Improving the Selectivity of the R-105A Receiver

I discussed my experiences getting my ARR-15/R-105A receiver on the air in a previous blog posting (here). As I mentioned in that posting, the selectivity is very broad -- so broad that, when using the receiver on, say, an 80 meter AM net, it really suffers from adjacent SSB interference.

I would like to eventually pair this receiver with an ART-13 transmitter I'm working on, but first I needed to improve its selectivity.

How best to do this?

The R-105A has a variable IF -- that is, the IF frequency varies from 450 KHz to 550 KHz, with it being 500 KHz at the BFO "detent." Interestingly, when in MCW mode (i.e. AM), the BFO is OFF when the BFO dial is in its detent ('0') position, but it turns ON as soon as this dial is turned away from its detent.

So, clearly, for AM operation the BFO dial should be left in its detent position (otherwise the BFO is ON and you'll hear it hetrodyning with the carrier).

OK -- so the IF frequency is 500 KHz when copying AM. I wondered if I could cobble a 500 KHz AM mechanical filter into the radio. But first, I needed to find a filter...

I looked around the internet to see if I could find one for sale. Finally I found one on the Fair Radio site (Lima, Ohio). Actually, I found two:

The first filter was contained within the AF Audio Amplifier Module (Fair Radio p/n 546-6053) for the Collins 618T HF Transceiver, and it was a 6 KHz wide, 500 KHz mechanical filter (F500 Y60 (526-9378)). The price Fair Radio was asking for this module, including filter, was $52.00.

The second filter was contained within the 500 KHz IF Audio Amplifier module (Fair Radio p/n 105-AA) for the RT712 / ARC-105 HF Aircraft Transceiver. Fair Radio doesn't specify the filter model number, except to say that it's an AM filter, and that the IF frequency is 500 KHz. The price for this module was $44.00.

(Fair Radio provided the latter with a copy of the module's schematic. I don't know if they also provide this if you order the first module.)

Fair Radio also mentioned that the ARC-105 is a pressurized version of the 618T2 transceiver, so I thought, what the heck, it probably has the same filter as the 618T series transceivers, which is the F500 Y60. And best of all, it was $8 less than the other module.

So I ordered it. When it arrived, I noticed that all of the boards within the module were covered with a clear conformal coating. This wasn't a real big deal, but it's probably one reason why this module costs less than the 618T module. And it kept me from determining what the filter part number actually is, because the conformal coating effectively glued the shield bracket to the filter, and the filter's label is under it.

Here's the module I received from Fair Radio (with its cover removed). The filter is the long cylinder in the upper corner...


Because of the filter's insertion lose, I would need some sort of amplification to compensate. I first tried using the amplifier built onto the filter module's PCB. It actually worked fine, and I initially considered using this PCB with filter and amplifier already built and working for my project, but the board itself is a bit too large for mounting within the R-105A chassis, and so I decided to roll my own amplifier.

I designed and bread-boarded the filter and amplifier circuit and, after some changes, I mounted them on a small piece of copper-clad PCB material. Here's the finished circuit, installed within my R-105A:


This design uses an existing hole within the receiver for mounting, so a purist can, at a later date, easily remove my modification and return the receiver to its original condition. I connected this circuit into the receiver's existing circuitry by unsoldering a wire from one of the pins of the IF transformer Z-119 and soldering in new wires in its place (and connecting another wire to the receiver's AVC line). Again, the new wires can be easily removed and the original wire reinstalled to return the receiver back to its original condition.

Here's a comparison of the receiver's bandwidth with the filter in and out (I've raised the filter's trace by 10 dB so that you can more easily see the difference).


(The above measurement was made by driving the receiver's antenna input with a wideband noise source (General Radio 1383 Random Noise Generator), and then, using a FET probe attached to the grid of the first IF amplifier, capturing the frequency response on my HP 8568B spectrum analyzer).

Here's the schematic:

(Click on image to enlarge)

Notes:

1. I don't have a data sheet for the filter. I'm leaving its drive impedance high (it's the first mixer's load, which, when the filter is switched-in, is simply receiver's IF transformer, unloaded). And I've set the filter's load impedance to 2K ohms. Correct? I've no idea.

