UK Vintage Radio Repair and Restoration Powered By Google Custom Search Vintage Radio and TV Service Data

Go Back   UK Vintage Radio Repair and Restoration Discussion Forum > Specific Vintage Equipment > Vintage Amateur and Military Radio

Notices

Vintage Amateur and Military Radio Amateur/military receivers and transmitters, morse, and any other related vintage comms equipment.

Reply
 
Thread Tools
Old 27th Nov 2020, 6:28 pm   #1
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Racal RA17C12 AGC

Yep, a controversial subject. My Racal is now my only expensive boatanchor that is not essentially fully satisfactory for SSB. The SP-600 does now have an outboard assist (a WWII surplus "Q5-er", which allows the radio to run with BFO off and AGC one, all of which works.)

But the Racal, while actually working fine if one turns the RF gain well down,
has hopeless AGC if the signal gets above S-9. (Being an RA17C12 it has an S-meter).

It has always been known that the AGC loop is only marginally stable.
Examination of the circuit shows that there are too many low-pass filters
(82K and .1uF) between the rectifier and the RF stage and thus too much phase shift.

But worse, if one instantly switches on a strong (say S9+20dB or more)
signal (whether a CW station coming on or an SSB break-in) the RF stage
cuts off badly. Because it has its own capacitor with a 470K resistor, it stays cut off quite a while.

Does anyone else notice this?

The circuit does have a feature that affects this, it R81B and R81C, together
a 20 meg resistor that is a current source biasing the ECC189 on, keeping it
on for a while after the capacitors start charging up. But this does not
work well.

Has anyone else thought of improving it? I understand why they
did all this (paranoid about getting the best S/N for low signals), but not
why the result is so bad. Its harmless for AM or RTTY once they get going of course, but its not harmless for even modestly slow CW.

In any case ... all those chained low pass filters are totally unnecessary ...
.0027uF instead of 0.1uF caps would provide adequate decoupling and
not cause instability.

Comments?

Edit: another thought.

That's the region of my set in which the "mad snipper" was most active. I was paranoid
about reconstructing it right, but maybe I goofed somewhere. Thus a serious enquiry.
dtvmcdonald is offline   Reply With Quote
Old 28th Nov 2020, 4:13 pm   #2
Keith
Heptode
 
Keith's Avatar
 
Join Date: Jan 2003
Location: Tewkesbury, Gloucestershire, UK.
Posts: 597
Default Re: Racal RA17C12 AGC

Hi,
I've always had trouble getting my head around this AGC arrangement - not helped by the way it is drawn. I agree there do seem to be a lot of time constants involved. I've also never seen an explanation for the inclusion of V24. Presumably it is a further "delay" for the AGC to the front end.

Mine had been working pretty well on SSB, so much so that I couldn't really see a requirement for a separate SSB adaptor. Just recently, however, something has changed and now the long TC seems way too short. I believe there is also a mod to speed up the attack time.

I think it's time to heave it onto it's side and get the scope on it again.
__________________
Keith Yates - G3XGW
VMARS & BVWS member http://www.tibblestone.com/oldradios/Old_Radios.htm
Keith is offline   Reply With Quote
Old 28th Nov 2020, 5:04 pm   #3
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 16,775
Default Re: Racal RA17C12 AGC

The AGC timeconstants are done with paper capacitors and the circuitry around them is high impedance. There are high value resistors of types known to go even higher.

A general test/swap of likely components could be productive, and you can check things are connected in the right place at the same time.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now   Reply With Quote
Old 28th Nov 2020, 5:45 pm   #4
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

"however, something has changed"

I can answer that around here! and the answer is absolutely wonderful!

Answer: much, much improved radio wave propagation, especially 20 meters and above.

I understand one odd feature. That's the 20 meg resistor to B+. It is a current
source (more or less) and the current through the 470K resistor to the RF stage
raises the voltage on the grid up to about 0, when the little diode tube conducts.
This means that the RF triode does not start losing transconductance until the
two 6BA6s are already at -5 volts on the grid. If it were a real current source,
it would mean that the RF stage would cut off when the 6BA6s were at about -14 volts.
Another effect is that the currect discharges the capacitors with some current that
does not drop exponentially with time. This means that if one removed the 470K
esistor to ground the discharge after a large signal burst would be linear in time.

What I don't understand is why the 1 uF cap (in long mode) is BEFORE the 6al5 section
that is inline (i.e. with the 2.2 meg resistor across it). One would expect that cap to be
AFTER that diode, so it would charge up fast and discharge slowly.

