It’s not very often certain very particular pieces of hardware turn up online. Lately I’ve been on a bit of a mission in regards to voice-band modem technology (many more articles to come), so I’ve been emptying my pockets quite literally to obtain various bits and bobs from eBay. Imagine my surprise when a pretty much mint in box Maestro Woomera second generation modem appeared for sale on eBay. It wasn’t cheap, but I put in an offer and ultimately, it was accepted.
The modem itself has a bit of a story. The modem is made by Maestro Communications, an Australian company who was contracted by Farmwide to produce a modem specifically tailored for use over difficult line conditions as would be experienced in outback farms of Australia, where most mainstream modems had difficulty achieving stable connections. The modems were trialled to great success, and made generally available to the public – at one point, even Dick Smith Electronics stocked them. The modem was, however, very expensive at about three to four times the price of competing modems, so you probably wouldn’t have bought one unless you absolutely needed it.
The Maestro Woomera has legendary status as possibly the best modem in the world. Users do generally concur that the Woomera performs well, magazine testing corroborates with it, and even other companies selling other products which interface with modems even recommend it. Despite its steep price, the power users definitely considered it worth the price. I definitely lusted after one as soon as I saw it … but I never did own one.
With such tall credentials, it would be a big disappointment if a high quality specimen were not to be collected in my hands to be documented, tested and preserved. This version of the Woomera is the (so called) new and improved version still available today, so in some sense, isn’t necessarily authentic to the original Woomera that wrote the legend.
The unit comes in a relatively distinctive cardboard box with red printing all over. It clearly shows the Australian background, noting that it comes from within the same state as the one I’m living in – in fact, I’m around 200km from there give or take.
The origin of this modem is slightly mysterious. I bought it from an eBay seller who claims they are an outlet for donated goods to raise funds for charity, so it’s likely this unit was probably new-old stock donated from a company or individual who may have bought it as a spare and never actually used it. The serial number has lots of zeroes in it, so I don’t suspect Maestro sold too many of these in a year.
The main star is the modem itself, which is a sizeable unit, quite a bit larger than most dial-up modems I’ve owned, albeit somewhat light and hollow feeling. If I were to compare it with something classically Australian, I’d have to say it’s about the size of a common boomerang.
The precariously-placed incorrectly-sized product information label seems quite odd for a product of this calibre and price, but it’s probably understandable as Maestro Communications is probably no large giant. Someone’s probably sticking these labels on by hand. This unit has a date of 1st August 2007, making it close to nine years old.
Looking at the back, the PCB layout seems to shove all the ports together in a sort-of amateurish way. It’s just all “shoved” together without any of the space you would expect to see as you do on regular commercial modems. It’s all part of the charm, I suppose.
As we saw from the label, the modem’s not particularly picky about power supply, accepting 7.5v to 9v DC which is just as well, since most farm areas in the outback probably don’t have stable electricity as some of them are on high-impedance single-wire earth return systems. The power adapter is a regular linear supply likely for robustness against surges, branded TE (likely unrelated to the connector company) and outputs 9V DC at 500mA. On the plus side, it seems the modem doesn’t consume much power, and it’s not likely going to suffer from bad caps.
You also get a telephone line bundled with a ferrite bead suppressor to take care of RFI. It’s attached to a detachable modular plug adapter.
They include a DB-9 (or DE-9) to DB-25 serial cable, as well as a Dynalink (unrelated to the modem company) USB A to B cable. In fact, this cable is still in its packaging which reveals that the branding was once the domain of Altronics, a local electronics parts distributor.
No manual is included, but there is a quick start guide which appears to have been photocopied and stapled by hand. It has some very interesting sections including details on the LCD and Maestro-specific commands.
Finally, there is also a driver CD (a burnt CD-R) in a PVC sleeve that’s never been closed up.
Taking apart the modem is not too difficult with just two screws to remove. One is precariously close to the power switch, so some care is needed. The halves of the case part along the seam.
Here, we can see the top lid and its molding marks which seem to be quite confusingly marked. I suspect it indicates the molding has undergone several revisions.
And of course, the bottom half which has all the goodies. It is a slight disappointment to see the inside so empty, as the first generation was likely to have used up all of the internal space. The PCB is also scored along the side with the switch segment connected through two through-hole zero-ohm links. That section is likely to have been removed for their Jetstream M240 model which is no longer available. The front LCD is connected by a soldered down ribbon, and a trimpot likely for contrast control is provided on the main PCB itself.
A closer look at the PCB shows that there was an option to socket the flash memory, but it was directly soldered to the board. There seems to be some Cypress CY7C1018CV 128kB x 8 SRAM for the chipset, which as we can see, is a Conexant with a date of Week 49 of 2006 making it a very late modem. The USB connection is provided by a SiLabs CP2102, with an Analog Devices ADM3307E Serial Port Driver with “Green Idle” providing the serial to TTL interface. It’s good to see that being an “older” style modem, they haven’t skimped on providing ferrite bead suppression on power and phone lines, and they also use a quality Omron relay instead of the silicon DAA switches which are very iffy signal-quality and storm-resistance wise in my experience.
