How connect PC Embedded to MDB (Vending Machine)

On Wed, 1 Oct 2008 15:45:14 +1000, "PhilW" <phil@microcoin.com> wrote:

>
>"Rich Webb" <bbew.ar@mapson.nozirev.ten> wrote in message 
>news:vlj5e4t6lqqprmagrakv9tp44joer11otm@4ax.com...
>> On Wed, 1 Oct 2008 09:42:46 +1000, "PhilW" <phil@microcoin.com> wrote:
>>
>>>
>>>Ok, then you have no choice but to write low level code to fudge the 9bit
>>>char by using the mark/space parity control.
>>
>> Sorry for jumping in here so far down in the thread, but to throw out
>> another option for the OP: Add a microcontroller as the glue between the
>> embedded PC and the MDB device.
>>
>> There are probably other chips but one family that I'm familiar with
>> that could handle this easily is the Atmel AVR ATmega. These natively
>> support 9-bit serial (in addition to the usual other formats) and, in
>> addition, can filter on the 9th bit set (MDB address mode). Almost like
>> they were made to support this protocol.
>>
>> Pick one with dual UARTS and talk to it serially, or get one with a USB
>> peripheral and communicate with it that way. It would take a teeny bit
>> of work but may be cleaner, in the end, than trying to dance with the
>> parity bit.
>>
>> -- 
>> Rich Webb     Norfolk, VA
>
>
>Hi Rich,
>
>If you have look back through some of the posts I mentioned why your 
>suggestion is not really a viable solution. It has to do with the msg 
>response time. In Sect 3 it says that a device, or host, must respond to a 
>msg within 5mS (5 char times).
>Therefore if you place a translator in series you will probably blow that 
>out of the water. By how much will depend on your implementation. If you 
>simply receive the command into the translator, convert it and resend it to 
>the device, your response time will be msg length(in chars @ 1mS each) * 2 + 
>device response time at best. with a message up to 36 bytes in length will 
>be 72mS + device response time. See what I mean.
>A better way would be to receive the msg from the host and convert on the 
>fly, retransmitting 1 char behind. At least you may be able to reduce it 
>down to 2mS + device response time. That only gives you a 3mS response time 
>window in the device to meet spec. On top of that, in both the above cases, 
>you will need to implement how you are going to talk to the PC and keep the 
>msgs in synch without an address/msg end byte (9th bit).
>
>Am I making sense?

Yup, totally. Interesting thread ;-)

I wouldn't do this as a byte-wise buffer device, but rather as a higher
level gateway. Implement the MDB state machine in the microprocessor to
behave as the communications gateway, peripheral, or bus master as
required/desired. The side facing the PC could run at a higher baud (or
even over a different medium) and would not necessarily use the MDB
message structure.

Personally, on the side facing the PC I'd try very hard to push
everything into a plain-text format. Easy to monitor and log, simple to
debug, and it can be digested by lots of analysis tools without a
separate binary -> ASCII translator pass.

-- 
Rich Webb     Norfolk, VA

On Sep 30, 7:42 pm, "PhilW" <p...@microcoin.com> wrote:

> Anyhow good luck. It can be done, I know because we support it. It's just
> not an ideal solution. Until you really know what you doing in it's
> implementation, stay away from virtual serial ports over USB because they do
> not act the same as a standard serial port. They move data in blocks and
> will not maintain the parity manipulation synchronously with the data bytes.

I don't think one should try to do the parity trick with a virtual
serial port.

However, some of the USB serial chips may be flexible enough to
implement a 9-bit
mode without resorting the parity trick.

Alternatively, some of the USB-capable microcontrollers certainly are.

And then there's the possibility of a parallel interface chip, and a
small CPLD.

None of these would be presented to the host PC as a serial port, but
instead as a custom interface.