But all those 0.1 uF caps muddy up the whole operation. When I get my bench
cleared of the SX-88 I'm going to try changing them to .022 or .01. My simulator says that that
should get rid of the incipient oscillation.

I really wishes seances were real: I'd like to call up the spirit of the designer of this
mess and ask him WHY?
dtvmcdonald is offline   Reply With Quote
Old 29th Nov 2020, 10:42 pm   #5
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

I decided to look at the AGC with a scope, putting a tube extender on the 6ES8 RF amp.
But ... it was not a 6ES8, a semi-remote cutoff tube, but rather a 6DJ8, a sharp cutoff tube.

Installing a 6ES8 resulted in an improvement. The "linear vs. time" behavior I described is quite obvious as is the voltage range offset of the 6ES8 from the same cause.

The instability (now just a 20% overshoot) remains. I suspect that C159A and C173 should be 0.022 to 0.047 uF with C182 going up a bit. This overshoot is simply not acceptable for SSB as it results in a serious volume drop while it is happening.
dtvmcdonald is offline   Reply With Quote
Old 29th Nov 2020, 11:44 pm   #6
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 16,775
Default Re: Racal RA17C12 AGC

The AGC system is a feedback loop. Changing things like sharp versus remote cutoff valves changes the loop gain and hence the stability margin which can be seen in amounts of overshoot.

Similarly resistor-capacitor time constants on the AGC rail don't give quite the effects you'd expect from their simple time-constant because there is a feedback loop closed around them which modifies their effect. Their overall phase/frequency characteristic modifies the phase response of the loop and we're back to the stability margin again.

Manufacturers of receivers for government contracts etc have to meet requirements specifying max change in audio output level over a large change in RF input level. The figures seen in the spec sheets of serious receivers mean that fairly high loop gain is needed in the AGC system, and some receivers employ DC amplifiers in the AGC line to help with this. In all cases there is enough available gain at DC in the loop that how the gain is rolled off, and the attendant phase delay have to be carefully planned to be sure of stability and good dynamic reponse.

Check very carefully that the components the manufacturer specified are all there and of the correct value to get yourself to a known starting point. This should work as per the manufacturer's spec for the receiver. If not, you've missed something.

Another curve ball these things throw you is that diode detectors on the end of the IF strip have a linear characteristic, while the voltage controlled gain of IF and RF stages tend to have a logarithmic characteristic of so many dB per volt. This means that the AGC loop gain varies with the signal level, and gets higher at higher levels. As already said, this change in gain changes the AGC dynamics across the range of RF signal levels.

As a consequence, changes of gain or AGC characteristic in any of the RF, mixers and IF stages will affect AGC action. Some checks to see if the set is a bit too gainy or a bit deaf can lead you to things which also affect AGC.

Designing an AGC system is not easy. Trying to fault-find a problematic one isn't much better. The only buoyancy aid you can cling to is the original design and its specified performance. Get it working correctly in that condition to be sure you've found all the faults, and only then contemplate making any changes.

The RA17 AGC speeds were tailored for AM (relatively easy!) and A1A morse. The syllabic problems of SSB hadn't yet come over the horizon, that was left for the SSB adaptors.

THere' some serious reading on AGC systems for SSB in one of the chapters of "Single Sideband Circuits and Systems" by Sabine, Schoenike and all the rest of the staff at Collins Radio.

David

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now   Reply With Quote
Old 30th Nov 2020, 3:19 am   #7
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

With the proper tube it meets specs, including rise and fall times, except for long time constant, which are double the spec because I have double the capacitance.
But there is no spec for overshoot.

I broke down and designed a circuit to simulate the AGC acccurately, not using
RF and a demodulator. I had done that previously, but I don't have the patience to
run it at realistic time constants and using nonlinear amplifiers. Amplifing 100 kHz for 10 seconds is not fast with nonlinear functions. I have in the past done such simulations
using real remote pentode tube models .... even simplified, 30 minutes to run. This did tell me the source impendance of the demodulator.

My current simulation is entirely DC and audio at 500 Hz. I'm using amplifier
gain models that are exponential in tube bias, different ones for 6BA6s and the 6ES8 and the 6DJ8. For the latter I use a sharp cutoff. I'm using real (and very good, including the current from the tubes in the absence of external bias) models of 6AL5s. After tweeking a bit the parameters this models the radio nicely with both the 6ES8 and 6DJ8.