Most interestingly, the logic behind the impedance match network is clear. In the bottom left of the board, four LCA110 SSRs appear to be used to switch different SMD resistances into the network to change the line impedance match. This is apparent because the ATS-091B is a hybrid transformer used to interface to the phone line. As a result, the four settings are primarily determined by the configuration of resistors, so it might be possible to create special versions as well. The MOSFETs could probably be driven by GPIO to connect different resistances to the transformer.
Sadly it seems they did skimp on the voice support, the blanking plate covers a lack of jacks on the inside, and the front seems to be moulded for a mic that isn’t there. Even the large speaker holes at the top are not really used as there is a tiny buzzer internally. The board itself is marked as M204I with a copyright date of 2005.
The heart of a modem is a Conexant chip marked CX06827 which is an ACF3 hardware controller modem. The modem supports coupling with a CX20437 Voice Codec, which would have increased costs marginally, but would have retained the speakerphone abilities of its predecessor. While voice capability is rarely used, on a premium modem like this, it’s sad to see such features being left off.
On the underside of the board, there isn’t much to see although there is some residues left on the board. There is one orange dot label with a signature or some marking on it.
The LCD is a simple 16×2 character display from Truly, based on the normal parallel HD44780.
As per the usual construction, the metal frame tabs are bent back through slots to retain the LCD under pressure on the elastomeric connectors. To accommodate this on the base, some handcrafting was done.
Likewise, the moulding on the bottom half of the case was equally as cryptic.
Support is still available from Maestro Communications’ website, and so I downloaded the latest drivers from there to find that the drivers do not install on Windows 7 x64. I suppose this isn’t unexpected, since the modem predates widespread adoption of 64-bit architectures.
As usual, it was basically a game of grabbing the original inf file, and adding a few bits to make it detect as an appropriate driver.
On the upside, the USB connection is based around a Silabs CP2102, which Windows Update automatically fetches the latest driver for. It’s nothing too special, as it’s just a regular USB to UART adapter commonly integrated in development boards and other products.
Ultimately, it’s a Conexant ACF3 chipset based modem, so there’s really not many surprises when it comes to compatibility, command set, or even the sounds the modem makes during connection. One interesting aspect, however, is the LCD screen.
On power up, and prior to any connection, the screen shows the firmware revision.
Once a serial connection is made, in this case, via the USB, then the modem screen shows READY and the DCE to DTE rate. Of note is that this unit supports 230400bps for faster performance as V.44 compression could theoretically make the DCE to DTE link the bottleneck (think 56k x 6 fold compression = 336k, although normally it doesn’t exceed about 3 fold which is still 168k and exceeds the old fashioned 115200bps serial port maximum).
Once a number is put in for dialling, the screen comes alive with a sequence of messages
Once a connection is made, Speed now indicates the DCE to DCE rate. The Net indicates the impedance network used – in this case, it was still trying to learn line conditions so it changes round-robin on each sequential call.
After 10 seconds or so, then the signal quality (SQ) and line level (LL) indications appear.
Of course, impedance network 2 is not the optimal for my setup which emulates the Australian complex standard, so with the modem reset to use network 1, it managed a blistering 50667bps connect rate over my VoIP ATA (with levels optimized, echo canceller disabled, etc). It’s a very respectable performance especially since ATA-based operation over VoIP is rarely conducive to reliable data mode connection. Other modems generally prefer to connect at 45333bps or slower under the same ATA with occasional 48000bps.
Having the screen is very handy as it lets you check the real-time line stats in a way which is not normally possible. While modems have diagnostic commands, the data cannot generally be accessed without breaking into the data stream, issuing an escape sequence (+++) and then the diagnostic command (AT&V1). When connected through Dial-Up Networking (PPP), there really is no option to do this while the PPP session is in progress. At least, now you can see why your connection is slow.
For good measure, I tried it with a fax as well, and it performed reliably as expected. Statistic figures are not available for signal quality as fax is a half-duplex protocol.
The other exciting part lies in the custom Maestro commands for checking the impedance match statistical data. The command to check the impedance data is AT%MR and gives output similar to this:
Call Records: Mod NW SQ RxL TxL BR+ BR- LRx LTx EQM Wgt Cnt 1 001 001 026 010 011 000 000 047 014 01C6 128 22 2 000 002 025 017 011 000 001 013 014 00E2 083 22 3 000 003 025 016 011 000 001 013 014 00E5 083 22 4 000 004 022 016 011 000 000 013 014 01E4 090 22 Test Cycle 1, Record 4 Modulation used . . . V34 or lower Data network used . . 4 Signal Quality . . . 22 Receive Level . . . . -16 db Transmit Level . . . -11 db Speed changes up . . . 0 Speed changes down . . 0 Last Receive Rate . . 26400 bps Last Transmit Rate . . 28800 bps EQM SUM . . . . . . 01E4 Call quality rating . . 090 Next network for test is 001 OK
In essence, it records historical statistics on your call to analyze and choose the best impedance network in a “learning” phase. In this case, four calls were placed – call 1 has V.90 or above modulation, so has 1 in Mod, other calls ended in V.34 resulting in 0. NW indicates the impedance network used for the call, and SQ indicates signal quality. RxL and TxL indicate receive level and transmit level respectively, and BR+ and BR- indicate retrains upwards/downwards. LRx and LTx indicate last transmit/receive rates coded as a number. EQM indicates the constellation error, Wgt indicating a weighting. I’m not sure what Cnt stands for – a count of something? the minimum quality level?