<juanma.freelance@gmail.com> wrote in message 
news:dbf851fb-651e-4c99-ac11-561f5a9b8e54@l64g2000hse.googlegroups.com...
On 1 oct, 07:12, "PhilW" <p...@microcoin.com> wrote:
> <juanma.freela...@gmail.com> wrote in message
>
> news:ed142d1d-3df5-40ee-8eea-d698f3fbb7be@m3g2000hsc.googlegroups.com...
> On 1 oct, 01:42, "PhilW" <p...@microcoin.com> wrote:
>
> > <juanma.freela...@gmail.com> wrote in message
>
> >news:ef120646-80af-45a8-91c1-d53caeabeb53@59g2000hsb.googlegroups.com...
>
> > > On 30 sep, 03:09, "PhilW" <p...@microcoin.com> wrote:
> > >> <juanma.freela...@gmail.com> wrote in message
>
> > >>news:73aaa712-5e75-4674-a2b2-3a62b78569a3@p25g2000hsf.googlegroups.com...
> > >> On 29 sep, 20:29, James Beck <j...@reallykillersystems.com> wrote:
>
> > >> > In article <29c16500-d3a1-4d46-a7d4-
> > >> > 59492ef13...@w7g2000hsa.googlegroups.com>, 
> > >> > juanma.freela...@gmail.com
> > >> > says...
>
> > >> > > On 29 sep, 17:56, James Beck <j...@reallykillersystems.com> 
> > >> > > wrote:
> > >> > > > In article <3834d44e-355e-4d5f-af57-87e9ec4c0a97
> > >> > > > @z72g2000hsb.googlegroups.com>, juanma.freela...@gmail.com
> > >> > > > says...
>
> > >> > > > > On 29 sep, 16:33, James Beck <j...@reallykillersystems.com>
> > >> > > > > wrote:
> > >> > > > > > In article <1da3a443-e530-487a-8cb7-e4166e206414
> > >> > > > > > @i76g2000hsf.googlegroups.com>, juanma.pal...@gmail.com
> > >> > > > > > says...
>
> > >> > > > > > > On 29 sep, 11:11, juanma.freela...@gmail.com wrote:
> > >> > > > > > > > On 29 sep, 03:32, "PhilW" <p...@microcoin.com> wrote:
>
> > >> > > > > > > > > <juanma.freela...@gmail.com> wrote in message
>
> > >> > > > > > > > >news:c4d5b1b0-ea81-446b-ba2a-4f7ebe2811c3@l64g2000hse.googlegroups.com...
>
> > >> > > > > > > > > > Hi!
>
> > >> > > > > > > > > > I'm sorry but I am not English and I speack English
> > >> > > > > > > > > > very
> > >> > > > > > > > > > bad.
>
> > >> > > > > > > > > > My question is: How connect a PC Embedded to 
> > >> > > > > > > > > > Vending
> > >> > > > > > > > > > Machine? I know
> > >> > > > > > > > > > that exist a protocol MDB that is used by vending
> > >> > > > > > > > > > machine,
> > >> > > > > > > > > > but this
> > >> > > > > > > > > > conector is different to pc conector (RS232, USB,
> > >> > > > > > > > > > i2c,
> > >> > > > > > > > > > etc). Can I
> > >> > > > > > > > > > connect a PC Embedded by MDB? How?
>
> > >> > > > > > > > > > Anybody understand me?
>
> > >> > > > > > > > > > Thanks you!
>
> > >> > > > > > > > > We understand you.
>
> > >> > > > > > > > > Read the spec to understand what is required to
> > >> > > > > > > > > connect.
>
> > >> > > > > > > > > However, even once you have done the hardware to be
> > >> > > > > > > > > able
> > >> > > > > > > > > to
> > >> > > > > > > > > connect to a
> > >> > > > > > > > > PC's RS232 port you will, more than likely run into 
> > >> > > > > > > > > the
> > >> > > > > > > > > problem of the 9bit
> > >> > > > > > > > > char format not being supported on a std PC serial
> > >> > > > > > > > > port.
>
> > >> > > > > > > > > You can fudge it using the mark/space parity control
> > >> > > > > > > > > but
> > >> > > > > > > > > it
> > >> > > > > > > > > is not very
> > >> > > > > > > > > nice.
>
> > >> > > > > > > > > Why don't you use ccTalk instead?www.cctalk.org
>
> > >> > > > > > > > > regards
> > >> > > > > > > > > PhilW
>
> > >> > > > > > > > Thanks you.
>
> > >> > > > > > > > I dont understand all, what is "ccTalk"??
>
> > >> > > > > > > > regards.
> > >> > > > > > > > Juanma.- Ocultar texto de la cita -
>
> > >> > > > > > > > - Mostrar texto de la cita -
>
> > >> > > > > > > I read a little about spec of "ccTalk", and this is other
> > >> > > > > > > protocol,
> > >> > > > > > > no?
>
> > >> > > > > > > I need "to talk" with a vending machine. The standard of
> > >> > > > > > > vending
> > >> > > > > > > machine is the MDB (i think).
>
> > >> > > > > > > Are you understand me?
>
> > >> > > > > > > Thanks you.
> > >> > > > > > > regards.
>
> > >> > > > > > Here is a USB to MDB bridge
> > >> > > > > > :http://www.geocities.com/bonusdata/analyzer/mdb_interface/mdb_interface.
> > >> > > > > > htm
> > >> > > > > > (sorry for the spanning)
>
> > >> > > > > > Here is the protocol
> > >> > > > > > :http://www.vending.org/technical/MDB_3.0.pdf
