[Post-publication note: This article first appeared in Blue Notes, magazine of San Francisco PC Users Group. Unfortunately, the editors rewrote it in places, distorting my meaning and introducing a large number of grammatical and punctuation errors (plus several errors of fact). Accordingly, I have withdrawn all publication rights to the article as printed. This original (and correct) version may be republished by anyone, with credit to the author.

Yes, "Ken Beseder" is a pseudonym. Two points if you spotted the rather arcane bilingual pun.

Bio note: Ken Beseder (beseder@mail.sfpcug.org) lives in a snit in San Francisco.

Hayes Optima 336 Business Modem (External)

Hayes Microcomputer Products, Inc.
5835 Peachtree Corners East
Norcross, GA 30092
770-840-9200 (main desk)
770-441-1617 (cust. svc. & support)
770-449-0087 (fax)
770-446-6336 (BBS)

This product absolutely screams "corporate market": Packaging, literature, and of course the premium price tag put it in that league. The good news is that it's an offering worthy of that niche.

Open the snazzy cardboard box advertising data rates "up to 230,000 bits/second" ("Hah!", I said, and "Hah!" again), and out fall the obligatory AOL, Spry Mosaic, and Prodigy sign-up packages, a Command Summary pamphlet, yet another anonymous power cube (mark it "Hayes 33.6" before you accidentally use the wrong one — a Very Bad Thing), four diskettes, and finally the modem itself in egg-carton packing. The disks are Hayes Smartcom LE for Windows (a "lite" Win16 terminal and fax package), Hayes Smartcom Message Center LE for Windows (voice-mail), and a "Modem User's Reference" disk.

The software is, well, very corporate — polished in appearance, and to all appearances quite competent. If you don't already have good modem software, this isn't a bad start (though, in my view, if you need voicemail, you need a more solid system than any modem is likely to provide). In my case, like most User Group types, I already had my favorites, and so concentrated on evaluating the modem itself.

This is the Hayes company that defined the modem as we know it. It's good to see them back, after recouping from Chapter 11 bankruptcy in mid-'96. It might be useful to consider how they got where they are.

Back in the Mists of Computing Time, some of us used modems that you plonked a telephone handset onto, to make them work. They ran at a breathtaking 300 bits/second, and "modem commands" were unheard of: You dialed by using your fingers.

Things weren't much further along when Hayes introduced the Smartmodem in 1981, which popularized the notion of controlling your modem through software. It responded to "AT" ("attention") modem-automation commands to make it do things, which innovation proved wildly popular.

That's all "Hayes-compatibility" means, really — responding to "AT" commands: There's so little standardization that you can count on "ATDT" doing dialing, and that's about it. Hayes made much of its "Hayes Improved Escape Sequence with Guard Time (U.S. Patent #4,549,302)" a few years ago, suing to compel license fees. It's just a simple way to get the modem's attention when it's on-line: pause, three plus symbols, then pause again. Of such things, lawsuits are made. There are alternatives, such as dropping the DTR (Data Terminal Ready) hardware signal, so it's not much of an issue.

Throughout the 1980s, other modem makers took the lead, making myriad, mutually incompatible extensions to the basic Hayes command set. Hayes continued to solidly appeal to the business market, though, and, around the time of the Chapter 11 reorganization, divided its Smartmodem product line into an "Optima" series (high-performance) and an "Accura" one (for the home market) — plus their Practical Peripherals subsidiary for the budget-minded. The current series serves
notice that Hayes is back.

Actually, this particular unit started life as an Optima 288 Business Modem, but Hayes kindly makes available a free upgrade to 33.6 available over the Internet at ftp://ftp.hayes.com/modem/3d66.exe, that rewrites the modem's flash ROM and soups the beast up. That's what I call service.[1]

Also in the good-service category is their customer support. The customer service and technical support people knew their stuff, and were prompt and efficient. Again, this is just right for the business market. (The five-year warranty doesn't hurt, either.)

The Hayes, internally, has one neat, well-made, surface-mounted circuit board with the aforementioned flash ROM chip and a Rockwell RCV288DPI chipset. There are the usual two RJ-11 telephone modular jacks (the second for an extension telephone), a DB25 for the serial cable, and microphone and speaker mini-jacks for the voicemail functions.

There are a number of bell-and-whistle functions: The Optima can report call Caller ID information, if you have appropriate software. It can accept reconfiguration of itself by calling modems, if so desired. It can implement password security on incoming and/or outgoing access. It can handle "distinctive rings" to answer multiple telephone numbers. Some of these functions are (I think) better implemented in software, but it's nice to know they're available.

