From "The Modem Report", Rockwell International - Digital Communications Modem Technology News. Issue No. 3, May 1993.

CLASS 1 OR CLASS 2, DOES IT MATTER?

By Mark Chapman, Chris Sneed and Tony Zuccarino

The widespread availability of low cost PC Fax modem boards has resulted in explosive growth in PC fax communication. This in turn has spawned a range of good PC fax applications software offering a variety of user friendly features.

To the user, the real questions are "how easy is it to send a fax" and "will the software work with my hardware".

There are many issues of standardization and much discussion in the industry over just what overhead a real time task such as receiving a fax will place on the PC.

Standards

International standards have facilitated reliable inter-working of modems and software by creating command interfaces for modem hardware and software. As long as any modem board and any communications software package comply with the same standards, they will work together.

Compatibility

To the end user, the interaction of the communication software and modem is irrelevant as long as the system works. But the question, "are there advantages to buying a modem complying with one standard versus a different standard?", is often raised.

The current industry debate centers on the relative merits of the two major standards for PC fax , the Class 1 and the Class 2 specifications.

About fax machines...

To understand the issues, let's review fax operation and how this affects the PC. A typical fax transaction consists of five phases:

Phase A. Call Establishment (e.g., dialing or answering the phone).
Phase B. Pre-message Procedure -- selecting parameters acceptable to the calling and the called station (i.e., speed, fine resolution, modulation etc.).
Phase C. Message Transmission -- usually image data for one page.
Phase D. Post-message procedure -- page break marker, and opportunity to change settings (e.g., drop to a lower speed because of too many errors detected during Phase C).
Phase E. Call clear down (i.e. hanging up the phone).

The actual operation of the five phases is specified by two CCITT standard protocols, T.4, and T.30.

T.4 defines the image transfer portion of the session used in phase C.

T.30 defines the negotiation and inter page phases (ABD&E). A normal multi-page transfer uses T.30 only at the beginning of the call, between pages, and after the last page.

Class 1 and Class 2:

Class 1 and Class 2 standards specify command sets for modems and software packages that allow modem manufacturers and software developers to build products independent of each other. As long as both of the vendors support either Class 1 or Class 2, their products will work well together. Class 2 is complex, supporting about 30 commands, while Class 1 supports less than 10. However, the main difference is that Class 1 requires the PC to handle both T.30 and T.4 protocols, while in Class 2 only T.4 is done in the PC.

                    Class 1        Class 2
   T.30 Protocol    PC software    PC fax modem
   T.4 Protocol     PC software    PC software

So how much does this additional processing mode affect the PC, especially one running Windows?

Significantly, Phase C, using T.4, takes up about 95% of the time and CPU effort during a typical facsimile transfer. This is because most of the time is spent scanning, compressing (or decompressing if receiving) and sending the pages, and very little time is spent by T.30 negotiating at the start and end of the call as well as between pages.

During the message transfer Phase C, Class 1 and Class 2 have the same performance. The two situations where there could be a difference between Class 1 and Class 2 in Phase C are:

(1) receive line quality checking (to determine if the received line is bad)
(2) padding the transmitted line (to meet minimum line transmit times)

Line quality checking is much more strenuous than line padding.

Due to the many methods of line quality checking, and the vagueness of the specifications, many, if not all, software packages perform this function in any case irrespective of the modem type. This means that line quality checking gets done in the PC irrespective of whether the modem supports Class 1 or Class 2.

As message transfer dominates fax transmission, most of the CPU intensive tasks are incurred by Class 1 and 2 alike. In fact as the number of pages increases, T.4 becomes by far the dominant factor in fax transmission and the difference between Class 1 & Class 2 becomes negligible.

Multitasking Environment

There are certain stages in call set up and interpage negotiation during which the modem requires the close attention of the PC. At these times, the PC software must respond rapidly and for Windows-based software (the application is controlling T.30), this means that the application cannot relinquish control to other applications during these critical times. It is this real-time constraint and associated interrupt overhead rather than any processor load factor which affects operation under Windows.

In practice, this effect can be largely prevented by suitable design of the applications software where the time critical T.30 portions are included at the driver level and these days all good commercial Class 1 fax packages work as well as Class 2 under Windows.

Class 1 Provides Flexibility

Class 2 is a "closed" standard, in that any changes to the T.30 protocol require a revision to the firmware (EPROM) on the fax modem board.

Modem and software manufacturers occasionally develop T.30 code updates in order to add additional features, correct errors in prior code revisions, or resolve interoperability issues with particular fax machines (the T.30 standard leaves some room for interpretation). While a Class 2 fax modem will require a firmware update (new EPROM) to incorporate the change, a Class 1 fax modem user can incorporate the change simply by downloading new software from a bulletin board, or ordering it from the software manufacturer.

There are about five different "Class 2" versions out there with no guarantee of compatibility (there are software packages on the market that do support all of the variants). However, most Class 2 fax boards support an interim working specification of Class 2 that has become the defacto Class 2 standard. Recently, the EIA-TIA approved a real class standard, but it will take 6 months to a year before software and hardware will emerge based on this new Class 2 standard.

Non-standard facilities (allowed under T.30) are not supported by Class 2 but can be easily added to Class 1 boards extending the life of modem hardware as new software or operating systems features become available.

In short, the Class 2 standard is fixed, whereas Class 1 is more flexible and can be maintained as software evolves. There are niche markets in which unique technical constraints dictate that Class 1 is favored and other niche markets where Class 2 is optimum but for the vast majority of the computing public, there is no discernible difference between Class 1 and 2. Both standards send and receive faxes reliably and the end user will neither care nor need to get involved comparing Class 1 and Class 2 as long as the application works.

Enough theory... is there actually a difference?

To evaluate the practical effect on CPU performance between Class 1 and Class 2 fax, a series of tests were performed on a representative set up using popular fax software packages. (See Figure 1). This was done by measuring the effect of receiving a fax in background mode while the PC in foreground is busy with a task.

(Figure 1 showwed time results of Class 1 and Class 2 using three Fax packages: Winfax, QuickLink II, BitFax on three CPU types: 386SX/25-MHz, 386DX/33-MHz, 486DX2/50-MHz. The times for a given CPU were within 5% of each other independant of Class or Fax Software.)

The task chosen for the series of tests was both CPU-and disk storage-intensive, and consisted of compressing a large collection of varied files. The received fax transmission consisted of 3 pages, including a cover page and representative test fax pages. The tests were repeated several times, and figures varied less than +/- 2% le