>
> > >> > > > > >www.vending.orgiswhereyouneedtogo for any info about 
> > >> > > > > >MDC/ICP-
> > >> > > > > >Ocultar texto de la cita -
>
> > >> > > > > > - Mostrar texto de la cita -
>
> > >> > > > > Thanks you!!
>
> > >> > > > > I know this all, but the "USB to MDB" is not a good solution
> > >> > > > > for
> > >> > > > > me,
> > >> > > > > should be others solutions...
>
> > >> > > > > I am interesting in ccTal PhilW but i dont understand how 
> > >> > > > > apply
> > >> > > > > this
> > >> > > > > to the vending machine.
>
> > >> > > > If the vending machine does not speak CCTalk then there is no 
> > >> > > > way
> > >> > > > to
> > >> > > > apply it at all, in this case. MDB/ICP is the way you will have
> > >> > > > to
> > >> > > > go.
> > >> > > > CCTalk was designed more along the lines of gaming systems and
> > >> > > > MDB
> > >> > > > was
> > >> > > > designed for vending machines and the like. Not that you 
> > >> > > > couldn't
> > >> > > > use
> > >> > > > CCTalk in a vending machine, but MDB was standardized for use 
> > >> > > > in
> > >> > > > them
> > >> > > > (vending machines) long ago.- Ocultar texto de la cita -
>
> > >> > > > - Mostrar texto de la cita -
>
> > >> > > Oh James Beck, i think that you have reason.
>
> > >> > > My problem is how comunicate the pc embedded with vending machine
> > >> > > with
> > >> > > the protocol MDB... If the unique solution the "MDB --> USB
> > >> > > adapter"?
> > >> > > If you sum the price that a pcembedded+adapter ... for one for
> > >> > > machine
> > >> > > is very expensive no? it must of having another solution, no?
>
> > >> > > Thanks you!
> > >> > > regards.
>
> > >> > Well, if you doing a one off machine, there will never be any 
> > >> > economy
> > >> > of
> > >> > scale. If I were doing a single machine/test station, I wouldn't 
> > >> > balk
> > >> > at paying the price for the adapter. If I were doing thousands of
> > >> > them
> > >> > I would look at doing something more task specific and cheaper. You
> > >> > can
> > >> > implement the protocol yourself with some kind of adapter for the 
> > >> > PC
> > >> > and
> > >> > a communications library, but that would be what the specification 
> > >> > is
> > >> > for.- Ocultar texto de la cita -
>
> > >> > - Mostrar texto de la cita -
>
> > >> Oh, I see that you understand me :)
>
> > >> I need to create adapter for MDB to any port (serial?) the PC ... it
> > >> is difficult? I know computer science, but the electronic... :( any
> > >> idea?
>
> > >> Thanks you!
> > >> regards.
>
> > >> An adaptor is not an easy thing to do, simply due to the fact you 
> > >> need
> > >> to
> > >> maintain the timing specifications of the MDB protocol. An adaptor 
> > >> that
> > >> just
> > >> recieves a command from the PC, re-formats it for MDB and then sends 
> > >> it
> > >> on,
> > >> waits for the response and forwards that to the PC will easily break
> > >> the
> > >> timing criteria of the protocol, considering that each char takes 1mS
> > >> to
> > >> transmit or receive.
>
> > >> Concerning USB adaptors. Can be good providing they support, and
> > >> implement
> > >> the mark/space parity support correctly. However, you still have to
> > >> provide
> > >> low level coding to fudge the 9 bits via the mark/space parity. In
> > >> addition
> > >> you will be required to add timing delays to ensure the parity 
> > >> changes
> > >> are
> > >> sent over USB before transmitting the next data or parity change. We
> > >> have
> > >> had success with the CP2102 from silicon Labs. It is sure however, 
> > >> from
> > >> experience, that supporting MDB over USB is a lot more painful than a
> > >> standard serial port.
>
> > >> Another alternative is to acquire an add-on serial port that natively
> > >> supports 9 data bits at the hardware level. Not cheap, hard to find 
> > >> and
> > >> expensive.
>
> > >> As for the vending machine implementation, you have not said, but I
> > >> conclude
> > >> that YOUare building the vending machine and YOU therefore have 
> > >> control
> > >> over
> > >> the protocol to use. Unless by some specification requirement that 
> > >> you
> > >> use
> > >> MDB, I see no reason why you cannot use ccTalk. I believe the
> > >> "protocol"
> > >> is
> > >> purely for internal machine use, aimed at connecting all peripherals 
