Reverse-engineering some old SINCGARS military radio boards

[D Straney]

I got some damaged boards (no more of #1, but this is #2 & #3) for cheap because they looked really interesting and I was curious about the design decisions etc. involved in high-reliability radios that are roughly as old as me, from before the era of doing everything in software (which makes for better functionality but much less interesting electronics).

First board is...
ARC-201 Dedicated Remote Panel board

Looking up the IC part numbers here at a 3rd-party-supplier website showed an end use of "RECEIVER-TRANSMITTER RT-1477A/ARC-201": the ARC-201 is a SINCGARS aircraft radio.  Finding a manual for the ARC-201 online showed that the RT-1477A is a remote control panel for this radio, and the "only one [control panel type] without a microprocessor".  The processor back in the radio communicates with this board to read off button states and set the display on this remote control panel.

The DIP ICs at one side are:

• A3012633-1 = NSN 5962-01-081-3931 = 4094, CMOS 8-bit shift register serial-in/parallel-out: this can only receive serial data, so probably brings in control or display signals from the radio
• (2x) A3012587-1 = NSN 5962-01-231-3217 (digital): no info on what this one does; might be transceivers and/or shift registers involved in the serial interface back to the radio's processor?
• A3012581-1 = NSN 5962-01-285-9161 = 4013, dual D-FF
• MM54C922 = 74x922, 16-key encoder: this probably handles the entire numerical keypad on its own, as it's a special-purpose logic chip that scans & decodes a 4x4 switch matrix, inc. contact debouncing, without needing a diode array or other external components. Its output is in the form of a 4-bit number and an "output available" signal.

I'd never heard of the 74x922 before but it's a pretty neat one: must've saved a lot of parts & space back in the day, as opposed to needing to waste extra CPU time or a lot of SSI/MSI logic chips on scanning a keypad matrix and decoding the outputs while handling debouncing, simultaneous keypresses, etc.

The large surface-mount LCC chips are, as would be expected, more application-specific and harder to track down.

• A3012933-1 = NSN 5962-01-304-5002 (digital), CMOS "3 decoder, decimal" in 52-CLFP: high pin count makes sense here, as 3x copies of (4 bits in, 10 outputs) = 42 pins + power/gnd, enable/latch/etc.  This probably does (part) of the display decoding, which turns the 4-bit BCD display digits from the processor into separate signals to drive some kind of display with individual filaments (either directly, or through a diode matrix to a 7-segment arrangement), just like the AN/ASN-112 displays here.
• A3012934-1 = NSN 5962-01-304-0809 (digital): no other info.  This probably does the same thing with the other 2 display digits, and might incorporate some other control or serial-comms circuitry?
• A3012931-1 (digital), closest are -2 = NSN 5962-01-305-8480, or -3 = NSN 5962-01-342-9092: no info on this one either, but again, the serial comms to and from the radio's processor is a big "missing piece" here so this plausibly has something to do with decoding display data.

There's also a crystal next to the A3012933 and A3012934.  The MM54C922, at a minimum, looks like it might need an oscillator to drive the keypad scanning, and maybe serial comms too - although I'd assume that would get clocked by the radio's processor.  This might point to some extra oscillator/comms functions being integrated in the A3012934.

[coromonadalix]

seeing  the damage on the pcb

the pcb have been punched out  to be sure they could not be reused ??


remove the chip's and use them elsewhere ??


and its a  MM54C922    16 key encoder

[D Straney]

Frequency Synthesizer Module (common across models)
There's too much to fit on one board here, so it's split up between two stacked boards:


This generates the LO used for tuning the radio in both receive and transmit modes.  As a PLL-based frequency synthesizer, it should have the following parts:

• A reference oscillator, and a frequency divider to divide it down to some common "comparison frequency"
• A VCO that covers the output frequency range
• A programmable frequency divider to bring the VCO output down to the same common "comparison frequency"
• A phase or frequency comparator to compare the divided VCO with the divided reference frequency

The block diagram from the RT-1439 maintenance manual I found online shows this well:


So let's look at the boards and see where the different functional blocks are...

Board 1:


The top section here behind the shield is the VCO.  The SINCGARS system, according to Wikipedia, covers a frequency range of 30-88 Mhz, and my guess was that the LO frequency was 10 Mhz above the RF frequency, which turns out to be pretty close to what the maintenance manual says, of 12.5 Mhz above the receive tuning frequency (it uses a smaller 7 Mhz offset during transmit).  Therefore, this oscillator has to cover a range of 30+12.5 to 88+12.5 = 42.5 Mhz to 100.5 Mhz.