[Update, 10 Jan 2010: I just came across the following schematic in Service Bulletin 2 for the Collins 51S-1 receiver in which they install a F500 Y60 filter into the 51S-1. Note that the newly-installed AM filter (bottom half of page) uses a total of102 pF (51 + 51) at the input of the filter and 113 pF at the output (51 + 62). Source impedance is a 10 mH inductor (coupled to filter via 1000 pf) and load impedance is 220K ohms (coupled via 470 pF). I've experimented with using a 10 mH inductor that's switched-in (via the relay) to be the load, in lieu of using the existing transformer (as I'm doing now), and I've also experimented with different values of filter input and output capacitance. When I varied the capacitance I really couldn't see any significant change in filter shape/symmetry, and different 1sy Mixer loads produced different gains, but...nothing else of any real significance, so, at this time, I don't plan to change my design. But I'm including the Collins schematic below in case you'd like to experiment, using it as a basis for your design...]

(Click on image to enlarge.)

2. I installed variable caps (and some additional capacitance using silver-mica caps, similar to what is used at the filter input and output on the RT-712 module) at the filter input and output of my circuit, but during my testing I couldn't really find much difference in filter shape and symmetry with the caps installed or not installed. So I removed the additional silver-mica caps, but left the variable caps (which were peaked to give max response). These could probably be removed, too. The schematic reflects the current implementation that uses only the variable caps.

3. Because of the conformal coating on the PCB, you need to remove the filter and its shield bracket together. But this isn't too difficult. I simply clipped the bracket's four mounting tabs flush to the PCB with a pair of diagonal cutters. The stubs of the tabs that remained on the bracket were sufficiently long enough to allow me to easily solder them to my copper-clad board to hold the filter in place.

4. The relay came from the RT-712 module, too.

5. The filter is switched into the circuit when when the receiver's front panel Channel switch is switched to Channel 10, and only Channel 10. For all other channels the filter is out-of-circuit, and the receiver's selectivity is back to its original bandwidth. (Note: The filter could be switched-in for more channels -- simply connect the additional channels at the Channel switch using individual diodes in a wired-OR wiring.

6. The pot is set to give the same AVC voltage, for a given signal, when the filter is in-circuit, compared to when it is out-of-circuit.

7. [Note (7 January 10): I changed the design slightly (raising the source impedance for the source feeding the filter, lowering the gain of the first transistor stage) from when I first published this post, so the photo above of the implementation doesn't exactly match the schematic, and the schematic revision is Rev B, not Rev A. The following discussion pertains to this new revision]

I first designed the amplifier with only one transistor, but when I tested it I found that there wasn't quite enough gain, so I added the second transistor. I didn't rebalance the gains when I added the second transistor, so the first transistor supplies the majority of the gain (41 dB, assuming an Ic of 0.9 mA and a load of 3.9K || 100K || 33K (Av = gm*Rl)). The second transistor is variable gain, and in my case the pot is set to 3.8 Kohms, so the gain of this second stage is about 9 dB (Av = Rl / Rf). Therefore, the overall gain calculates to be about 50 dB.

50 dB is a lot of gain. I don't know why it needs so much. Hmmm...could I have made a math error in my gain calculation above?

Nope. I just simulated the amplifier using LTSpice IV (a great program, available here for free), and at 500 KHz the gain is 49.3 dB in the simulation.

Why so much loss? I don't know, but...I don't plan to investigate any further: the mod works to my satisfaction, and I've other projects to work on!

8. You can find schematics for the 618T series HF transceivers here.

9. And, as a reminder, my earlier post on the R-105A is here.


And of course, the following always applies...

Standard Caveat...

I hope you find this information useful, but please, use these modifications at your own risk -- although they worked for me, I cannot guarantee that they'll work for you. (After all, I could have made a mistake in transposing them from my lab notebook to this post.)

If you do find any errors, or if you have any questions, please let me know. Thanks!


Note:  November 2, 2013:

I've just received a very helpful note from Cliff, WB6BIH.  He says: 

There is no reason for needing extra gain to compensate for the mechanical filter.  I fear that you may have had a defective mechanical filter.  I bought a few of these several years ago, and found one that I had marked "bad" and it was the SSB filter. The wide AM filter I have works fine.

