On 20 Aug 2006 12:46:25 -0700, "
wrote:

shortend

Read EMRFD page 6.16 (ARRL press) they tested the 1350 and at the
point where the gain cell has equal conduction on both legs the noise
rises significantly. I duplicated the test fixture and yes, it's
noisy, from around 6db to around 11db in my fixture when gain is
reduced by 10db and that was at 16mhz. In a reciever that used
it I went to two cascode stages using JFETs and the difference noise
was notable for weak signals just into the agc range. I restrict the
1590/1350/ca3028 for lower perfomance recievers now.

Apparently I hit some nerve on my disagreement.

My first experience with the MC1590 was in 1973 and a need to
operate over 55-64 MHz. Electronic gain control was essential
and it had to be fast. Motorola supplied some additional
information which was later incorporated into appnotes.

The MC1350 was marketed around '73 along with the MC1330
video detector as a TV IF system. It didn't sell that
well in quantities (presumably) and both were dropped
from active production (Lansdale acquired masks and now
makes the MC1350). The 1350 (8-pin DIP) should use the
same die in the metal can MC1590. While neither one was
ever touted as a super-champ low-noise device, it is what
I consider respectable as to NF.

Wide open and at 20db reduction the noise figure is not bad at
all. Only that the apparent increase is noteable. I've seen it
occur with other topologies including simple bipolar or FET
stages (even tubes).

The fact that it has
differential input and differential output is convenient
from the standpoint of circuit design. Especially so when
input impedances (each side) has a dependable 5K R in
parallel with about 5 pF total capacitance. Gain of both
begins to fall above 75 MHz with output loads of 100 Ohms
resistive. I've found no noticeable difference between
differential input v. single-ended.

Those are the feature of the part that makes them desireable.

That IC is what I term a "double Gilbert cell" in that
AGC control current affects both differential inputs
equally (or very nearly so). Whether one connects to
both inputs or just one shouldn't make any difference
other than output gain.

I call that circuit an analog four quadrant multiplier for what it
does not how it's made.

The feature of the Gilbert cell that applies for the AGC use is the
lack of DC shift at the output points keeping downstream
DC coupled stages at their undistrubed bias points.

However the active device that is in the agc control
positions is still a noise generator (as are all active devices)
and as agc increases it's contribution is additive to the
RF path devices. Makes little difference if the node where
outputs are combined see no DC shift the various diff amp
transistor as individual pairs do see a significant shift (100%
collector current to near 0).

It's easier to see using the older MC1550 or CA3028 diffamps
or even discretes in a diffamp with current source.

I also verified that the 1590 does same and also the CA3028
wired as differential AGC. Even tried three 2n3904s and
same result. The agc range was good and at full gain the
noise was ok but the noise increase at partial agc was surprizing.

I've never encountered any "surprising" increase in noise
at any AGC input to a 1590 or 1350 causing partial gain
reduction. That is as true in 2005 as it was in 1973. If
there is a SNR of 10 db at an RF carrier input of 3 uV and
a gain reduction of 10 db for a 10 uV RF input results in
3 db more noise in the front end, the SNR with a 10 uV
input is still higher than the one at 3 uV. What has been
"lost" there?

The surprize is that I'd not considered the possibility that
the SN+N/N could degrade unevenly due to applied agc.
So I'd never paid attention until I was trying to improve an
earlier reciever design (ca1978) of my own and at the same
time aquired a copy of EMRFD and did some testing to verify
their resuts. Since the design was optimized for low RF
gain and high overload thresholds I was revisiting anything
that could better the design without loosing those features.
Note it's a single conversion system with high IF. The
problem was a MDS of -136dbm but the 10db Signal+N/N
point was around -110dbm and at ~121dbm it was worse
than at -130! The front end was common gate RF amp
(2n4416s) driving a pair of 4416s in a single balanced
mixer. Low noise but limited gain for better overload
performance. No agc before the IF. If needed there
are switchable resistive attenuators (3, 6,12db or 19db total).
Measured gain from antenna to IF is only 16db (after all losses).
Disable the agc or increase the threshold and it wasn't as
measurable or appeared to disappear.. The hunt was on.
The results were a surprize as there is no data for noise output
with no input or signal to noise with gain reduction. My Moto
databooks go way back, as do my National, RCA and
Signetics library. More current datasheets do not reflect any
improved information.