The last call data is also printed, similarly to what you might get with a regular ACF’s AT&V1:
TERMINATION REASON.......... NONE LAST TX rate................ 26400 BPS HIGHEST TX rate............. 26400 BPS LAST RX rate................ 46667 BPS HIGHEST RX rate............. 46667 BPS PROTOCOL.................... LAPM COMPRESSION................. V42Bis Line QUALITY................ 037 Rx LEVEL.................... 010 Highest Rx State............ 67 Highest TX State............ 67 EQM Sum..................... 00BE RBS Pattern................. 00 Rate Drop................... FF Digital Loss................ 2000 Local Rtrn Count............ 00 Remote Rtrn Count........... 00 V90 9481834246E0 OK
This modem also supports using AT&V1 of course. It also has AT#UD support for unimodem diagnostics but that is printed in a non-human readable form. The human-readable AT#UG command is not supported.
Flashcom32? Yep! New Firmware!
There was a firmware update package for the Woomera available on the site. I downloaded it and proceeded to try it out.
It’s a new version of FlashCom, this one capable of running under Windows 7 x64 even. It doesn’t blindly detect COM ports anymore, and relies on TAPI to find modems. As a result, if you only have the CP2102 drivers installed and not the modem drivers (hacked above), then you will find that you can’t flash the modem.
Even if you did have it installed, as it turns out, the default INI configuration points to the S37 files at an absolute path of C:\flashcom32 v5\, which is not where you would likely have unzipped the archive. As a result, you will find your flash loading will fail, unless you update these paths first. As Maestro doesn’t provide the corresponding unloader, you cannot backup your firmware – this is a one-way upgrade.
I love the extra care Maestro has gone to in rebranding the tool … Anyhow, modem drivers installed and off we go!
The process is quite similar to FlashCom, so I won’t bore everyone with another video. Step 1 is to validate the modem by checking its ATI3 responses match the expected result in FLASHCOM.INI.
Once we’ve chosen the modem, it asks us to check the files we are sending/receiving as the tool could have a library of .s37 files attached to it for working with other modems. In our case, we have no choices.
Normally a stern vendor warning not to turn off the computer, modem or interfere with the process is displayed, but Maestro didn’t bother customizing this, so it’s now just a perplexing set of Warnings.
The files are uploaded as expected, and the modem’s screen nicely changes to say Flashload to let you know it’s in download mode.
Post-upgrade, we can see the suffix of the firmware changed to a g from a c, so the upgrade was indeed successful, and I can sleep soundly knowing the reprogramming will keep the flash chip good for at least another 10 years.
As promised, the modem was in excellent, virtually mint condition when received. By tearing it down, and reading the guide carefully, the impedance match system can be seen to be a simple call-quality driven statistical analysis of connection quality versus which one of the four impedance networks are in use. The impedance match networks are simply switched in and out by controlling a few MOSFETs, likely from GPIOs. As a result, it’s not really dynamically matching impedance nor perfectly tuning the impedance in the same way an antenna tuner for an HF radio might.
That being said, the provided 600 ohm (regular), 900 ohm (x2) and 1200 ohm options seem very suited especially for longer lines, where DC resistance of the loop would increase the impedance, thus allowing better quality matching which would improve signal to noise and echo figures and ultimately improve stability and connect rates.
The LCD display is a nice touch, allowing you to see the DCE to DTE rate and DCE to DCE rates at a glance, and also monitor the line conditions in a way that most PPP users had no ability to do (as they couldn’t interrupt their session, issue +++ and the AT&V1 to check the stats). The one disadvantage, of course, is the refresh rate of the display when it comes to RX/TX indications.
The modem itself is a solid modem, and that mainly comes down to the use of a Conexant ACF3 chipset, namely one of the most regular chipsets around. This means the modem is a true hardware modem with its own intelligence, decoupling it from potentially older underpowered PCs which could interfere with real-time DSP needs.
The included parts seem to show they were sourced from local electronics distributors without much care to branding, and the PCB seems to show that there are commonalities between other models (likely the Jetstream M240) which would have used the main section of the PCB. It seems the second generation device loses out on onboard voice speakerphone capability that was integrated into the first generation devices based on the case moulding.
Was the modem as exciting as I had expected it to be? Sort of. I was a little disappointed to see just how much of it was empty space, the lost voice/speakerphone capabilities and how amateurish the port layout seemed to be. Even Maestro’s efforts in bundling parts and in providing firmware update tools show an Australian level of care. But the modem was in great condition, it performed well even in my VoIP ATA-based phone line simulations, so I’m still happy that I can finally say that I’ve had my hands on the legend (or at least, the son of the legend).