This is the oscillator itself, with the two black diodes here being the varactors that tune the VCO, from an external tuning voltage.  The part number A3012733-2 = NSN 5961-01-224-5434 = KV2301A, which is a 22V "VHF hyperabrupt tuning varactor".

The 5x coils with vertical tubes are different selectable inductors for the oscillator, to break up its tuning range into 5 sub-bands.  Tracing out the connections shows that the blue diode next to each coil is a switching diode used to selectively connect one of the coils to the oscillator circuit, similar to standard PIN diode switching of RF signals.  I'm not sure if the inductors are selected one at a time (for 5 sub-bands), or are used in some kind of binary combinations (for 32 sub-bands).  Even 5 sub-bands though, if equally sized, reduces the tuning range necessary for the varactor from 100.5/42.5 = 2.4:1, to 1.2:1.  This helps in one way by reducing the range of tuning voltages needed for the varactor, and also helps in a different way by making the frequency hopping faster.

According to Wikipedia, SINCGARS does frequency hopping every 10 ms.  The big limiting factor on how fast you can change frequencies is how fast your frequency-synthesizer-generated LO can re-tune to a new frequency, and settle to a usable level of accuracy (here, the 25 channel spacing is 25 kHz, which is only 250-590 ppm relative to the LO frequency, or 0.03-0.06%, which means that the LO has to settle to roughly 100 ppm or 0.01% accuracy!).  Adding more switchable sub-bands within your entire frequency range adds a "feed-forward" mechanism that speeds up the re-tuning process significantly: rather than having to wait for the frequency synthesizer's control loop to respond to a large step change in the setpoint, the VCO can "jump" to within a much closer distance of the final target frequency, by switching to the sub-band that contains it.  This significantly reduces the "distance" that the control loop has to travel, to settle at the target frequency, and so speeds up the re-tuning process.


After the VCO, there's 3 switchable crystal filters (with blue diodes and adjacent series inductors, to switch between them), and a set of transistor amplifiers to boost the VCO signal and send it to the two coax connectors (mostly removed, but you can see their pads at the top-left of the photo) where the LO travels to both the receive and transmit modules of the radio.


There's also two ICs here: the DIP-8 (SC31459) is listed as a prescaler, although I can't find any data about it as it doesn't seem to be an off-the-shelf part.  This could be the switchable "÷32 / ÷33 prescaler" shown in the block diagram above.  The gold CLCC package is a custom-branded A3012937-1 = NSN 5962-01-225-9116, which is listed as a 7-bit CMOS counter.  This probably forms at least one part of the programmable frequency divider, scaling down the VCO frequency before being sent to Board 2 through the rectangular header next to it.

You can also see the manufacturer's CAGE code (31550) here, which shows that this was made by Harris.

Now on to...
Board 2:

It's immediately obvious that there's much more digital control going on here.

Frequency reference
The first thing that stands out is the reference oscillator, in the giant metal can.  This is a 3.2 Mhz crystal oscillator, temperature-compensated for precision (again, remember that having 2000+ 25-kHz-spaced channels means that this frequency synthesizer needs to hit the correct frequency within roughly ±100 ppm).  Next to it is a separate crystal, which could be the 3.2 Mhz filter used to "clean up" the reference oscillator output that gets used in the transmit mode (although I am a little suspicious of this being listed as an LPF, vs. crystals inherently being BPFs):


Small ASICs
The chip in the gold CLCC package next to the giant metal can is probably the reference divider that divides that 3.2 Mhz frequency to whatever "comparison frequency" is used by the synthesizer. The other similar-looking chip in the corner may implement the PLL phase comparator (which can be as simple as an XOR gate: see the standard CMOS 4046 datasheet) among other things, because one of its pins seems to be involved in the VCO tuning circuit.  Their part numbers are A3012939-1 (different rev. = NSN 5962-01-304-2054) and A3012936-2 (= NSN 5962-01-297-8002), which are both listed as digital ICs, but can't find any other details about them.

Controls
The large chip (A3012947-1 = NSN 5962-01-304-5005) seems to be the "main controller" for the synthesizer.  The block diagram shows a serial bus used for tuning, as well as some kind of frequency table ROM, and this chip likely contains both the serial interface and the ROM lookup table that probably maps the channel numbers to specific VCO-frequency-divider settings.  Its outputs drive the two "2074" ICs in DIP packages, which are the military versions of the standard 2074 quad Darlington array.  Tracing the output pins from these shows that these drive the RF-switching diodes on board 1, as mentioned above, and so are responsible for selecting the 5 VCO frequency bands and the 3 alternate LO filters.