When you're done admiring the hardware, you look around for the manual. Then, you spy the "Modem User's Reference" diskette. Oh. The disk's the manual. It's for Windows, and the only way you can print it out is one tiny page at a time. Ouch. However, the tri-fold Command Summary does the trick for most purposes.

Alphabet Soup Time

The Optima is classified as an "ITU-T V.34" modem. That means that it meets the International Telecommunications Union - Telecommunication specs for operation at a 28,800 bits/second modulation rate (plus the optional 33,600 and 31,200 rates), and does the implied V.42 error-correction and V.42bis hardware-level compression. Hayes implements a Hayes-only superset of V.42bis, that's claimed to reach twice V.42's theoretical-maximum 4-to-1 compression (when supported at both ends). That's how they claimed "230,400 bits/sec speed" on the box: 28,800 x 4 x 2 = 230,400. You'll never see that in the real world — or much else that's claimed about high-speed modems.

While V.34 was being worked on by the ITU-T's predecessor, the CCITT (Geneva's Comité consultatif international télégraphique et téléphonique), two quasi-standards enjoyed a brief vogue: V.FC, a 28.8 kbps spec from Rockwell, and V.32terbo, a 19.2 one from AT&T. They're still around, so the Optima supports V.FC.

However, the fun doesn't stop with V.34's 33.6 limit. U.S. Robotics (along with Texas Instruments) has announced that it will make available upgrades to 56 kbps ("x2" being USR's trademark for this), later this year.* In response, Rockwell, Lucent Technologies, and Motorola announced a competing standard, to come out about the same time — and Motorola promptly filed suit against USR for patent infringement. Hayes's (Rockwell) upgrade from 288 or 336 will cost $99 (or free if you're a recent buyer); USR's deal is about the same. There won't be an ITU-T V.something standard for years, so the competitors will have to slug it out in the meantime.

The 56 kbps war is largely madness, anyhow. There's an inherent limit of about 35 kbps in voice telephone lines (described in mathematical information theory as Shannon's Limit). You can go faster, in theory, but only by assuming that the noise floor is lower than usual. Already, it's very common for V.34 modems to "step down" in real-world situations, to 28.8, 21.6, or lower speeds.

To the extent it works at all, there will be noteworthy limits on 56k operation: The far end must be all-digital, there cannot be extra analog/digital conversions between you and the destination, and all parties warn that it "depends on the local loop" (i.e., the quality of local telco service). Those limitations** plus the fact that the higher throughput will be receive-only mean it will primarily be used for Internet access, and, especially, Web browsing. USR/TI already have an advantage in Internet Service Provider acceptance, so, for that and other reasons, most are betting on them.

Any Port in a Storm?

With any of these modems, present and future, you have to decide what to plug them into. Your existing serial ports may not handle the data flow: Fire up MSD from a DOS session (not under Windows), to check. If the port has a 16450 or 8250 UART (Universal Asynchronous Receiver-Transmitter — the port buffer chip), you're in trouble and need an upgrade. If it's a 16550, that's fine for 28.8 and arguably for 33.6 (but barely).

A 16550 chip (preferably a premium-quality 16550A or 16550AFN[2] chip from National Semiconductor) can run the serial port at 57.6 kbps. Remember, modern modems do data compression at the hardware level: V.42bis tops out at a theoretical 4:1 maximum ratio, and Hayes claims twice that (between recent Hayes modems). If you had a 33.6 kbps modem connection (a very noise-free line: ideal conditions), in theory that means the serial port might be asked to handle 33.6 x 4 = 134.4 kilobits/second, or 268.8 kbps for a best-case Hayes-only connection. That's a lot of data!

In the real world, 1.5:1 or at most 2:1 is about all you get; a little more when sending pure text, much less while sending compressed files. So, a 57.6-capable serial port is actually within reason, and many 16550-equipped ports will even do 115.2.

However, if you want to make sure there's no bottleneck, there are options: Hayes will be glad to sell you an Enhanced Serial Port (ESP) v. 2 card, capable of 921,600 bps and possessing 2048 bytes of buffer[3] (versus 16 in a 16550). ESP cards require special drivers, and cost about $90 (one port per card). There are several competitors with similar products, and soon there will be a "Universal Serial Bus" port on upcoming Intel-chipset motherboards, that will similarly provide surplus capacity.