> > >> to
> > >> a
> > >> host controller, namely your embedded PC. Any external communications
> > >> is
> > >> usually by some other communications link, modem, ethernet, wifi, 
> > >> etc.
> > >> Why
> > >> ccTalk? It's simple and most importantly can be easily supported by 
> > >> any
> > >> PC
> > >> serial port, USB adaptor etc that supports the standard 8bit char
> > >> length.
> > >> It
> > >> could be any other protocol besides, that uses the 8bit data length 
> > >> as
> > >> well.
> > >> It's just that ccTalk is well documented, well supported and widely
> > >> used.
>
> > >> We produce products that support both MDB and ccTalk in addition to
> > >> others.
> > >> We much prefer ccTalk for its non requirement of the 9bit data 
> > >> format.
>
> > >> regards
> > >> PhilW- Ocultar texto de la cita -
>
> > >> - Mostrar texto de la cita -
>
> > > Oh, no, I am not a building the vending machine and I haven't control
> > > over the protocol to use, it is the problem. I need use the protocol
> > > standard in the most vending machine, that is, MDB.
>
> > > Do you understand me PhilW? Sorry for my English.
>
> > > regards.
>
> > Ok, then you have no choice but to write low level code to fudge the 
> > 9bit
> > char by using the mark/space parity control.
>
> > You do this by programming the UART 8bit with parity, that makes the 9
> > bits.
> > The parity you program to be manually controlled using the mark/space
> > controls to simulate the ninth bit.
> > When you sent the first byte of the message, you first make parity to be
> > Mark, this makes effective 9th bit set = address byte. Send the address
> > byte.
> > You then make parity to be Space for remainder of message = 9th bit 
> > clear.
>
> > To receive a reply just set to space parity and look for parity error 
> > for
> > end of message. Since this is a fudge you will be at risk to
> > miss-interpreting a real parity error, but that is the consequences of 
> > not
> > having full 9 bit support in hardware.
>
> > Anyhow good luck. It can be done, I know because we support it. It's 
> > just
> > not an ideal solution. Until you really know what you doing in it's
> > implementation, stay away from virtual serial ports over USB because 
> > they
> > do
> > not act the same as a standard serial port. They move data in blocks and
> > will not maintain the parity manipulation synchronously with the data
> > bytes.
> > They will need delays to allow the control commands to transfer before
> > sending the data etc. This will slow down your effective data rate under
> > MDB.
>
> > Also you should make an investment in a good serial coms monitor tool,
> > that
> > can capture and timestamp every event on the link between the host and 
> > the
> > device. It will pay for itself easily in the time you save discovering
> > what
> > is really happening in the transfer. We used a tool called "ViewComm
> > Async"
> > fromwww.greenleafsoft.com. Such a tool is worth it's weight in gold.
>
> > regards ...
>
> > leer m&#4294967295;s &#4294967295;- Ocultar texto de la cita -
>
> > - Mostrar texto de la cita -
>
> Oh, I think that I understand you.
>
> But, I have a question. How connect MDB to Serial Port, for use the
> monitor tool??
>
> Thanks you!!
> regards.
> Juan.
>
> Juan,
> I think you need to read the spec. There are to many details to deal with
> here.
>
> To connect to the PC you must build and interface circuit. How you do that
> is up to you.
> Sect 4 of the spec tells you all you need to know. It even gives a sample
> schematic.
>
> The bus is a multi drop bus, many slaves with one master.
> You still have not made clear which one you want to be. I know you have 
> said
> you are not building the vending machine, but you also have not said what
> you want to be, Host or slave. Are you the host controller or are you one 
> of
> the slave devices.
> I assume you do understand that the protocol is an internal machine 
> protocol
> and that you do not talk to the machine from the outside using MDB?
>
> Anyhow, as for connecting to the PC. Assuming you are the host, the
> interface you build will give an output of RS232 that you can connect to 
> the
> PC. The monitor you simply insert in between the interface and the PC.
>
> regards
> PhilW