It also likely contains the rest of the programmable frequency divider which started with the 7-bit counter on board 1.  Having 2320 selectable tuning frequencies would require at least 12 bits.  Implementing the frequency synthesizer the simplest possible way would involve dividing both the 3.2 Mhz reference and the VCO down to 25 kHz (because it's the channel spacing, and so every increment is an integer multiple of this), and comparing the two 25 kHz frequencies.

However, I don't think this does things the simple way with a 25 kHz "comparison frequency" and all-integer division factors: the block diagram shows the "÷32 / ÷33 prescaler", which if you look at the math doesn't actually let you select 25 kHz steps with all-integer division factors, and it also shows a "dual-mode counter".  These both suggest to me that it uses a comparison frequency well above 25 kHz, and a fractional frequency divider for the VCO.  You can create a fractional frequency divider essentially by dithering: you change your division ratio for each output pulse (3x cycles of ÷10 and 1x cycle of ÷11 gives you an effective division factor between 10 and 11, for example), and your analog loop filter smooths out the unevenness of the output pulses, if done correctly.  This is called fractional-N synthesis, and you can get a better overview of it here and here.

One of the main benefits of using fractional-N synthesis here is that it allows using a much higher "comparison frequency" than 25 kHz.  The phase comparator's output is in the form of pulses at the "comparison frequency", which need to be filtered through a low-pass filter (the loop filter) to be smoothed into a DC level that gets fed to the VCO tuning control.  The lower the frequency of this LPF, the less ripple in the output voltage (and therefore tuning jitter/inaccuracy & extra sidebands on your LO) remains on the tuning voltage, but also the slower the filter responds to changes, which makes the frequency synthesizer slower to re-tune.  This is an inherent tradeoff that pits the ~100 ppm tight precision requirements of the frequency tuning, against the need for fast re-tuning for the frequency hopping.  Using a higher "comparison frequency" makes for a better set of tradeoffs: with the same low-pass filter, you can get less ripple (= less tuning instability) for the same re-tuning speed, or you can use a higher filter corner frequency and get the same ripple with a faster re-tuning speed.

Frequency control loop
The TL072 dual op-amp and the A3012595-11 (= NSN 5962-01-277-2621 = Siliconix DG201 quad analog switch), along with the array of metal-can transistors, seem to implement the analog portions of the frequency control loop.  I didn't do a full tracing of the connections due to lack of time and being less worthwhile due to all the custom parts, but the array of transistors seems to take phase comparator output pulses from the A3012936-2 and convert these into switched current pulses (typical PLL "charge pump" stuff) into an integrator capacitor next to the TL072, which buffers this voltage and generates the VCO tuning voltage that drives the varactors on board 1.  There's an extra input from the A3012939-1 which is seems to control the current gain, so I'm not sure if this is some kind of loop gain control to keep the loop characteristics stable across the wide range of frequency bands, or speed up the tuning slew rate when the target frequency is far away from the current frequency (I'm not confident enough in the partial schematic I drew to post it).

Two of the analog switches have their controls ganged together, controlled by the "main control" A3012947-1, and seem to be a reset switch for the control loop, probably needed to start it back in a "neutral" position after a new frequency band is selected on the VCO and it's better to start from the center of the tuning range than to have to travel to the final frequency from wherever the control loop last happened to be.  The other two analog switches are also ganged together and seem to do something with the op-amp, likely to implement some kind of switchable control loop gain.

Here's some more close-ups of board 2:




Hope you enjoyed.  Would've liked to do a full connection tracing and draw a schematic for the whole thing, but really do not have the time for that right now.

[D Straney]

RT-1439 Exciter Board


This is part of the transmit signal path, and the RT-1439 manual I found online has a helpful block diagram showing how it works:


Overall, the purpose of this module is to generate a frequency-modulated RF signal at the chosen tuning frequency (7 Mhz below the LO).  This happens in 2 stages:

• 1: The input audio/data signal is first used to drive the modulation input of a crystal-based VCO (VCXO) at 3.9 Mhz, whose output is then mixed with the 3.2 Mhz reference oscillator (from the frequency synthesizer module I looked at before).  This produces a 7 Mhz output.  A PLL is used to stabilize the VCXO's frequency based on the reference oscillator, while reacting slowly enough that the frequency modulation remains intact on the 7 Mhz output even though the "center frequency" remains very stable.
• 2: This fixed-7-Mhz FM signal is mixed with a 2nd VCO that covers the whole tuning range of the radio, to produce the desired RF frequency, with the frequency modulation from the 7 Mhz signal.  A 2nd PLL is used to stabilize this 2nd VCO's frequency, by mixing it with the LO from the frequency synthesizer (7 Mhz above the desired RF frequency), but again reacting slowly enough to not "cancel out" the frequency modulation on the output, while still keeping the overall tuning spot-on.