Behavior is embaressingly bad with the wrong tube in the RF amp, in reality and in simulation.

So I played with the simulation to get it to work better for SSB. This re
dtvmcdonald is offline   Reply With Quote
Old 30th Nov 2020, 4:29 am   #8
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

Quote:
Originally Posted by dtvmcdonald View Post
...

So I played with the simulation to get it to work better for SSB. This re
I waited too long to edit the message. It continues ...

This required what I had expected, including a different
method of keeping the bias on the 6ES8 in range of its remote cutoff.

The designers never ever expected its users to do CW slow enough that the overshoot problem would
occur except on the first word. With the proper tube it does not cut off, its just a bit weak.

Next step is to try changes and see if the simulation predicts the results.
dtvmcdonald is offline   Reply With Quote
Old 30th Nov 2020, 6:50 am   #9
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 16,775
Default Re: Racal RA17C12 AGC

So, you've found the fault in your set (wrong RF valve) and now you're looking at design inadequacies. The RA17 AGC system is not very good even when working properly. The overshoot with the first word or so of CW is known. They come from an era when most CW operators would switch AGC off and drive the IF gain manually, along with the input attenuator.

Here in Europe, we had megawatt broadcasters in 7.1MHz upwards and the 40m band was 7.0 to 7.1MHz. Receiver overload while trying to hear distant stuff was a bad problem with almost all receivers. The RA17 front end with three mixers before you even get to any narrow selectivity was well known as an underperformer in the overload stakes.... Actually, the plain RA17 flavours had a narrow capacitor tuned filter in the 2-3MHz IF to get rejection of the image from mixing next to 100kHz. That helped a little. I had the RA117E. This variant has no tracking 3-4MHz filter, it has an extra IF and bandpass filters, so it misses out on even that narrowing (This was so the 'kc/s' LO could be switched to external for use with an exciter or synthesiser.) The RA117E first stage was different too, a double triode cascode. This stage went into some of the later RA17 models.

The whole structure of many stages without much selectivity is a recipe for intermodulation trouble. Using one of these receivers in Europe meant you had to always use the RF preselector (RA117 had a wideband position on the preselector range switch - you never used it!) and to run with RF attenuation in most of the time.

I think the AGC design must have been designed to try to not make the overload issues any worse, in terms of how gain was distributed at various levels of signal. The detector of course comes right at the end of the receiver selectivity and sees essentially just the wanted signal, while many stages, up to the 100kHz IF get tenderised by any strong signals within up to a whole MHz block.

Racal followed up the RA17 family by going transistorised. The RA1217. This is a very cute receiver, but with the RA117 block diagram. With the weedy overload behaviour of early transistor circuitry, it was a very limited receiver. AGC was better done, it had gone to crystal block filters with proper dedicated USB, LSB and CW bandwidths (plus AM and RTTTY) all depending on optional fit. There was a proper product detector. The British Army bought loads and many went in Land Rover mounted radio vehicles, using whip antennae of modest size. The small rack height of the RA1217 was the attractive feature. It was convenient not having to rack up an SSB adaptor with it, but the poor RF/IF dynamic range was a major limitation and a setback from even the RA117.

Eventually, someone at RACAL cottoned on to the issues and their next generation, the RA1772 paid a lot of attention to RF linearity and overload. They licensed the use of the Rafuse D-MOS mixer (from MIT?) and went straight to a crystal filter at ... was it 34MHz? This set fixed a lot of problems. But it still needed manual operation of separate control knobs for its preselector, agc was better in some ways, but it was a hang circuit which would hold for half a second before decaying to higher gain. This meant short impulses could deafen it. Around the same time, much more expensive receivers from Marconi (H2900) and GEC (RC410) had servo-motor controlled preselectors tracking the tuning of their synthesisers.

After the RA1772, RACAL went backwards again. The RA1792 retained the Rafuse mixer, but replaced the preselector with a bank of suboctave filters. It is very much a cost cutting exercise. For a while their second hand prices were rather high.... a later generation receiver from RACAL had to be better, didn't it? But eventually reality sank in... even the bearing for the tuning knob was cheapened to a nylon bushing!

So you may be stuck with the AGC flaws of your RA17. Doing much about it risks worsening overload problems at various signal levels and offsets.