Yes, I want that my pc embedded will be a Slave.

I want that my pc embedded will be as a Communications Gateway
(Section 8) for to receive all the information from the VMC. It is
possible?

regards.
Juan.


Hi Juan,

Yes it is possible, but everything I mentioned previously remains the same.

To be honest we have not looked at Section 8, we are mainly involved in 
coin/note validation. However, nothing else changes. The device you want to 
build is still just a device that must meet the standard MDB interface 
specification.

Connection to the MDB bus for you as a device is much simpler than that of a 
host. You can get away with a couple of opto couplers, just like in the 
sample schematic in that section 4. You would be for eg slave 1.

PhilW

"Rich Webb" <bbew.ar@mapson.nozirev.ten> wrote in message 
news:nbs6e4lmu620eun7b83cj91243nd5smsfb@4ax.com...
> On Wed, 1 Oct 2008 15:45:14 +1000, "PhilW" <phil@microcoin.com> wrote:
>
>>
>>"Rich Webb" <bbew.ar@mapson.nozirev.ten> wrote in message
>>news:vlj5e4t6lqqprmagrakv9tp44joer11otm@4ax.com...
>>> On Wed, 1 Oct 2008 09:42:46 +1000, "PhilW" <phil@microcoin.com> wrote:
>>>
>>>>
>>>>Ok, then you have no choice but to write low level code to fudge the 
>>>>9bit
>>>>char by using the mark/space parity control.
>>>
>>> Sorry for jumping in here so far down in the thread, but to throw out
>>> another option for the OP: Add a microcontroller as the glue between the
>>> embedded PC and the MDB device.
>>>
>>> There are probably other chips but one family that I'm familiar with
>>> that could handle this easily is the Atmel AVR ATmega. These natively
>>> support 9-bit serial (in addition to the usual other formats) and, in
>>> addition, can filter on the 9th bit set (MDB address mode). Almost like
>>> they were made to support this protocol.
>>>
>>> Pick one with dual UARTS and talk to it serially, or get one with a USB
>>> peripheral and communicate with it that way. It would take a teeny bit
>>> of work but may be cleaner, in the end, than trying to dance with the
>>> parity bit.
>>>
>>> -- 
>>> Rich Webb     Norfolk, VA
>>
>>
>>Hi Rich,
>>
>>If you have look back through some of the posts I mentioned why your
>>suggestion is not really a viable solution. It has to do with the msg
>>response time. In Sect 3 it says that a device, or host, must respond to a
>>msg within 5mS (5 char times).
>>Therefore if you place a translator in series you will probably blow that
>>out of the water. By how much will depend on your implementation. If you
>>simply receive the command into the translator, convert it and resend it 
>>to
>>the device, your response time will be msg length(in chars @ 1mS each) * 2 
>>+
>>device response time at best. with a message up to 36 bytes in length will
>>be 72mS + device response time. See what I mean.
>>A better way would be to receive the msg from the host and convert on the
>>fly, retransmitting 1 char behind. At least you may be able to reduce it
>>down to 2mS + device response time. That only gives you a 3mS response 
>>time
>>window in the device to meet spec. On top of that, in both the above 
>>cases,
>>you will need to implement how you are going to talk to the PC and keep 
>>the
>>msgs in synch without an address/msg end byte (9th bit).
>>
>>Am I making sense?
>
> Yup, totally. Interesting thread ;-)
>
> I wouldn't do this as a byte-wise buffer device, but rather as a higher
> level gateway. Implement the MDB state machine in the microprocessor to
> behave as the communications gateway, peripheral, or bus master as
> required/desired. The side facing the PC could run at a higher baud (or
> even over a different medium) and would not necessarily use the MDB
> message structure.
>
> Personally, on the side facing the PC I'd try very hard to push
> everything into a plain-text format. Easy to monitor and log, simple to
> debug, and it can be digested by lots of analysis tools without a
> separate binary -> ASCII translator pass.
>
> -- 
> Rich Webb     Norfolk, VA



Yes I agree with you that would be the better way to go. However, it is 
stepping away from the MDB protocol concept. It also adds a lot of work and 
maintainance.

