Emax

After some experiments during the weekend, the current status of my 
little RS422 project is that I know how to do *fast* bank 
loads/unloads on the Emax via RS422. 
This should reduce the total data transfer time to 25-30 seconds on 
the Emax-I instead of the 2-3 minutes of Alchemy. Most probably this 
was also the total load time of the OMI CDS3 system, which is - let's 
say - "acceptable"... 
That's about the same speed as loading from a floppy :-) but the 
biggest advantage would be that one would have immediate access to 
hundreds of banks on the PC harddrive instead of having to copy 
individual banks to floppy disks first... 
Still SCSI is a much better alternative... for those having a rev2 or 
rev3 board, and for those using the Emax-II. BTW at RS422 speed, the 
data transfer time on a fully loaded Emax-II Turbo 8M would be about 
7 minutes :-). Fortunately every Emax-II is equipped with SCSI.

I have to write some decent software now which supports the full Emax 
handshaking protocol. But I'm pretty sure that the USB<-->RS422 
converters will not be the best solution for this communication - 
just like with the EII the communication seems to be quite unreliable 
when transmitting data from the PC to the Emax, as a consequence the 
total transfer time increases dramatically due to handshaking 
overhead. 
At the end of the week I will have a PCCard RS422 port on my laptop. 
This piece of hardware does not suffer from USB latency, so I hope it 
will work better...

///E-Synthesist


  
 

