The Ware for December 2010 is shown below, click on the image for a much larger version.
This is a reader-submitted ware; thanks to Martin Ling of the Edinburgh Hacklab for the submission!
Happy New Year to everyone!
[edit: Nobody’s guessed the ware yet — first time this has happened — so I’ll add a hint. The silver box on the top left hand corner whose label has been blocked out is a precision 70 MHz oscillator by Vectron. Hope that helps!]
Hmm, an interesting one.
I’m going to say either a software radio of some type or some type of spectrum analyser / protocol analyser.
Ok, let’s take a look at it.
We have:
1. a massive PCB with no apparent effort put into making it smaller
2. 27 different removable / programmable chips including eproms
3. 9 TI DSP chips.
4. 9 Xilinx chips – probably FPGAs of some type
5. 12 identical connectors with no obvious places for standoffs for daughter boards
6. 8 additional connectors that appear to have been modified to swap the two pin rows.
7. A radio or analog to digital circuit of some type in the top left corner including 2 SMA connectors
8. A *lot* of modifications, especially to the area mentioned in 7 – and damage to the top SMA connector.
9. 9 7(?) segment displays on the bottom left
10. A reset switch
11. What appears to be an RS232 port top right
12. +12, +5 and -12 volt power connector – top right.
13. A surprising combination of surface mount and through hole components including both old and newer style DIP chips below the RF section mentioned in 7.
1, 2 and 4 suggest to me that this is a low yield, highly specialised component of some system. The few thin wires that appear to be to fix up problems with the PCB back this up. (See top left, next to the eprom)
The combination of eproms (2) and both surface mount and through hole ICs (13) suggests to me that this is a fairly old board. Probably dating from the 90s. The orange driver chips for the 7 seg displays back this up (9)
The connectors (5) remind me of the radio connectors on the USRP device that someone was building a while ago. So I’m guessing that these are for connection to something else, but their apparent lack of order and lack of any sort of mounting hardware confuses me.
The DSPs and Xilinx chips (3, 4) make me think that some serious processing is happening on this – which makes the single “radio” (7) a little out of place – I’d expect that there’d be more input for such a large amount of processing.
All the modifications make me think that this has been modified, probably both software (the stickers on the chips) and hardware (the modified connectors (6) and radio modifications (8))
So I’m going to say that this is some sort of signal / protocol analyser, however it’s been modified significantly from it’s original spec.
I’m going to go out on a limb and say that it’s a GSM analyser that’s been modified support different frequencies or CDMA. I’m guessing that this is some part of some cell phone hacking attempt.
It’s very very nice to look at:
– The layout of the RS232 port looks almost cute with the line level converter, capacitors and crystal arranged almost symmetrically in a line.
– The apparent lack of order (the DSP numbering goes all over the board, but the Xilinx chips are arranged numerically almost top to bottom) and the crazy connectors make it strange and beautiful to behold.
– And the mish-mash of older style components (eproms, through hole AD chips, and the 7 seg displays) with newer surface mount components is also kinda confusingly beautiful.
Anyway, it’s a stellar ware to pore over, and I love the way that it’s functionality is so evident without having to look up component numbers. (I’d probably guess better if I did =) ) – Thanks again!
This is a wild board, the number of chips is terrifying.
Lots of EPROMs on there. Let alone 7 TI 16Bit DSPs .
I Guess, from the layout of the board that it is the bottom most board of a series of lab devices. It’s on the receiving side of things, i guess on the left are two antenna ports, one of witch is only a loop back out since there is only one EM-Shielded blob at top. The white plastics a the bottom left are raised up 7 segment displays. 9 of em may be indicating a GHz Range signal. At the back there is a Serial Port going out reinforcing my theory of a lab measurement device.
There are also 10 FPGAs from Xilinx on board
On the board we can see:
* 9x TMS320C50 – 80MHz DSP, labeled DSP1 to DSP9 both on the silkscreen and on the corresponding EPROMs
* 10x XC4010 – 5V, 400-cell, 10k-gate, “3rd generation” FPGA, labeled “COR1” to “COR9” and Master Control
Therefore there is probably one FPGA paired with each DSP.