I've never had an R390, they'r fairly uncommon over here. I have had Marconi H2900 and GEC RC410 (GEC and Marconi competed against each other even though GEC owned Marconi) Racal receivers have been RA117E and I still have an RA1217 and RA1792. I've always had my eye open for an RA1772 at a good price, I think this was their best.
GM4ZNX was on the air for many years with an NOS RA117, and a Redifon GK203 multimode exciter plus a home built PA. I had an HP spectrum analyser looking at the RA117 IF to see what was about.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now   Reply With Quote
Old 30th Nov 2020, 11:19 am   #10
turretslug
Dekatron
 
Join Date: Nov 2011
Location: Surrey, UK.
Posts: 3,644
Default Re: Racal RA17C12 AGC

Is it possible that a previous owner had substituted an original 6ES8 (ECC189) with the 6DJ8 (ECC88) in an attempt to pep up its ultimate sensitivity? Over decades, I've noticed that there is any amount of spiel around about allegedly super-charging radios with "hot toobs" such as frame-grid pentodes intended originally as wideband TV IF stages, or high-gain double-triodes typically widespread in oscilloscope amps connected in cascode, but there is never, ever even a murmur on the subject of compatibility with existing AGC sytems as regards the contest-sweeping new RF device. It makes me think of earnest kids "souping-up" a mundane first car or motorbike, but with scant consideratiion of the effect on all-round useability! No doubt there are examples of well-considered and genuinely useful examples of updating existing sets as vacuum tube technology improved, but I'm inclined to think that a real, professional RF designer would suck his teeth and say, "Um, there are many knock-ons, and it's more difficult than you think to improve upon the performance of an existing designed-together system just by popping in one different device on the strength of a brief scan of a few details of a single operating point".
turretslug is offline   Reply With Quote
Old 30th Nov 2020, 5:41 pm   #11
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

Around here there are only three overload problems: 580 and 1400 kHz and 94.5 MHz.
The latter is so awful its even audible in the RA17 if every screw is not tight, and in the R390A (barely) even if they are. It appears due to mixer drive harmonics. But other than that no tube (valve) set is anywhere near bothered. My Airspy, however, direly needs really good prefilters, and on good propagation days there is one station at exactly 6MHz that is a severe problem. Those problems were solved
with a set of prefilters that cost more than the Airspy itself plus the Spyverter combined.

I love making mods ... tinkering has been my whole life. Its served me well as a grad student and professor. Especially fun is mixing electronics and liquid helium. Its took lots of tinkering to get a noise figure of -16 dB, and an input impedance of 1 megohm in parallel with 1 pF.

The mods I propose will have an effect on overload only in the range between when the overall AGC delay starts kicking in and when the current source RF stsge delay kicks in, and that change will reduce overload. It won't reduce useful sensitivity as the
overall AGC delay is in fact probably a bit too much.
dtvmcdonald is offline   Reply With Quote
Old 2nd Dec 2020, 10:18 pm   #12
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

A part for my SX-88's dial indicator problem got delayed, so I was able tp get the
RA18C12 on the bench. First I went looking for mistakes I made when recreating
the AGC after the Mad Clipper got at it. There were none.

I did lots of experimenting. My simulations turned out to be reqsonably sane.
The unfixable problem was as I suspected: the 2nd IF stage simply can't produce enough output current to charge up the AGC fast enoufgh above a certain
signal level. Decreasing series resistance values does not help, its in the tube itself.
The only cure would be a big fat transistor emitter flowwer.

That said, of course up to some signal level it CAN and mods do help, quite a bit
it turns out. Here's how I have it at present, presented as a recipe.

1. Remove C159B and C173, both 0.1 uF.
2. change C182 to 0.27uF
3. Disconnect one end of the R81B - R81C chain and check to be sure its in fact 20 Megohm. Replace if bad ... mine was way off.
4.Replace R118 with 4.7 Meg
5. Disconnet the wires that go to the wiper of S6, the AGC speed switch. Stow them away for a moment.
6. Remove the wires that go to the "Short" side of S6 and connect to the wiper.
7. Put capacitors to taste from the Long and Short sides of the switch to ground.
I used a 1.0 uF 250 volt film cap and 1 4.7 uF 63 volt electrolytic, negative side to the switch of course. For electrolytics I recomment 105 degree expensive Nichicons but Panasonics are OK too.
8. Take one of the wires from Step 5 and stick through one of the unused holes in the switch, and solder to the other one. This holds them securely enough.

That's it for changes. The meter will need a rezero. The AGC connection on the back
can still be used to monitor it or you can add even more capacitance across it.

The main difference that I noted is that the fall time to rise time ratio is noticeably higher, dramatically so if the signal change upward if small enough to not overload
the signal amplification chain. If it does overload the unpleasant effects are
smaller.