The ISDN Option

The 56 kbps figure should ring a bell with many, as that's ISDN (Integrate Services Digital Network) territory. Specifically, it's the home-service "single-line ISDN" figure (either 56 or 64 kbps, depending on telco needs) whereas full business-grade "Basic Rate ISDN" is 128 kbps.[4] In either event, ISDN's an all-digital service over ordinary telephone wiring. ISDN equipment ought to be cheaper than modems, because of its much simpler circuitry (lacking modems' complicated encoding and digital/analog converters), but isn't yet, because of manufacturing scale. ISDN also cannot do hardware-level compression.

Why use ISDN instead of modems? One, modems are twitchy. Because they have to work out all the myriad protocols each supports, when you connect, the initial "negotiation" phase can take up to a minute, and then they can fail to connect, or "fall back" to slower standards. With ISDN, it always works, it's always fast, and negotiation takes 5 seconds or less. Two, ISDN has far superior "connection latency" — response time when data must travel both ways, such as during any interactive usage (e.g., keyboard echo, when typing across a connection). Three, of course, ISDN provides more bandwidth (information capacity) than do most modems. Four, ISDN is simply more elegant, eliminating completely the four conversions between analog and digital that a modem link requires.

However: Installation is expensive. (Watch for promotional deals, to help.) It won't help if the other end of your dial-up is a bottleneck already, e.g., an overburdened Internet provider. Also, expect to pay more for ISDN service from such providers. (They're no fools: They know you expect to draw more on their capacity, and charge accordingly.) For that matter, providers may not have ISDN service, yet. Last, but most important, ISDN is really a chore to set up, often requiring long hours working with PacBell.

Further down the line will be something called ADSL (Asymmetrical Digital Subscriber Line), a digital technique still in testing that will furnish download rates in the megabytes/second, with the reverse channel running at ISDN speeds, still using ordinary telephone wiring. Don't hold your breath waiting, though.[5]

Miscellaneous Recommendations and Tricks

Modem configuration is an art. Most people ignore the issue by picking a driver that's assumed to cover everything, which implicitly sends someone's idea of a good initialization string (a group of setup directives) to the modem. You can indeed just pick "the" driver, and then whistle in the dark about potential problems.

However, what's good in one situation may not be in others. Also, the "driver" string may be for an earlier (or later) modem with the same model designation. For instance, many very different modems have been called "U.S. Robotics Sportster", and two different ones have even gone by the exact name "USR Sportster
28.8".

Some people go around urging others to adopt their extremely long init strings, purporting those to be applicable to some range of modem models. For the above reasons, this is an extremely silly habit, and they're doing their audiences no favor.

Nonetheless, here are a few settings that work for me with this particular Optima modem. Your own modem's adjustment can be reasonably done only by reference to that modem's own manual. Anyone who tells you otherwise is deluded.

AT M3 N1 W1 &B1 &C1 &D2 &K3 &Q9 +FClass=1 S11=45 S95=43

That tells the Optima to enable speaker after dialing until connect, connect at the highest modem speed possible, show call-progress messages, use any workable modulation method, don't ignore the carrier-detect signal, hang up if the computer so signals on the DTR hardware wire, use hardware flow control, use any workable error-control method, use Class 1 fax (more conservative), dial using 45-millisecond touch-tones (the default 70 is slow), and enable modem-speed, error-control, protocol, and compression result codes.

One can enter that string in a terminal (modem) program, then type "AT &W0 &Y0" (write current state to stored profile #0, and set modem to read profile #0 on power-up). Having done that, you can then use a generic modem driver to avoid having some driver interfere with your settings. (Disable Plug'N'Play, if that applies, or your machine may "find" the modem and insist on substituting some fascistic driver for the generic one.)

Don't forget to disable Call Waiting, if your telephone service has it. This is done for the duration of a call by specifying a "*70" dialing prefix in your modem program. Otherwise, a Call-Waiting "click" may well disconnect your modem call, as many have found to their annoyance.

Another common annoyance, interrupted modem calls when someone in your house picks up another extension telephone, can be cured by putting an inexpensive "Line Protector" device on that other line, e.g., Radio Shack's TeleProtector. This doodad keeps the other extension disconnected as long as your modem call lasts. Handy.

If you suspect noise or hum may be keeping your residential telephone line from furnishing good modem connections, try temporarily disconnecting your house's telephone lines from the telco "demarcation" panel, and connecting your modem directly to it. (This will probably entail moving your PC to the garage, for the test.) If performance and reliability suddenly go up, then you need to re-wire internally. Also, listen to the level of background noise and hum while on a voice call. You may need to get PacBell to investigate (or, if nothing else works, order a second line, then cancel the first one).

