of-channel signals around.

The FRG7/RA17/RA117 have many stages in their front end which have to be at least a whole MHz wide, and the RA117/RA1217 have an added IF which is still appreciably wider than the final selectivity. Fertile ground for out of channel overload issues. These sets do indeed have reputations for soft front ends and you really have to operate their preselectors and RF attenuators with some skill to get the best from them.

It's no coincidence that Racal moved with the RA1772 to a structure where the channel defining filters were a lot closer to the antenna, in stage count as well as net gain.

David

[Wendymott]

VERY interesting David. I guess its like the audio quality in TV's.... gets least design thought...the front end FET indeed is given the same voltage on G2 as the later I.F devices. I guess its a case of... if it works.. ish..get on with designing something else... its only Amateur product.
To compare two receivers (FRG7's)with decidedly different apparent gains, is to disconnect the AGC and check each stage to see where things change appreciably, or is it a subtle change which is accumulative. I then wonder if the "Good" receiver has been "tweeked" and I am looking the wrong way.
Still it keeps me off the streets.