* 20x EPROM – for DSP code, FPGA config
DSP EPROMs labels hint the product was under development in 1996
* 5 x IDT7024 – 4kx16 dual-port SRAM
* 8x IDT7133 – 2kx16 dual-port SRAM
* 4x IDT71321 – 2kx8 dual-port SRAM
The DPRAMS are probably shared by DSPs and FPGAs and used for buffering / communication
* 12x (Some long Paradigm chip), always paired, probably RAM
* alotof x socketed PLCC PAL (programmable glue logic)
* 2x AD9048 8-bit 35MS/s high-speed ADC with what seems 2x Cypress DPRAM near them
* AD667 – slow 12-bit DAC, uP interface
* some RF analog front-end, probably I-Q demodulator given the 2 ADCs, or maybe the ADCs are interleaved to give a higher sample rate.
* 9 edge connectors, left
* serial port (?), right
* lots of kludges, clearly a prototype under development
So it’s a high-performance nine-core DSP board, probably designed around 1990, with a RF receiver front-end. No transmitter is apparent (no dual high-speed DAC).
* The 9 white things on the left are probably 7-segment LED displays mounted to right-angle (stand-up) dip-sockets. The yellow things next to them are resistor arrays, most likely the series-resistors for driving them. Alternatively, they could be configuration switches (with the resistor packs being pull-ups, I cannot determine the resistance value from the scan).
* The 26 MHz Quartz Y2 would hint to something GSM related (typically everything there is related to a 13MHz clock). But normally it’s derived from a external timebase with a PLL, or using a high-recision oscillator which Y2 clearly isn’t (according to the datasheet it’s also not a voltage controlled oscillator as used in a PLL).
Y1 appears to be a temperature controlled crystal oscillator (TCXO). There’s your precision source. Unfortunately, the markings have been blacked out, so I’m guessing the details of make and model would giving away too much information.
Looks like there are 10 Xilinx XC4010s (labelled COR1-9 and MST CNTL) and 9 TI TMS320C50s (labelled DSP1-9). I’d hazard a guess that each DSP is linked to a COR, and the MST CNTL coordinates all 9 units. I’d say the parts at the bottom left are connectors of some kind, supplying a channel to each DSP/COR unit. If it needs 9 seperate DSPs and 10 FPGAs, it’s doubtful that it’s using the serial port for any real data I/O, so I’d imagine that’s what the two gold plugs (coaxial, maybe?) at the top left are doing.
That integral-of-dt logo might belong to IDT, in which case those chips are all probably SRAM. Assuming I haven’t missed anything (and I probably have), it looks like there is 656 kbits total on board in varying widths and densities.
If I had to guess, I’d say it *might* be a WildChild board from Annapolis Micro Systems (judging by this: http://webcache.googleusercontent.com/search?q=cache:http://www.ece.northwestern.edu/cpdc/Match/accomplishments.html) , if only because they’re the only people I can find who’ve built and sold a DSP board with 9 XC4010 FPGAs on it. That said, I haven’t been able to find a picture or even proper diagram of the board, and there doesn’t seem to be enough RAM on board going by the 2MB claimed on the aforementioned page.
My guess is it’s an old (ooollllddd) GPS dev board or lab/calibration/debug tool for GPS. Guessing that the CORs are correlators. Although the (horrific!) rework around the RF entry makes me question whether or not the front end could actually detect a real GPS signal. Possibly there is an offboard LNA/frontend and the inputs are quadrature either at an IF or at the native frequency.
I see that Analog currently sells the AD15252 (dual 12bit, 65M/s rate) for use in antijam GPS receivers. So for application in GPS it matches two 8bit 35M/s nicely.
On the other hand, in 1996 or so GPS receivers where down to the size of a cigarette box already, so I cannot really believe that this huge board would have to be built to replicate the functionality.
My first thought was GPS (or other global nav, such as GLONASS) related. Maybe someone wanted to proto before doing an ASIC? I read the multiple high density connectors as debugging aids for a scope or logic analyzer.