This is all you need; a .01 from the top of the mixer plate to the filter and I used a combination of disk capacitors of about 120 pf to resonate the filter input coil.  At the grid of the first IF I used a 100 pf mica (that probably would have worked fine for both).  There is more than enough overall gain after a second alignment using a Simpson 260 on the AVC voltage as an indicator. The shield cover goes on over this.  Also, larger capacitors on the last two audio stage cathodes will gain quite a lot more signal.  I will dig into that later, mostly to remove all the odd and useless controls and other connections to switches that only keep it from working.

I worked for the Navy before I retired and somewhere I have the Collins source control drawing on these filters.  I will send you a copy if I happen to come across it.  I spent a lot of time trying to get a smooth passband for these mechanical filters by varying the load and source impedances, but I don't think it matters (helps) much.  Just resonate them with about 110 pf in and out. 

This thing is the darndest technical oddity I have ever seen that you could call a "radio" and I appreciate your notes on getting started. 

Thanks very much, Cliff.  This is a simple mod, and I'll need to give it a try!

- Jeff, K6JCA

R-105A/ARR-15 Receiver

[Update (4 January 2010): Additional info on modifying the R-105A to improve selectivity can be found in my new blog posting here.]


I picked up this receiver, along with a companion ART-13 transmitter, a couple of years ago. Both are in "well-used" (beat-up) condition, but...what the heck. I'd been looking for an ART-13, and the ARR-15 intrigued me. And no, the tuning knob isn't original.

Here's a picture of them in the radio operating position of a military plane. (Photo is from this website: 51H-3.)

(Click on image to enlarge.)

Although I'm well familiar with the ART-13 transmitter (having disassembled one for parts back when I was in high-school), I've never seen (nor heard of) the R-105A receiver. It's the military version of Collins 51H-3 receiver, manufactured post-World War II (mine has a 1951 contract date). And, apparently, it was Collins first remotely-tunable receiver (tunable to 10 preset frequencies).

Although intended to be used on 10 preset frequencies, the receiver can also be tuned the "normal" way via a tuning-knob and band-switch on the front panel. Frequency coverage is 1.5 - 18 MHz in 6 bands, and modes are MCW (AM) and CW.

The R-105A is designed to be powered from 26.5 volts DC, and it uses an internal dynamotor (DY-34) to convert this voltage to 220 VDC for tube B+ voltage. If using the dynamotor, I believe an external power supply should be rated at 15 amps, 26.5 VDC.

My radio did not have the dynamotor installed. Instead, a previous owner had wired the B+ line to one of the spare pins on the back connector. My receiver's power requirements are:

• 26.5 VDC (filaments/motor): 1.4A normally, 5A (or a bit more) when Autotuning.
• 220VDC (B+): about 70 mA.

Here are some photos. Despite the relative shabbiness of the exterior, the interior is actually in nice shape.

(R105A, Top View)

(R105A, Bottom View)

(R105A, Right Side View)

(R105A, Left Side View)

Getting It Up and Running...

OK, the only documentation I had was a schematic that I downloaded from the web (see "Resource" section, below). The radio had no dynamotor, but the previous owner had brought B+ out to pin 18 of the rear connector. So I attached a 220 volt supply between pins 18 and 9 of the rear connector ("plus" to pin 18), and 26.5 volts between pins 17 and 9 ("plus" to pin 17). I attached a pair of headphones and an antenna, then switched on the power supplies, turned on the radio's front-panel switch, and...

Nothing. The dial-lights were lit, but I couldn't hear anything -- it was as if the receiver was dead.

I looked at the schematic again and noticed that the resistors in the cathodes of the RF Amplifier and the First IF Amplifier weren't grounded, but were instead going to pin 3 of the rear connector. Clearly they needed to be connected to something (such as ground), but what exactly should this be?

One of the websites I visited mentioned that, in CW mode, the front-panel Gain pot is used to control RF, rather than AF, gain. Hmmm...RF gain as in, perhaps, the cathode of the RF amplifier? Ah ha! A clue!

I noticed in the schematic that there was one section of the gain pot, R139C, that, when the radio was in CW mode, was connected to pin 20 of the rear connector. Could it be as simple as connecting pin 3 to pin 20 on the rear connector?

Yes! I connected these two pins together, applied power, and...signals!!!

