HyperTerminal sees a COM port but VISA does not

Question

It should not be possible that HyperTerminal sees a COM port but VISA does not. Does anyone know why or have any ideas?
Here's some more info: I wrote a LabView (LV) program and deployed it (via the Application Builder) to some of our production department's PCs (some Windows XP and some Windows 7). We use it to set up products we design and manufacture through our device's USB serial port. The USB port is actually a virtual COM port (VCP) - a USB to Serial converter or "bridge". We have a custom driver for our device and since each device's USB bridge chip is programmed with a different serial number, every time we connect a new device to the PC the installer runs and assigns a new COM port number. This is so our customers can use multiple devices on a PC. Fine. On one of our production PC's we were close to  COM900 ports (HyperTerminal sees only up to COM256) when suddenly Win7 would not allow any more. We were able to uninstall orphaned (non-present) COM ports from DeviceManager ("DM") by starting it from a command console (DOS window) with a special command-line switch, "set devmgr_show_nonpresent_devices=1" then "start devmgmt.msc" (I have this in a batch file). In DM View menu you can "Show hidden devices" and uninstall them as if they were connected. I've also tried uninstalling them the normal way when the device was actually connected. Either way, sometimes, not always, depending on the COM port number, after reusing that port number by installing a new device which picks up that old, freed-up number, LV VISA doesn't see it. In other words, when "Refresh" is clicked in the COM port VISA list box, any new devices installed should be and normally are seen. (Refresh is not needed if the application is started after the device's driver is installed. It is only need if the LV application is open with the new device is added.) However, HyperTerminal does see it and lists it in the Properties screen, the "Connect using..." list box. And the port works with HyperTerminal. (Of course we've tried rebooting after uninstalling and/or reinstalling the device, with the same results.)
When I've spoken to the NI Applications Engineers in the past about serial port issues (not about this problem) they have confirmed that if HyperTerminal can see the port then NI VISA should also see it. This is crazy. I am posting this in hope that someone can shed some light on this problem.
Thank you. -Ed

Answer 1

Good catch. Yes, we're using the Silicon Labs CP2102 and have created a "custom driver" installation package. Our new PCBs have a factory-programmed chip, so the PC reuses that COM port (COM3 on that PC). The problem is when we we re-program the CP2102 (setIDs exe) with our PID, and device strings (including the S/N), the PC runs our installer and assigns the next COM port number. I was hoping that maybe it would go up to 64K or something! Anyway, it surpassed COM256 but we ran into this problem when almost up to COM900.)
There is a registry entry for each SN device, which shows the assigned port. If we delete the key with regedit, the COM port database (don't know the file name) still has the port number. That's where Device Manager comes in. It seems to allow us to manually delete the ports, but it would take quite a while to delete 900. (I was trying to make a batch file for DevCon, but couldn't get it to see the HW ID of the devices in need of deletion.) So we tried manually deleting a few. After doing that Device Manager shows that Windows (and HyperTerminal) can reuse the port number when a new device (new SN) is installed, but, as you know from this post, VISA cannot see unless I manually add it in MAX.
I can't see how it has anything to do with the Silicon Labs driver. I might have to delete ALL of MY devices, then uninstall the driver files, reboot, then reinstall, as you suggested with the SiLabs driver. However, I don't see how VISA will now see these changes. VISA must have it's own database. That's what I thought deleting the ports in MAX would clear out, like a fresh install, but apparently not.
I've been through Silicon Labs' tech support on other issues and they've been able to help on some. I could run this by them.
BTW, have you noticed that Silicon Labs' chip and driver, while faster than NI's own "bridge" or serial-USB converter (I returned it because at $100 it's performance was worse than a $20 converter you can buy in Staples), is substantially slower than a real serial port? For example, our device can transmit 250 readings per second (a 12 byte string). A real serial port used with my LabView program can keep up, i.e.., graph & tabulate at that rate, but with a VCP we only get about 70 RPS. Same code. Can't figure this one out. They say the've tested the driver to 900 KB claimed. We're using 115KB here. Anyway, the VCP driver makes it look like a COM port which LV thinks it is. Why is it slower than a real port when the USB should be able to work almost 10 times faster than I'm using it (up to 1024 bytes per mSec per the full-speed USB spec, which SiLabs claims they verified)?
Anyway, I appreciate your help. Thanks! -Ed