Second thought was cell site base station equipment, but as razvan784 pointed out, no fast DACs, so it looks like it’s just the receive side.
What’s strange to me is the I/O split: input side is high frequency (SMA connectors, so obviously RF), but output is apparently just the UART (slow) and the 7 segment displays? Prototype of something (turbo code receiver?) for the deep space network? That might explain the I/O balance, slow data rate, and fairly decent ADC rate.
Wikipedia shows turbo codes in ’93, so I’m liking that…
BTW, all those old EPROMs next to the TQFP packages hurt my eyes. Why not slap some flash on there?
My first thought upon seeing all the DSPs and FPGAs is that this board is a Software Defined Radio board, so I’m going to continue under that assumption.
During the mid 90s, there were a couple military type radios under development, Speakeasy phase I and phase II. I’m guessing this is some sort of test/debug board for phase II. Here are some reasons that lead me to this guess:
1. Under development in 1995, produced a product in 15 months. Matches the dates on the stickers over the EPROMS, indicating dates that code was programmed into the boards.
2. According to the Wiki Article, there were 9 modules, corresponding with the 9 DSPs on the board.
3. I think some of the control modules mentioned in the IEEE article were implemented in the FPGAs, or on a higher board in a stack of boards.
4. The DSPs on the board are similar models of the ones on the Phase II project.
There are also more modules, as well as input and output for voice, which leads me to believe that this is just a bottom board of at least a 2 board stack, using the connectors on the board to interface with the next level. I also think that is is probably a development board, not a production board, based on the patches.
There also appears to be no input/output for voice signals, I think that would be on a board above this in the stack. The serial port is probably for control/debugging, not for data.
Now here are some challenges to my assumptions:
1. Military hardware prototypes, especially ones dealing with comms/encryption may not be available to the public at this point. It’s only been 15 years since this board was under development. This leads me to believe this was some sort of proto/dev board that somehow got released.
2. There was supposed to be clear separation between RED/BLACK sides of the design, I can’t say that I can see this on the board. Perhaps it happened on a board higher in the stack.
Here is my source for some information: http://ieeexplore.ieee.org/iel4/49/16463/00761042.pd (need subscription, sorry)
So final answer: Speakeasy Phase II dev board, bottom RF and SDR board in a stack.
More on Speakeasy:
http://211.232.57.48/Documents/document/EN/COMMUNICATION/radio%20control/PDF/bons_s.pdf
Phase I was VME, phase II PCI. If this was an early proto, maybe they didn’t have the bus worked out yet?
Interesting guess, but I’m with Julian Calaby as I don’t think you’d stack board on those connectors without standoffs for mechanical support.
And what’s with the 7 segment LEDs, but no buttons for user input?
My guess is that this is a Rake receiver http://en.wikipedia.org/wiki/Rake_receiver. Each DSP/FPGA pair is dedicated to a finger and the resultant data rates for mid 90s CDMA or WCDMA cell phone technology could be output over a fast RS232 port. I imagine that the LEDs are used to display the center frequency of the carrier or the status of each finger.
It is some kind of precision clock, probably using GPS as at timebase (as opposed to WWVB or shortwave). I suspect Y1 is a TXCO so it can maintain time accuracy down to parts per gazillion.
There were handheld GPS, but those didn’t do strange things like check for phase variatioins in the signal (it may be merely a military version that does decryption, but I’d doubt it) which can allow for precision GPS or RTK (realtime kinematics).
Is it a receiver for something like GPS timing or atomic clock radio timing signal?
I’m wondering if this might be a part of a test jig for a mixed-mode analog-digital ASIC.
The leds are usually for displaying POST messages and displaying error codes for the tests.
Y1 might be a rubidium oscillator which is usually used in tv systems, cellphone and gps.
I like the way they used bits of solder wick to properly ground the caps of the smaller oscillators.
This might be a random number generator.
You don’t need a board for a random number generator:
http://xkcd.com/221/
http://dilbert.com/strips/comic/2001-10-25/
:-)
So, I’ll chime in again. After pondering a little over this riddle, I want to make the following long and rambling remarks:
The board looks quite hastily developed, with the “Master Control” and “DSP” boards just put arbitrary on the board, and then autorouted. Whoever designed this part was sure that the DSPs will work on the first try without much effort and just wanted to get this done as fast as possible. If some actual research would have taken place, the board would look much tidier.