I'm not an RF expert and haven't designed any high-performance radios from scratch, but it's probably easier to inject a small FM tuning signal (esp. considering the very narrow band spacing of 25 kHz on top of a 30-88 Mhz carrier, which means that the actual frequency modulation range is tiny) into a lower-frequency oscillator, than into a higher-frequency oscillator that has to span the entire tuning range.  Also, adding a certain modulation voltage to a fixed-frequency VCO translates to a consistent frequency deviation in the output, while the tuning control of a wider-range VCO like the 30-88 Mhz one will respond very differently to modulation over its wide frequency range, due to the (necessarily) wildly-non-linear characteristics of varactors.  I'm guessing some combination of these 2 reasons is why the frequency up-conversion was done this way, in 2 stages.

The VCXO is easy to spot, as there's a crystal in one corner with a few transistors nearby:

The part number is A3012956-3 = NSN 5955-01-304-2022, listed as 3.9 Mhz, which confirms this is the 3.9 Mhz VCXO.  The two black diodes right next to the crystal are probably back-to-back tuning varactors, similar to in the frequency synthesizer's VCO from the last post.  The large metal can in the top of this photo is a A3012613 = NSN 5962-01-305-4268, which is listed as a linear Motorola IC, but with no other data I could find.  This is probably the mixer that produces the 7 Mhz IF, implemented as an analog multiplier.


The small A3012936-1 here is the same part that turned up on board 2 of the frequency synthesizer, and which at a minimum seemed to contain a phase comparator.  I think this chip is probably some kind of PLL control building block, with some other unknown functions (gain control on the phase comparator?) which can be set externally.  A phase comparator isn't needed here because the frequency comparison is being done by the mixer.  This, together with our good friend the TL072 dual-op amp, form the control loop for this first PLL that generates the 7 Mhz IF.

The larger chip, A3012938-1 (= NSN 5962-01-340-5405 "digital"), is probably responsible for the digital control of the exciter module.  The block diagram shows the serial interface controlling both PLLs in some unknown way through two "ROM decoders", which are probably contained in this chip along with the serial interface.


This is the second VCO and PLL that produces the final modulated RF signal, to go to the power amplifier through the rectangular connector at the right-hand edge.  The discrete-transistor circuitry at the bottom of this photo should be the 30-88 Mhz VCO, and the "MCL" Mini-Circuits metal can is a mixer, which mixes the output of the VCO with the LO that entres through a coax connector (just barely off the edge of this photo).  The same A3012936-1 PLL controller is visible here too, to control this 2nd PLL.  I don't know how high of a frequency the A3012936-1 can handle, so the SN54LS196 here (divide-by-2 and divide-by-5 counter) is probably dividing the output frequency of the mixer before feeding it to the A3012936-1.  Again, though, its phase comparator isn't needed here as the frequency comparison is being done by the Mini-Circuits mixer, so besides doing other PLL mystery control functions (??) this IC is probably involved in taking the mixer output (which will be in the form of an "error frequency", the difference between the two compared frequencies) and turning it into a DC erorr signal that tunes the VCO.  This could be something as simple as a zero-crossing detector followed by a one-shot pulse generator, with some kind of phase-sensitive section, to produce VCO tuning "increment" or "decrement" pulses based on the incoming error frequency.

The block diagram shows band-switching for this 2nd VCO, controlled by the "ROM decoders", but nothing stands out as obviously as in the frequency synthesizer's VCO.  There are a few diodes in the VCO section (excluding an anti-parallel-looking pair which probably limit the output amplitude), so these might switch between a couple different axial inductors visible, or maybe they switch between different tuning caps instead.

Finally, at the top of this photo, the entire row of transistors along with a few resistor arrays could be drivers used to select different RF switching diodes (as in the frequency synthesizer) to select the "three band LPF" that the block diagram shows as part of the power amplifier.

[Buriedcode]

I love this stuff. Thanks for the explanations and diagrams too. One of those threads I'll sub and read all the way through over coffee 

[Njk]

 It can be noticed that all the boards are not conformally coated. In old Soviet equipment, conformal coating was widely used in every more less serious application. I guess that's because Soviet laminate was of lower density so it was easy for moisture to get inside the PCB causing a malfunction.

[D Straney]

Interesting bit of materials science there!  The frequency synthesizer boards are actually both conformal-coated (the reflectiveness doesn't come through in the photos), but the other 2 are not.