PhilW

<cs_posting@hotmail.com> wrote in message 
news:da3e7bf3-6a2c-4159-8f92-fe328ec201e4@k7g2000hsd.googlegroups.com...
> On Sep 30, 7:42 pm, "PhilW" <p...@microcoin.com> wrote:
>
>> Anyhow good luck. It can be done, I know because we support it. It's just
>> not an ideal solution. Until you really know what you doing in it's
>> implementation, stay away from virtual serial ports over USB because they 
>> do
>> not act the same as a standard serial port. They move data in blocks and
>> will not maintain the parity manipulation synchronously with the data 
>> bytes.
>
> I don't think one should try to do the parity trick with a virtual
> serial port.
>
> However, some of the USB serial chips may be flexible enough to
> implement a 9-bit
> mode without resorting the parity trick.
>
> Alternatively, some of the USB-capable microcontrollers certainly are.
>
> And then there's the possibility of a parallel interface chip, and a
> small CPLD.
>
> None of these would be presented to the host PC as a serial port, but
> instead as a custom interface.


Hi,

I'm not aware of any USB chips that support full 9 bit mode. Probably for a 
couple of reasons (off the top of my head). They are designed emulate a PC 
serial port, the CDC spec does not really support it. The CDC spec is rather 
broad in its possible applications, but the section that is utilised for 
virtual serial ports is based on the modem type interface. As far as I'm 
aware modems don't use 9bit.

Even though some USB-capable chips are able to support 9bit serial, 
supporting it through the USB is another question.

PhilW

On 2 oct, 00:30, "PhilW" <p...@microcoin.com> wrote:
> Hi Juan,
>
> Yes it is possible, but everything I mentioned previously remains the same.
>
> To be honest we have not looked at Section 8, we are mainly involved in
> coin/note validation. However, nothing else changes. The device you want to
> build is still just a device that must meet the standard MDB interface
> specification.
>
> Connection to the MDB bus for you as a device is much simpler than that of a
> host. You can get away with a couple of opto couplers, just like in the
> sample schematic in that section 4. You would be for eg slave 1.
>
> PhilW

PhilW, Have you anyone schematic to make a adapter MDB to RS232 or any
pc port (usb, rj45, etc...)??

Thanks you!

Juan.

On 5 oct, 18:10, juanma.freela...@gmail.com wrote:
> On 2 oct, 00:30, "PhilW" <p...@microcoin.com> wrote:
>
> > Hi Juan,
>
> > Yes it is possible, but everything I mentioned previously remains the same.
>
> > To be honest we have not looked at Section 8, we are mainly involved in
> > coin/note validation. However, nothing else changes. The device you want to
> > build is still just a device that must meet the standard MDB interface
> > specification.
>
> > Connection to the MDB bus for you as a device is much simpler than that of a
> > host. You can get away with a couple of opto couplers, just like in the
> > sample schematic in that section 4. You would be for eg slave 1.
>
> > PhilW
>
> PhilW, Have you anyone schematic to make a adapter MDB to RS232 or any
> pc port (usb, rj45, etc...)??
>
> Thanks you!
>
> Juan.

Anybody Can answer me?

But.. have you considered using the Executive Protocol ?
Of course, it depends on the function you have to implement, anyway it
is less difficult,
even if it isn't possible to connect directly the PC seria port with
the vending machine.

Talking about mdb I've found only the PC2MDB product, but I haven't
used it.
Let us know if you try it..

>> However, some of the USB serial chips may be flexible enough to
>> implement a 9-bit mode without resorting the parity trick.
>> Alternatively, some of the USB-capable microcontrollers certainly are.

>I'm not aware of any USB chips that support full 9 bit mode ...
>Even though some USB-capable chips are able to support 9bit serial,
>supporting it through the USB is another question.

I'm unsure if there are affordable off-the-shelf
USB to 9 bit adapters
(Black Box probably has an adapter for $$$).

Atmel AVR and Microchip PIC chips (such as the PIC18F14K50, PIC18F66J50
as featured in the October 2008 Circuit Cellar magazine)
have 9 bit UARTS and USB 2.0 on chip, but require your own programming.
Manufacturer's Application Notes and user support might give you all
the building blocks, perhaps entire programs to handle the problem.