* The analog section obviously took some rework to get running, with the ground-lugs soldered to the SMA(?) connectors, most likely to have a good ground for attaching a oscilloscope probe when working there.
There has been additional grounding done on the metal cans, which might be mixers to mix down the RF-input with the output of VCXO(?) Y1 to the 35 MHz bandwidth of the ADCs. Some trimpots have been added around the OP37s (40 MHz G*Bw Opamps), possibly to tune the gain in front of the ADCs: With only 8bits you don’t want to give away resolution by starting with too low of a signal.
So my guess here is, that whoever built or used that board in the end found out that signal-to-noise in what was supposed to be a non-critical circuit (only 8bit, no heavy-duty shielding or separate ground planes) was not as good as he had hoped and had to tweak.
The single DAC might be used to either adjust the (?) VCXO Y1 feednig the (?) down-mixers?
* To the “digital” part:
The small 14pin jumper-headers with the bent pin for sure have been used for /something/, and it wasn’t just for connecting an oscilloscope, because the order of pins wouldn’t have been critital to warrant the green-wire adjustments that have been made to swap it. So I’m pretty sure that some device with a matured design connects to these, and you weren’t able to just insert a “pin swapping” adapter cable for some reason.
There are 9 almost identical sections of DSPs/CORs, but they have roms with distinct labeling (those that haven’t already fallen off, I presume…):
DSP4 0x03055e 0x030669 delta=267
DSP5 0x020edc 0x020fc5 delta=233
DSP6 0x020eda 0x020fe3 delta=265
DSP7 0x020f51 0x02105a delta=265
DSP8 0x1a18d0 0x1a19b9 delta=233
Two hex numbers with similar ranges, they also seem to overlap…
So I’m pretty sure that those 9 sections did the same numerical work, but just differ by the part of the “spectrum”(?) they were programmed to analyze. But why not just make the number a configurable parameter? I don’t know? Maybe something like special-purpose fft- or convolution routines have been generated for these exact coefficients and lengths?
* So to make up a story that could make plausible how such a board could come into existence:
Someone had a signal analysis problem which was understood in principle and prototype hardware already was working and debugged (not *this* board). Only it wasn’t scaled to the required number of parallel channels yet, and this additional computing power was needed fast.
The signals to be analyzed were not too low (not much amplification visible) neither was the SNR too critical (only 8bit ADCs, no effort put into schielding).
Output most likely was through the 9 reworked jumper-headers with the green wire-nest below, they match exactly the 9 DSP sections. The big black high-density connetors to me serve no obvious use, they have way too many pins for the expected low data rate (after all it’s only 2×35 MHz raw input data, and it will be preprocessed by the DSPs, won’t it?)…
The reworked connectors are likely for a JTAG or programming/emulation pod. The 9 displays may be one status display per DSP? It’s interesting that the layout for the 9 DSPs are different; if it were a 9-channel something, I would expect at least the component placement to be similar from one to the next. This hints to 9 separate steps, or someone used an autoplacer.
/mike
My wild guess is a component of a radar system. I have no facts to back this up :)
Going to guess that this is an early GPS-disciplined time/frequency standard with an eight or nine-channel receiver. The displays show days:hours:minutes:seconds. One of the SMA jacks takes the signal from the GPS antenna, while the other puts out a sample of the frequency-corrected OCXO signal.
It may have come out of the R&D department at Datum or Trimble. One of the most interesting clues is what isn’t shown — namely, the nomenclature on Y1. If that’s a 5 or 10 MHz part, or if it was labelled with the name of a major GPS vendor like Trimble, the submitter might have guessed (correctly) that the challenge would be too easy.
Well, Y1 is an OCXO but it’s frequency is obscured. If it where 10 MHz it would not say much, but if it is some other frequency it would be such a give-away. However, the Y2 is probably a VC-TCXO and frequency of 26 MHz *is* a give-away.