There were still some issues, though. I could hear distortion on AM signals, and I could see that, for whatever reason, there was way too much gain -- so much so that the AF Amplifier was being driven into distortion for reasonable-level signals.

When I looked at the R-105A schematic that I had downloaded from the BAMA site, I quickly realized it did not match my receiver. In fact, that schematic is for the R-105 (non-A) version, and there are some significant differences, particularly in the Audio stages. So I traced out my receiver's circuit from the detector up to (but not including) the AF Amplifier. Here it is:

(Click on image to enlarge)

Regarding the distortion and gain issues, my primary suspects were the limiter and the AVC circuits. But I looked at my schematic and quickly realized there were some strange things in the design and that I had no idea how the limiter and AVC were really supposed to function. I poked around with a scope and DVM for a few days but didn't make any headway. What I needed was a good description of how these circuits were supposed to operate. Usually the military tech manuals contain some sort of theory-of-operation descriptions...it was time to try to round one up...

After a bit of searching, I found someone on the web that could sell me a manual reprint (see "Resources" below), and I ordered it. It proved to be quite useful...

The first thing that I discovered upon reading it is that pin 3 of the Limiter stage (V110), during normal operation, should be higher in voltage than pin 8 of the same tube (that is, both diodes are conducting). In my radio pin 3 was lower than pin 8 (despite the fact that the plate voltage of V105A was higher than V107A) and the diode of V110B wasn't conducting all of the time. Oh oh. Cap C133 looked fine -- must be a leaky 12H6. Unfortunately, I didn't have a spare tube in my tube-stash, so I made a solid-state replacement using an octal plug, two 1N4006 diodes, and an 80 ohm, 3 watt resistor (to mimic the tube's filament load -- I made this using 3 power resistors I found in my junkbox). The octal-plug was wired as follows:

• 80 ohm resistor between pin 2 and 7
• 1N4006 Anode to pin 3, Cathode to pin 4
• 1N4006 Anode to pin 5, Cathode to pin 8
  

I plugged it in and...the voltages were now OK! (This mod should suffice until I can find another working 12H6 tube.)

But there was still a gain issue -- during modulation peaks, loud signals would flat-top at the output of the AF Amplifier. (Note: the front-panel gain control does not control the level of the signal fed to the AF amplifier, it actually controls the gain (via attenuation) right at the headphones. Thus it's reasonable to expect the AF amplifier to operate at a high level (to get the best dynamic range), but...it should never go into clipping!)

I spent quite a bit of time exploring the AVC and audio stages...was there too much gain in the audio? Was there not enough gain (or leakage) in the AVC circuit? Or...?

Although there's quite a bit of gain in the audio stages, it looked to me, from the component values and from what I was measuring, that the gain I was seeing was reasonable (and I reduced the gain of the AF driver as much as I could by setting R156, an internal pot, to its max value). I checked the AVC line for leakage or loss (the AVC line drives the grids of the RF amplifier the 1st IF Amplifier) -- it looked fine. Finally, after much poking around, the only explanation I could come up with was that there simply wasn't enough AVC control-voltage being developed to keep loud, highly modulated signals from clipping.

How could I develop more negative AVC voltage?

Looking at the schematic for the AVC circuit, it is is unlike any I'd seen before. Although there is a diode detector (V106A), this is only used to change the signal-level threshold at which the AVC begins operating, rather than, as is typical, developing the AVC voltage itself.

Instead, it is the second section of V106 (V106B) that actually develops the AVC voltage.

It does this by acting as a variable load on the AC-coupled IF signal (coupled to the tube via C123). If there is no AGC action, this IF signal sees R121 (1 Meg) as its load, and R125/C129C low-pass filter the signal across this load.

With small signals, the cathode of V106B sits at about 17 volts (this level is set by the voltage divider formed by R132, R122, and R133). For signals whose amplitude, at the plate of V106B, is less than 17 volts, the tube is in cutoff and, effectively, out-of-circuit. Thus the IF signal only sees R121 as its load, and because the IF signal is AC-coupled and R121 is unchanging, the AVC voltage, after the IF signal has been low-pass filtered, is 0 volts. (That is, the low-pass filter is essentially an "averager", and the average of an AC signal that is symmetric and centered on 0 volts is...0 volts.)

