Way back in 1994, IBM promised the owners of the WindSurfer that they would have v.34 in their hands by September 1994. Well, here it is, December 1995, and finally, v.34 is available for the Spectrum OfficeF/X. A beta-quality version for the WindSurfer (known as version 59A) was released late 1995 as well. This review will cover the new features and bug fixes in the 2.0 release of Spectrum's OfficeF/X drivers. The rest of the features are covered in my first review.
I installed the 2.0 upgrade on an Intel 486DX/33 with 8meg RAM running Windows 95 (retail build 950). Spectrum expects the drivers to work "better" under Windows 95 than the previous drivers, but they make no guarantees. I found that they worked quite well. A few performance problems were noted, but these were most likely due to Windows 95.
The installation went smoothly after one crash. I did notice a bug in the modification of autoexec.bat. The install forgot to prepend "%path%;" when it added a "set PATH=" line. This lost the path that was originally set by Windows 95. This may or may not be a problem with Windows 3.x.
Spectrum's phone number for support is now long distance instead of an 800 number. I guess Mwave has become popular. They also do not have an Internet email address for support. As a result, I didn't contact Spectrum for support, so any problems I ran into will be stated without any response from Spectrum on the matter. I will update this review if any problems are resolved.
v.34 (28.8kbps) worked fine for me when connecting to my Internet Access Provider (IAP). I'm not the BBS-head I used to be, so my testing was limited to my IAP. I usually get 21.6kbps when connecting, and an occassional dropped connection. From my understanding of the state of most v.34 implementations, this is to be expected. Sometime in the next year modem manufacturers will get their acts together and produce modems that connect reliably and consistently across brands. I'm hoping the Mwave modem will follow the trends as they develop. Mwave is a bit late in the game, so it might not have the advantage that the other manufacturers have. I briefly retested the other modem features and they seem to work as expected. The v.32bis (14.4kbps) modem is solid as ever.
The modem is now automatically loaded when a communication application accesses the COM port. Aura has been dropped from the package as a result. A small amount of DSP resources are used when the modem is waiting for someone to open the COM port.
Originally I had reported problems with the DOS modem in a Windows DOS box. Dummy me, I had the wrong IRQ. Both the version 1.0 DOS box modem, and the 2.0 DOS box modem work fine.
I had a lot of trouble with the native DOS modem. Spectrum told me that I had to install the DOS modem into the default directory of c:\officefx for it to work. I asked around on the Mwave mailing list on the Internet, and others indicated they had installed into c:\officefx and had no problems. Unfortunately, I have to fail Spectrum on this one. I have so much junk on my machine, that I try to keep it well organized. Drivers in the root directory make me cringe. If you want to use the DOS modem, don't forget to install into c:\officefx. Hopefully it will work for you.
While trying to troubleshoot this problem, the config.sys file pointed to a readme file in c:\officefx\mwd. I looked for it and found a readme.txt file with zero length. Oops.
Mwave's Digital Simultaneous Voice over Data (DSVD) feature allows two Mwave modems to connect and provide both Data and Voice transfer in both directions. It can only run at v.32bis (14.4bps) probably due to the strain of the Voice Over Data (VOD) task running on the DSP. When DSVD was running, 19.482 MIPS were consumed on the processor. This means DSVD won't run on the older MDSP1xxx boards (ACE5000, etc...).
The voice sound quality was very low, as expected, but it was intelligible. While talking, modem throughput was reduced, but only in the direction of the speech. If I talked while I was sending a file, I noticed a slowdown. If the other party talked while I was sending a file, there was no impact. The transfer that I did was reduced to about 1000 characters per second (cps) at 14.4bps (Error Correction, no Compression) with a moderate amount of conversation. Normally I get around 1680 cps with the v.32bis modem.
The modem automatically does Class II now. Bring up a FAX app that opens the COM port, and the modem will launch itself. As a pleasant surprise, Manual FAX Receive with Delrina WinFax works just fine. I noticed that the handset still stays connected while the two modems are talking, so you have to hang up quickly or risk lost data. This is a minor annoyance when weighed against the convenience of being able to pick up the phone and press the Manual Receive button when you hear those FAX tones. The AA light still stays dark when a FAX program is waiting for a FAX. I'm guessing this is because Delrina detects rings by watching for the word "RING" to come over the modem.