The FPGAs labeled COR1 through COR9 is correlators. It if would be a GSM receiver it is the wrong age (1994) and also would correlators not be relevant for GSM reception (which is GMSK), but it would fit in nicely for a EDGE or even UMTS receiver research prototype. I’m a bit weak on detailed knowledge of that radio interface, so I am not able to pin-point closer than that from the top of my head.
I’m guessing it’s something from a PBX based on the form factor, all those DSP chips and the power connector on the rear. It doesn’t have a big backplane connector, so I don’t think it’s primarily designed to shift data to other cards, but would sit in a PBX chassis for convenience, showing the LEDs at the front and with the reset button to hand.
cell phone network base station of some sort.
I think its some sort of satelitte receiver development board.
(70MHz / L-Band) maybe?
Well, 70MHz is a common satellite and microwave IF frequency, so we’re back to a radio application. The front end is elsewhere, and connected to this board through the top SMA. Possibly a radar or radio telescope application?
Early Wi-Fi test board maybe?
VideoCipher encoder?
try 2 … ham radio board ? 70Mhz is common for ham radio applications as well as SDRs.
Is it too old to be a proto for one of the direct broadcast satellite receivers, such as used by DirecTV?
Or maybe a proto for XM/Sirius radio? (Again, I’m thinking the board is too old for this.)
Proto hardware for part of the Iridium Communications network? The timeframe seems about right.
Could it be the front-end for some type of digital communications receiver (based on QPSK or QAM)?
The hints seem to be:
– 2 inputs (I and Q)
– the obvious “I” label left U29, which is a 35 MSPS flash A/D
– correlators are certainly used in FEC – implemented in DSP
– 7-segment LED displays used to show symbol/data rate
– the multiple IDT dual-ported memories imply different stages operating at different data rates, or needing to be fed packetized data chunks
– 70 MHz is a common IF frequency used in satellite communications
Conceivably, it could also be a modulator board – by running the above process in reverse. But, the A/D does not fit the equation then – unless I misread the part numbers…
Radio astronomy or SETI receiver? Still seems like there’s not enough output for those apps, though.
I’m with Joe here… if it’s not radio astronomy or SETI, it could be a deep space network receiver. I recall that the SETI receiver outputs a disgustingly low bit rate, after analysis. Deep space network would require tons of front end to analyze multiple doppler and chirp rates simultaneously. SETI, RA and DSN would all use rake receivers to recover signal under difficult conditions, while SETI and RA would use them to keep bitrates down on the output despite wide signal inputs.
Satellite modem? Perhaps military?
I’m going to guess it’s part of a PET scanner. These devices depend on being able to correlate the reception of particles 180 degrees apart to localize events (generally within tissue). Perhaps the scintillators deliver their data to the multipin connectors near the DSPs, which compute the correlations. A fair amount of math is done to “estimate” actual pulse timing from the known pulse shape and the sampled data.
main board for the Chumby Three
This is the receiver end of an attempt to cram nine conversations into one 4-meter carrier frequency, probably using single-carrier frequency-division multiple access.
In the early nineties, several researchers[1][2] were interested in computation-intensive signal multiplexing techniques, especially as applied to low-bandwidth channels. Someone must have noticed the emerging international popularity of the 4-meter (70 Mhz) ham radio band, as more and more countries ceased commercial VHF TV transmission. In 1993 and 1994 almost no 70 Mhz equipment was available for amateurs. This band was the sparsely populated home of personal equipment designers and builders. I have heard that most of the traffic on 4 meters was about how to come up on 4 meters.
I’m guessing that this board was made as part of that general climate of experimentation.
-Wang-Lo.
——–
[1] J. A. Bingham, “Multicarrier modulation for data transmission: An idea
whose time has come,” IEEE Communication Magazine, May 1990.
[2] J. M. Cioffi, “A multicarrier primer,” Online,
http://www.stanford. edulgrouplcioffilpdflmulticarrier.pdf, 1991.
I’m so dumb I could kick myself.