I will report back later after trying it under many different band conditions. No local
S9+50 signals have appeared yet.
dtvmcdonald is offline   Reply With Quote
Old 3rd Dec 2020, 5:00 pm   #13
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

Its morning and 40 and 20 meters are alive again, as is the
flamethrower on 580. SSB is up to 15 dB over S9 on 40, but as usual noise
level is at S5.

The status as of last night was good but still quite irritating due to
the odd rise time situation. The "differential delay" on the RF amp remained.
That means that the gain on that stage starts going down at higher signals
than on the other two tubes. This is done with a constant current through
R91a. However, there are capacitors (three 1 nF ones) remaining downstream
(from the AGC source) from it. These are clearly for decoupling purposes.
They would be hard to remove, especially since I actually have never
found one of them, C48. The problem is that while in that situation the
rise time of the AGC voltage is short enough, when it stops rising the
system is not in equilibrium. There is too much reduction in gain on
the 6BA6s and not enough on the ECC189. The overall gain reduction is
too much. So it has to get back to the equilibrium situation, which has
less driving force than the original rise, and so a vastly slower rate.

The only cure for this is to reduce the extra time constant. That can
be done by reducing the 470k R91A or reducing the three decoupling caps.
Doing the latter is either physically bothersome or, as I said,
impossible since I can't find C48. So I reduced R91A to 27K and the
even less needed R102 to 4.7K. This reduces the extra AGC delay by
a factor of 20.

The result is dramatic: no more noticeable extra time constant. Even
with the 5.2 uF slow time constant cap the AGC rise time is about
35 mS or so, almost as fast as the risetime spec for "fast" in the
manual. Its almost inaudible. The fall of the AGC becomes
truly linear in voltage (or approximately linear in dB, very roughly
15 dB/sec if I calculate right on slow. I may want to make it even slower.

So the big question is ... does this result in the ECC189 actually
cutting off at extremely high signals? Answer: the Mullard datasheet
says that the ECC189 is still working at -12 volts on the grid. The
strongest AM signal I can muster is 0.1 volt RMS into the set. That
generates -15 volts of AGC. The AM signal sounds a bit strained.
Cutting the signal 10 dB and it sounds just fine, at about -12 volts
AGC. Note that a 10 Meg scope probe is not high enough impedance,
I'm using 40 Meg.

I keep playing to learn more. I need higher signal level SSB.

P.S. Adding two instruction steps:

9. Replace R91A with 27K
10. Replace R102 to 4.7K
dtvmcdonald is offline   Reply With Quote
Old 4th Dec 2020, 12:05 am   #14
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

I just noticed: my Step 1 should be C159A, not C159B.

This evening some SSB signals reached S9+20dB and still sound fine.
dtvmcdonald is offline   Reply With Quote
Old 6th Dec 2020, 2:16 pm   #15
dtvmcdonald
Pentode
 
Join Date: Mar 2016
Location: Champaign, Illinois, USA.
Posts: 194
Default Re: Racal RA17C12 AGC

Its done, buttoned up, and back in its place of honor next to the
R390A and SP44.

One more comment on the original "design". I found the schematic of a
earlier pentode RF amp interation.

There is no AGC on that pentode! There is AGC only on the two IF amps.
The circuit for that is almost the same as the L or C12 up to the 470K resistor
before the teensy diode. "Almost" because there is no 1 uF cap switched in
on "slow", just a shorting of the 6AL5 series diode.

I don't understand the reason for the lack of AGC on the RF amp. This means that
big signals will overload all the mixers!

In any case, when they did put AGC on the RF amp in, perhaps in order to
keep that sensitivity claim, they put in the "added" extra AGC delay of the cascode. When they did, they threw performance on big, fast, signal rises out the window.
The design they used is, it must be said, elegant. Leaving the current source in
is quite a pleasure to listen to, with the constant (in dB/sec) rise in gain. There is no irritatingly slow final return to full sensitivity. I added it long ago to my modified
R390A AGC.
dtvmcdonald is offline   Reply With Quote
Reply

Thread Tools



All times are GMT. The time now is 10:30 am.


All information and advice on this forum is subject to the WARNING AND DISCLAIMER located at https://www.vintage-radio.net/rules.html.
Failure to heed this warning may result in death or serious injury to yourself and/or others.


Powered by vBulletin®
Copyright ©2000 - 2021, vBulletin Solutions, Inc.
Copyright ©2002 - 2021, Paul Stenning.