The new Telephone Answering Machine (TAM) drivers seem to be overly confident that a quiet caller is actually no caller at all. Several times I've had messages that are cut off due to the drivers thinking the person had stopped talking. This is particularly a problem with long distance callers and renders the TAM features of the board pretty much useless.
I discovered a binary file in the MWW\worldtrd directory named W07_001.DAT and it contains a parameter that is related to this problem. Comparing it with the OfficeF/X 1.0 drivers, and sure enough, it was changed. The parameter appears on the "Quiet-USA" line of this file (part of it is ASCII text) and it is called "Min Lvl dBm". This is most likely the level that the drivers consider silence. In 1.0 the value was -60 (REAL quiet), in 2.0 it is -48 (not so quiet, somebody may actually be talking at this level). Let's hope a patch is made available real soon, or there will be many angry customers.
The way I use Switchboard, I didn't notice a huge difference in the new 2.0 version. The contact list has some new fields, but still no OLE container for dragging and dropping links to files associated with a person in your addressbook. The annoying "second load on the line" that I noticed in 1.0 has not been fixed. A new phone ring simulation is provided. You can choose any WAV file to play when the phone rings or you can specify no sound at all.
The speakerphone is now heading toward half-duplex. In version 1.0 you could hear a faint echo of yourself in the speakers and the other party would hear the beginnings of your words cut off. Now it sounds like the beginnings of the words are getting through, but to eliminate the echo, the speakers are muted when you talk. So much for full-duplex speakerphone.
The phone line is still kept off hook when recording greetings through the local handset. The ability to record greetings and messages as WAV files is not yet supported. According to IBM the WAV to VOI conversion drivers in OfficeF/X 2.0 do not work under Windows 95.
The discriminator is now invisible and launches automatically like the modem. I tried Voice/FAX discrimination and it worked quite well. A class II FAX application like Delrina WinFax will not show up in the discriminator window as in version 1.0 even though it will answer the phone as expected.
I finally succeeded in testing Modem/Voice discrimination. With the discriminator set to do DTMF discrimination with *2 as the modem code, a calling modem with a dial string ending in @*2 was able to connect. The "@" in the dial string tells the modem to wait for silence. Without the "@*2" things worked fine as well. The Mwave discriminator will hand the call over to the modem if the voice application hears silence. The fastest connection was obtained by using a carefully selected number of commas followed by a "*2". The commas would pause until just after Switchboard picked up the phone. Then the caller's modem would send a "*2" and get my modem.
Lastly I wanted to do an all-out Voice/FAX/Data discrimination test. This came to an abrupt halt when I realized you can't have a Class II FAX application sitting on the COM port along with a Data Modem application. Trio DataFax which uses the proprietary Mwave MCI FAX interface would probably work fine since it doesn't use the COM port.
As I mentioned in the Modem section above, I had some trouble with the DOS drivers possibly due to my choice of installation directory. The new Mwcfg utility for DOS has an audio mixer to allow mixing the line input, and setting the volume of the SoundBlaster emulation and General MIDI synth. I couldn't get Mwcfg to work, but it does look useful. It also includes the Bass, Treble, QSound, Reverb, and Chorus controls found in the new Windows mixer.
The SoundBlaster emulation did work fine even though I had no control over volume. The DOS driver footprint is around 5k and it loads into UMBs. The emulation has moved up a notch from SoundBlaster classic to SoundBlaster Pro. This means 20-voice stereo FM synthesis. The PCM audio is still 8-bit mono. Let's hope SoundBlaster 16 is the next step.
I think I finally found a Windows program that accesses the SoundBlaster hardware directly. Activision's Action Pack 3 causes a window to pop up that says, "Unable to play DOS Game - Mwave is unavailable or in use by other applications". Action Pack 3 occasionally reports, "FM portion of soundcard not detected error". I've seen similar messages with Windows 3.1 and the OfficeF/X 1.0 drivers. I don't think the Mwave drivers have ever properly supported SoundBlaster under Windows.