The revised RX used two stages of mpf102 Jfet in cascode
plus a diferential pair of 2n3904s to resolve the 5-10V agc to be
compatable with the new fet amp to replace the two MC1350s
and the problem of decreasing signal to noise as signal increased
with agc active disappeared. Not to say the fet amps did not do the
same thing only that the rate of noise increase was a smoother curve
from max gain to min gain. In retrospect a delaying AGC to the first
of the two 1350s could potentially have the same effect but was not
investigated. I may revisit it at some time as I still have the
original if module in the junkbox.

Let's look at the original problem starting this thread:
There was a claim of "increased noise" with AGC on, but
no quantifiable data.

Thats a problem, the lack of data or information on the circuit.
Also I've repaired a few commercial radios that due to component
failure or "golden screwdriver" had the various operating conditions
sufficiently altered as to cause a similar problem.

The sudden segue to stating that a
certain IC is "bad" is a leap that defies good design
practices to me. I'm not impressed that the ARRL had some
test data in a publication; having been hands-on with this
Motorola design for a number of years, I have a number of
RCA lab notebook pages filled with my testing of it along
with a patent involving it granted 1974...besides my own
hobby notebooks.

That's nice but are the test results in error from two different
sources? No.

However, it's was a noteable weak point. But calling it bad is
your words. It's a point that needs to be understood and
allowed for. In a design with more RF gain and/or less mixer
noise it many not have been a factor or less of one. Also in the
case that brought it to a point for me even altering how agc is
applied might have achieved a better result.

Since the 1350 is at IF for most designs the noise is likely from
front end causes should be investigated first. One would hope the
design had secured the system noise performance before
the IF. However in low gain systems or system with no gain
before the mixer and first filters this may be problematic.

Low-noise input amplifier design is an entirely separate
subject and there are a number of other active devices
which can do lower NFs than 5. What was orignally needed
was some way of getting some numbers and test configuration
of Andrea's problem...to pin down a possible reason for
alleged increased noise with AGC applied, presumably a
"partial AGC" application. [I can't quantify "partial"
as a numeric value...maybe others can?]

I do most of my RX experimentation at 6/ 2M and 70cm SSB so
noise and overload perfomance are important to me. Images
are also a big problem as I'm near a lot of VHF/hf broadcast.

[shrug I live about 6 miles from 50 KW KMPC on AM...]

10 miles from the Needham towers in MA. Not less than 8 VHF
broadcasters, then the usual crowd of UHF and now the
HDTV-UHF broadcasters and no small party of FM
broaccasters.

Oh and WKOX 1200 AM three miles away.

Then I have 9 hams within a 1 mile circle and two within 1500ft
running KW level at VHF. RFI are us. I understand overload
as +15dbm on coax is common here.

It's an interesting design challenge to do low noise figure RX
and at the same time be overload resistant in a harsh environment.

If we can get back to the original claim of "increased
noise with AGC applied" we might be able to help Andrea
some. We don't know what Andrea has for a main receiver
and interjecting some "badness" remarks by the ARRL about
a certain IC isn't going to help clarify Andrea's problem.

Having tested and understood the problem I would say the
authors of EMRFD did a fair job of pointing out the points where a
device needs better understanding. A blanket "it's great" is lore,
testing it and understanding it is engineering. Having done the
work to understand it better I can appreciate the perfomance of
the part and it's limitations.

I still use it and have a tube of them because it's a useful part.
Just like the often reviled SA602 mixer.

It's relevence is I've seen this before and understood it's origin
and also elsehere. The other aspect is that if a commonly
accepted part is not fully understood and can lead to undesired
effects then, why not others. AGC is not a trivial thing to be
tacked on and considered a problem solver. Protects the ears
but it's place in the reciever is not always understood. It does
not always solve things like gross overload at the front end or
possibly further down or outside the agc detectors bandwidth.

But a lack of information about his radio doesn't help us either.
We do not know for instance what topology is used for RF
and if any agc is even applied to it.


Allison