Some good modem numbers to call for testing: 800-US-HAYES, 800-392-2432 (Multitech), and 847-982-5092 (U.S. Robotics).

The Linneus Bottumus

At $300 a pop, this had better be a good modem — and it is. In comparison with U.S. Robotics Courier V.Everything externals, the Optima gets similar throughput, and mostly similar results. The only exception I found was on noisy telephone lines, where the Optima sometimes failed or fell back to lower speeds, while the Courier didn't.

With the Hayes, you get solid support and an attitude that particularly caters to the business market. (You also get Caller ID support, call-type recognition, and Plug'N'Play, if that matters.) With the Courier, you get absolute-best performance, the current favorite in the 56 kbps sweepstakes, a real manual, and a slightly lower price ($259 at discount).

It's a close race, and you'll get a good product either way.***

[1] You find out, when doing the upgrade, that you shouldn't have thrown away the cardboard box, because it has the unit's real model number on it, required to find the correct upgrade file. For the record, (for this modem) it's Part Number 08-02353, version 6.0, which owners should scrawl on the chassis for future reference.

[2] There are also some newer variants: The 16552 is a pair of 16550 chips in a single package. That's fine. The 16650 is a slight improvement on a 16550, having twice the buffer space. Finally, some I/O boards (or motherboards) have serial circuitry inside surface-mounted VLSI (very-large-scale integration) chips. Good luck determining how good those are. Consult the documentation for your best hope of finding out.

[3] Buffer memory houses data in transit (a little like cache memory). It's important because the CPU may not be available to pick up a byte at the time it arrives in the receiving system's UART chip (a phenomenon called "interrupt latency"), resulting in overflowed buffers and resulting data loss. The problem is especially acute in OSes that have serial-driver problems, or that are poor at multitasking. In particular, if you have Windows 3.1, you have my pity and should at least upgrade to 3.11 and/or use a replacement shareware serial driver. Non-Microsoft OSes seldom have these problems.

[4] There are other variants, such as the more elaborate Primary Rate ISDN, which we won't get into.

[5] You're probably wondering when I'll discuss "cable modems", intended for use in the future with our good buddies at TCI and the like. OK, but I'll keep it short: "Feh."

Post-publication footnotes:

* USR "x2" modems shipped in Feb. '97, hard on the heels of this article going to publication. "x2" upgrades to existing USR modems are scheduled to ship March 31st. Rockwell/Lucent/Motorola products are just now starting to ship, in late March. This competing standard has recently been dubbed "K56flex".

Both competitors claim that their current 56 kbps modems will be upgradeable for free to the eventual ITU-T 56k standard. However, USR/TI products will be upgradeable in software, by re-burning the flash ROMs, whereas "K56flex" units will need certain of their chips replaced.

Curiously, Hayes appears to have started hedging its bets, by announcing that its Practical Peripherals subsidiary will be shipping modems using "x2" chipsets licensed from US Robotics. Hayes-branded modems, however, will continue to be "K56flex" only.

Late note: The 56kbps standard from the ITU-T finally did arrive, and is called "V.90". You'll want to look for that, among other things (not just "V.34", and certainly not "K56Flex", which still lives on to haunt us).

** Now, this is really trivia, but I'll note for the sake of completeness that no 56 kbps modem will be able, even under theoretical-ideal conditions, to deliver 56 kbps throughput, for now, because FCC regulations limit analog-line bandwidth to 53 kbps.

*** Although I was taking pains to be polite, and although the Hayes remains a worthy modem, I hope my point is clear that the USR Courier V.Everything external is both better and cheaper. I would probably not have bought the Hayes: I was given it for free and asked (by SFpcUG's Product Review Coordinator) to write this review.

Copyright (C) 1997,1998 by Ken Beseder.

From: Ken Beseder 
To: "'Lucky_Pierre@compuserve.com'" 
Cc: "'Ybarbero@earthlink.net'" ,
        "'tommcl@mail.sfpcug.org'" ,
        "'kirk@mail.sfpcug.org'" ,
        "'cjewel@mail.sfpcug.org'" 
Subject: Optima Review — Letter to the Editor
Date: Mon, 24 Mar 1997 21:18:11 -0800

Letters to the Editor
Lucky_Pierre@compuserve.com
cc: BN Editorial Board