Windows sound recorder records 22kHz, 8bit, Mono without trouble on my system. Anything higher is a bit of a problem and I think this is the fault of Windows 95. At top speeds (44.1kHz, 16-bit stereo) there was very noticeable glitching that didn't occur with Windows 3.1 and 4meg of memory.
The mixer application has been given a face-lift. New features include Bass/Treble controls, Reverb which is actually digital delay with what appears to be a fixed delay time, fixed mix, and adjustable feedback. The chorus effect is stereo with a fixed modulation rate and adjustable depth. QSound is now available from the mixer instead of just the MIDI Synth. You can QSound process your stereo WAV files, line-input, and CD audio. QSound affects the entire mix going to the speakers.
The MIDI synth includes a new set of EMU samples. This sample set is smaller than the old Prosonus set, and many instruments are missing. Fortunately, the old Prosonus samples still work if you have them on your hard disk. The Samples= line in the [MIDISNTH] section of mwave.ini can be changed to point to the directory where your Prosonus samples live, and all is well. Upgrading may delete these, so it's a good idea to copy them someplace else on your hard drive before installing the 2.0 upgrade. You will find the Prosonus samples in the c:\officefx\mww\midisnth directory of the old 1.0 install. I copied them to a \prosonus directory before installing 2.0.
The voice stealing bug that I reported in my last review has finally been fixed. The synth still runs at 22kHz sampling rate, and the number of voices depends on the free resources available on the DSP. The synth selection application is gone in favor of dynamic DSP resource allocation. I was able to make a 2400 baud modem connection after the MIDI synth was up, or before bringing up the MIDI synth, and there were no error messages. The synth gives up processor power to the modem when asked. This philosophy of shielding the user from the ugly details of DSP resources is a move in the right direction. Of course, now I wonder how many voices it can play simultaneously when nothing else is loaded. The new modem does use some DSP resources when idle, so the polyphony may be reduced.
I was unable to get this working, but it purports to be the same EMU sample based synth used in Windows. It receives MIDI through the MPU-401 port onboard and uses the EMU samples to produce the audio. My problems may be due to the fact that I didn't install into the default install directory (c:\officefx), see above. For me the MIDI data continued out through the physical MIDI port instead of being handled by the Mwave MIDI Synth.
The blast of garbage out the MIDI port that I reported in the last review is still here. This is probably a hardware problem, so I don't expect it to go away. There is still no support for SoundBlaster MIDI emulation. Only MPU-401 is available.
Aura is gone. All of the individual features of the Mwave drivers have been taught to start themselves up automatically, and avoid fighting during class. Once again, I find this move toward automation very user friendly.
The JPEG viewer is back. This made its first appearance in the WindSurfer drivers, and it was really bad. Well, it is the same JPEG viewer. Every JPEG file I try to load causes the DSP to hang, requiring a hard reset of the system. A very small subset of JPEG files load successfully. I've tried discussing this with IBM, but no one seems to want to fix the problem. My advice is to avoid the JPEG viewer. It does associate itself with JPEG files, so be careful, it may come up when you don't expect it and crash your Mwave board.
The added automation makes using the OfficeF/X a lot easier than before. There's no need to worry whether the modem is up, or if the MIDI synth will work while using the modem. New features like v.34 are what the programmable DSP is all about. The answering machine hanging up on my long distance callers, and the DOS modem requiring installation in the c:\officefx directory did not make me happy however.
For the bigger picture, I still haven't heard much about OS/2 drivers. Windows NT drivers are rumored to be in the works, while true 32-bit Windows 95 drivers have also been promised. As it was, the OfficeF/X 2.0 drivers performed well under Windows 95 with only a few performance and installation problems.
v.34 OfficeF/X (card + 2.0): $299 MSRP; v.34 OfficeF/X Software Upgrade (2.0): $59 MSRP from Spectrum Signal Processing, 1-800-667-0018.
<- Back to my Mwave page.
All product names are copyright and registered trademarks/tradenames of their respective owners.
The content of this review is primarily the opinion of Ted Felix, and no one else. Ted Felix is not affiliated with Spectrum Signal Processing, or IBM in any way.
This review Copyright ©1996, Ted Felix, all rights reserved. Permission is hereby given to distribute this review in a not-for-profit fashion. Any other use is a violation of copyright law. Violators will be prosecuted.