--- In emax@yahoogroups.com, "esynthesist" <esynthesist@...> wrote:
>
> Yes, I was also thinking there must be some dedicated command (set) 
> for fast load/unload. But the fact that John remembered a load time 
> of 5 minutes for the OMI cdroms made me doubt again... On the other 
> hand it is true that OMI cdroms could only be used after the 
release 
> of OS 3.2, so this is indeed an indication that additional commands 
> have been added, or at least some changes have been applied. I also 
> observed that the v 3.2 MIDI protocol is not 100% behaving as 
> described in the v 3.0 document, e.g. the timeout handling is 
> different. So there are also changes in the 'normal' SYSEX/MMA 
> protocol.
> By the way: the OMI drive also required a firmware update in order 
to 
> be compatible with the Emax. Question is of course whether this was 
> just a small firmware update (to support the newly added commands 
in 
> the Emax OS) or a huge piece of Emax-specific code (to implement 
the 
> full SYSEX/MMA command set - which is indeed quite unlikely) ...
> 
> The Emax-II and EIII indeed have a filesystem which is optimized 
for 
> handling different banksizes; I have the specs here because I 
needed 
> them for EMXP. The EII and Emax are using filesystems with fixed 
> filesizes in a sequential order.
> 
> Since I don't have any Emax OMI cdrom disk I'm not even sure 
whether 
> the banks on these disks are "EMX-like" 8-bit images or expanded 12 
> bit images. It makes sense that they are 8-bit, because this 
allowed 
> OMI to put more banks on a CD, to transfer them faster to the Emax 
> (if EII-like commands have been implemented in the Emax OS of 
course) 
> and to use the same bank layout as on the Emax floppy and Emax 
> harddisk banks.
> 
> So despite the "5 minutes load time" note from John, I think we can 
> still assume that there is some specific command set in V3.2 which 
> enables fast bank loads. I will try to find them out during the 
> weekend, either by experimenting or by looking into the OS 
> binary/disassembled code...
> 
> ///E-Synthesist
> 
> 
> 
> --- In emax@yahoogroups.com, tu@ wrote:
> >
> > 
> > Thursday, November 6, 2008, 4:23:43 AM, you wrote:
> > 
> > >
> > But the 5 minutes load time may have been reality...
> > 
> > This can explain why I don't know anyone and find no reference at 
> all 
> > of anyone who actually used this CD-ROM drive with the Emax. If 
> this 
> > 5 minutes load time is true, this must have resulted in a 
> commercial 
> > failure for OMI when they launched the Emax OMI cd disks... but 
> they 
> > probably released these disks also in Mac/SD format ?
> > 
> > 
> > 
> > 
> > Yes, such a slow load time would have been a major marketing 
> problem. I find it difficult to imagine that Emu would not have 
added 
> the small amount of 
> > extra code required to load in a bank quickly via RS422. If they 
> wanted to sell Emaxes then surely there was a strong incentive to 
> make the sound 
> > library efficient to use. I suspect the OMI CDROM system for the 
> Emax was not a major market success because of the cost. The OMI 
> CDROM drive, or 
> > even a Mac with a CDROM drive, would have cost a significant 
> proportion of the cost of an Emax. The average musician probably 
> would not have been 
> > able to justify that additional expense. Particularly so given 
that 
> early CDROM drives were rather fragile.
> > 
> > >
> > 
> > And yes, Emu has done strange things. E.g. the EII cdrom kit 
> > supported a "folder" or "category" system: the banks on a disk 
> could 
> > be put in folders (like bank "piano" in folder "acoustic 
> keyboards") 
> > to make navigation much easier. This feature was not available on 
> the 
> > Emax and EIII harddisks. Maybe Emu considered this to be a 
feature 
> of 
> > OMI and not of Emu themselves, but they could have learned from 
> > that...
> > 
> > 
> > 
> > 
> > Surely the category organisation was only a feature of the OMI 
> CDROM system, as the EII had no control over it. So it was OMI's 
> CDROM format, not 
> > Emu's format. But this also gave OMI the flexibility to put as 
many 
> banks on the disk as they wanted and to organise them as they 
wanted. 
> There was 
> > no restriction on how OMI could do this as long as they could 
serve 
> up the full memory image of each bank to the EII via the RS422 port 
> when required.
> > 
> > Don't the Emax and the EII hard disk formats allocate a fixed 
> number of banks for the disk? I believe you cannot fit any more 
banks 
> on the disk even if 
> > the existing banks are only half full of samples. Presumably this 
> is because the Emax and EII use a fixed memory size for each bank 
and 
> the complete 
> > data for the bank is copied directly between memory and disk when 
> you load or save a bank. Each hard disk bank is a the equivalent of 
> one floppy disk 
> > image minus the OS data. I believe the Emax II and EIII use 
> variable sized banks. Therefore the number of banks stored on a 
hard 
> disk or CDROM 
> > depends on how much data is contained in each bank. But I believe 
> there is still a limit of 100 banks per disk. 
> > 
> > >
> > 
> > 
> > 
> > RS422 Communication with the Emulator was designed based on 
> following 
> > key principles:
> > - all communication, including request/reply for parameter 
changes, 
> > occurs at 500 Kbaud
> > - a whole bank can be downloaded/loaded with one special designed 
> > type of command (a command which actually directly reads/writes 
the 
> > RAM memory segments in which the bank data is residing)
> > - bulk data load/unload occurs with data packets sized 256 bytes, 
> of 
> > which each byte represents 1 sample point (data is transferred in 
> > compressed format)
> > 
> > On the Emax, they seem to have decided that choosing for a 
> *standard* 
> > medium speed protocol was more important than choosing for a 
> > *proprietary* high speed protocol. So they went for the MIDI 
> > SYSES/MMA approach:
> > - all basic communication, including all commands/instructions, 
> > occurs at 31.25 Kbaud, no matter if the DIN5 MIDI sockets or the 