If the signal amplitude on the plate of V106B exceeds the voltage of the cathode, V106B begins to conduct (the amount of conduction is determined in part by the cathode-grid voltage: note that the grid is tied to ground). When the tube conducts, it acts like a finite-valued resistor in parallel with R121, the 1 Meg load resistance, and thus the load resistance seen by the IF signal (coupled via C123) is lowered. Because the tube only conducts on postive peaks, the IF signal sees this smaller load (and thus more attenuation) only during its positive peaks, but not during the remaining part of this signal's cycle. Thus, there is more attenuation for positive peaks than for negative peaks.

Because the positive peaks are attenuated compared to the negative peaks, the "average" of the signal is no longer 0 volts, but instead it is a negative voltage. And this is the AVC voltage.

V106A is used to lower the cathode voltage for strong signals to drive the AVC voltage more negative -- if the cathode is lower than 17 volts, the tube will begin conducting at a lower positive signal amplitude, and thus more of the positive peaks of the IF signal will be attenuated compared to the negative peaks, and thus the AVC will become more negative.

Essentially, V106A acts as a diode detector, detecting the IF signal coupled to it via C132 and developing a negative voltage which, when fed to the cathode of the second stage of V106 (via R123), subtracts from the 17 volts that is normally there (fed to the cathode of V106B via R126). C186 filters out the high-frequency IF signal, leaving only its negative audio envelope.

I needed to develop more negative AVC voltage during loud signals. After experimenting, I was able to get reasonable performance with this simple mod (which can be easily backed-out if one is a purist and wishes to keep their receiver in original condition) :

• Parallel R123 (470K) with a 47K resistor.
• Parallel C186 (470 pF) with a 4.7 nF capacitor.

This modification drives the cathode of V106B lower (on average) than occurs with the stock 470K resistor in R123 (because R123 is smaller, there is less voltage "lost" across it, due to the voltage-divider action that takes place with R126, and hence the cathode of V106B is driven lower (but never less than about 0 volts).

The change in the value of C186 matches the change in R123 and keeps unchanged the time constant of the filter formed by R123 and C186 (which filters out the IF frequency, leaving only the modulation envelope).

It seems to work well. In my listening tests (and measuring with a scope) there is certainly less distortion with the mod than without it.

That's it! Besides that, I haven't changed anything else in the receiver.

Other notes:

Althought the receiver really isn't designed for SSB use, it can be used in that mode, although tuning is a bit too fast.

The autotune is really very cool! (There are 10 channels you can preset.)

The IF is quite broad. It reminds me of using a Command Set receiver.

To Mute the receiver during transmit, add an SPST switch between pin 3 and pin 20 on the rear connector. This switch should be closed during receive and open during transmit.

With its octal-tube sockets and well laid-out design, the R-105A is a real pleasure to work on, especially when compared to typical ham boatanchors in which components are often buried under other components, making access difficult, if not impossible.


Resources:

Schematics here (BAMA site). Yes, they are small and difficult to read. But...they're the only schematics I could find on-line, and they're better than nothing at all. IMPORTANT NOTE: Although the BAMA site lists these as being schematics for the R-105A, they are actually for the earlier R-105 (non-A) version! There's a crystal rectifier detector shown in the schematic in lieu of a detector implemented with 1/2 of V105 (as my R-105A has). And V107 is shown as a 12SJ7 instead of a 12SL7. (By the way -- there's a mistake, too, in the schematics: they incorrectly show R107 connected to the same line as R111 (First Mixer's Cathode resistor). Instead, R107 should connect to B+. And I've no doubt there are other differences...)

AN/ARR-15A feature summary.

51H-3 Good information and a great picture of an ARR-15 / ART-13 pair aboard a P2V anti-submarine patrol bomber.

More pictures here.

Tech Manual: AN 16-30ARR15-3. [You can purchase reprints of this manual (as of 30 Oct 09) from WA5CAB.]

Rear-Connector Pin Assignments (traced from the schematic: click on image to enlarge):
And finally, a reminder that my later post (on improving the R-105's selectivity) can be found here.

Standard Caveat -- take everything I've written with a grain of salt. I could have easily made a mistake.

Thanks!

- Jeff, K6JCA