I have been staring at this thing for a month and it only just now occurred to me to ask: if the board was built before 1994, and one of the EPROMs bear a sticker dated 3/8/96, how were any of the EPROMS reflashed, if they’re soldered to the board?
At the same time, I’ve been baffled by the presence of the 12 HD connectors scattered around in a way that would make quite a tangle of the 12 ribbon cables plugged into them. But there would be no tangle if only one connector at a time had a cable attached. Like, say, from a ROM burner.
-Wang-Lo.
Don’t kick so hard, Wang-Lo! Those EPROMs appear (to me anyway) to be socketed. Look at the ones in the top left, for example.
Your Stanford link makes me wonder if this ware could be an early DSL modem prototype, but again, the output bandwidth seems limited by that DE-9.
Thanks. I thought I could see little blobs of solder, but now I can’t make up my mind. I should have paid attention to the length of the shadows cast by the EPROMs, which indicate that they’re too proud to be soldered.
So I’m still baffled by those 12 connectors.
The Stanford link is to the cited papers. Bunnie says the board came from Edinburgh.
-Wang-Lo.
I think the 12 connectors were meant to support an upper-deck PC board. One peculiar thing about this board is that the power entry connector looks a bit wimpy for all those chips. It may be that the Samtec-style connectors brought power as well as data connectivity to the board.
One thing’s for certain, this was no ham project. Hams are cheap by nature, and this thing wasn’t cheap.
The trouble with the daughterboard theory is that it practically blocks the nine 14-pin (well, 13-pin, after mutilation) connectors.
I concur with your other points.
-Wang-Lo.
GPS Signal Simulator ?
could be a ham radio repeater but multiple conversations i dont get. maybe multiple frequencies ?
most repeaters just handle two conversations at the same time.
I have been trying to make this into something that connects the 4-meter ham band to the internet, like a computer-controlled conversation concentrator, or a very early prototype super-reflector. But every application I can research for it always leaves too many parts left over.
So it is time for Wang-Lo wild guess #2: This is the front end of a test instrument used to verify the operation of 70/140 Mhz IF satellite modem prototypes.
I base this guess on three observations.
(1) If it was a prototype itself, it would probably have a 140 Mhz TCXO and a frequency divider instead of the 70 Mhz;
(2) It has lots of connections to report in-band and diagnostic data (especially 12 30-pin ribbon cable jacks left over from guess #1); and
(3) Severe damage to the top SMA connector such as might be caused by several years of distracted technicians tugging on a probe lead.
In any case, it is a stellar puzzle ware, and I expect to study it for another month after Bunnie tells us what it is.
-Wang-Lo.
Hello
It could be a (very) early prototype of some radar receiver, doing some correlation.
From the construction style, and the fact that it’s available 15 years after prototyping, i would rather exclude a military use.
Could be acaddemic research, rather, for example :
http://www.era.lib.ed.ac.uk/handle/1842/4074
It seems edinburg is an active place for radar technology…
This is great fun. I don’t know much of the theory behind digital hierarchical broadcasting receivers or DSL modems which I think, this ware could be a prototype board for.
Some of the EPROMs is marked with a label with the text THP and THP could mean Tomlinson-Harashima Precoding.
–cut–
• Independently developed by Tomlinson and Harashima in
1971.
• Uses a Decision Feedback Equalizer at the transmitter
instead of the receiver
– receiver computes DFE coefficients during startup and sends
coefficients over to transmitter
– advantage – allows for block processing and decoding at the
receiver.
– advantage – reduces complexity of receiver analog front end.
– drawback – increases complexity of transmitter.
— cut —
-Morten
If I had to guess, I’d say this looks like an early prototype built by Hunt Engineering. They’re based in the UK, even though Bristol is a fair distance from Edinburgh. They’ve been DSP/FPGA weenies for a long time.
70 MHz is a common IF. It could be for radioastronomy, which would also explain why there is only a receiver and no transmitter. 70 MHz is also used for satellite uplink/downlink and terrestrial microwave, but I’d expect both TX and RX, except for RX only satcom such as GPS.
As an official guess, I’m going to say a Hunt Engineering prototype for radioastronomy.