> DB9 
> > RS422 port are being used.
> > - loading/unloading banks requires the full set of SYSEX 
commands. 
> > Hence to simply download the parameters of just one voice of just 
> one 
> > preset, already multiple commands must be exchanged with the 
Emax. 
> > This is due to the fact that in general only one parameter can be 
> > transferred per command. And this must be done at the slow 31.25 
> > Kbaud speed...mmmm...
> > - bulk data (sample) load/unload occurs with data packets sized 
> only 
> > 120 bytes (MMA standard). Moreover each sample point requires 12 
> bits 
> > now instead of 8 bits on the EII since data is transferred in 
> linear 
> > format instead of compressed format.
> > As a consequence, loading/unloading banks is much slower than on 
> the 
> > EII. Of course, once they released the Emax-II, they would have 
> faced 
> > problems anyway. This machine could have up to 8MB banks and 
> doesn't 
> > use compression, so even at full 500 kbaud speed and using only 
one 
> > command - which is impossible in reality - the Emax-II would 
> require 
> > at least 2.7 minutes for loading/unloading banks. Fortunately 
there 
> > was something invented called SCSI :-)
> > 
> > 
> > 
> > 
> > Have a look at the MIDI spec for the Emax V3.0 software. The fast 
> (RS422) dumps use a protocol based on the MIDI SDS but slightly 
> modified. The 
> > sample data is dumped as 12 bit linear but the samples are packed 
> so that two 12 bit samples are transferred in three 8 bit bytes. It 
> is also of note that 
> > sending 8 bit wide data in this way violates the MIDI standard, 
as 
> bit7 is always reserved as an indicator of a status byte. Of course 
> this is not really an 
> > issue here as the 500k baud RS422 data is only being transferred 
> to/from the Emax so no other MIDI devices will ever see this 
> violation of the 
> > standard. But the outcome is that dumping samples as 12 bit only 
> takes 50% longer than dumping as 8 bit compressed. Doing a proper 
> MIDI SDS dump 
> > of 8 bit or 12 bit data actually takes the same amount of time as 
> only 7 data bits can be transferred for each byte in the message. 
So 
> an 8 bit dump 
> > takes two bytes per sample (7 + 1) while a 12 bit dump also 
> requires two bytes per sample (7 + 5). 16 bit dumps are even slower 
> as they require three 
> > bytes per sample (7 + 7 + 2).
> > 
> > As you have said, the failure to provide a means of directly 
> transferring banks into memory via RS422 seems to be the problem in 
> the Emax, at least as 
> > documented in the V3.0 MIDI spec. But if the V3.0 spec already 
> provides all the functions required to load banks from the CDROM 
> drive using MIDI 
> > SYSEX and RS422, then why is V3.2 or the SE software claimed to 
add 
> OMI CDROM support? It still seems likely to me that some extra 
> functions were 
> > added in those versions to support fast bank loading via RS422. 
If 
> not, then the OMI CDROM drive would have to be converting the 8 bit 
> compressed 
> > sample data on the CDROM to 12 bit linear in order to dump the 
> samples into the Emax. The Emax would then have to convert the 12 
bit 
> linear samples 
> > back to compressed 8 bit samples. The transfer of samples would 
> also take 50% longer for 12 bit linear compared to 8 bit 
compressed. 
> And of course 
> > there would be no way for sequencer data included in the bank to 
be 
> loaded into the Emax. I could be wrong, but it just seems unlikely 
> Emu would have 
> > made it so difficult when a small software update to the Emax 
could 
> make bank dumping work in much the same way as the EII.
> > 
> > >
> > 
> > Nevertheless I will still do some experiments to find out if the 
> Emax 
> > OS doesn't have any "fast bank load" commands... 
> > By the way: does anyone know whether the binary code of the Emax 
OS 
> > can easily be de-compiled/disassembled in some way in order to 
get 
> > some kind of source code ? Is a simple Z80 disassembler 
sufficient ?
> > 
> > 
> > 
> > 
> > Unfortunately it seems the only way is to experiment and see what 
> can be uncovered. The Emax NS32000 main CPU code can be 
disassembled 
> but it is 
> > not a common processor. The hard part of analyzing the 
disassembled 
> code is working out where the program and data begins and ends as 
> well as what 
> > interrupt routines are being handled at runtime and how they 
> interact. You would need to combine together the code from the disk 
> OS image and the 
> > EPROM into a processor memory map. Various hardware peripherals 
> will also probably exist at certain addresses in the memory map. To 
> pull  it all 
> > together you will ideally have the circuit schematics, the memory 
> map, CPU/chip documentation plus a detailed design description of 
how 
> the system 
> > works. Often much of this data can be found in the product 
service 
> manual. Then you need to determine which routines are called when 
> MIDI/RS422 
> > interrupts are handled. Testing with a logic analyzer probing the 
> CPU would certainly make that easier.
> > 
> > /Tristan
> > 
> > >
> > 
> > 
> > ///E-Synthesist 
> > 
> > --- In emax@yahoogroups.com, tu@ wrote:
> > >
> > > That seems excessively slow, as the EII could load a similar 
> sized 
> > bank from the same CDROM 
> > > drive in 12 seconds. Its hard to imagine Emu would not have 
> > implemented a similar load time on 
> > > the Emax if all it took was adding a software routine. But then 
> > again, stranger things have 
> > > happened...
> > > 
> > > /Tristan
> > > 
> > > Quoting John Silveria II <john@>:
> > > 
> > > > Somewhere, and I can't remember where, I read that the CD-Rom 
> > drive
> > > > took 
> > > > up to 5 minutes to load a bank. I wish I could remember 
where. 
> So
> > > > indeed 
> > > > it was not only as slow as typical SYSEX load, it could 
> actually 
> > take
> > > > 
> > > > longer.
> > > > 
> > > > -- 
> > > > Yahoo! Messenger: EmaxJS
> > > > The Silveria Family Website and Emax and Emax II User's Group
> > > > http://www.silveriafamily.com
> > > > 
> > > >
> > >
> >
>