$Id: tda9887.html 1131 2006-04-29 17:48:52Z isely $
Mike Isely <isely at pobox dot com>
This page describes issues involving the pvrusb2 driver's use of the tda9887 module that is a part of Vidoe4Linux. What is described here is general information about the chip and its module, some tips on making it work for the pvrusb2 driver, and a short history of driver changes related to this module. If you are having problems related to bad tuning or no sound - or if you suspect TDA9887 problems - reading this page might help.
Note that for pvrusb2 driver snapshots from 20051231 onward, this should all be automatic now. So hopefully if I got everything right then you can ignore the rest of this page. On the other hand...
You can find the main pvrusb2 driver page on the web at http://www.isely.net/pvrusb2/pvrusb2.html.
The TDA9887 is an I2C-accessed chip that is involved in demodulation of the TV signal. To do this job, it needs to be given some configuration information relating to the video standard in use, the mode (radio vs TV), the surrounding circuitry, and maybe also related to the frequency being tuned (not sure about that last part). This job is handled within V4L by the tda9887.ko module (we'll just call it "tda9887" from here onward).
Like any I2C module in V4L, the tda9887 module when loaded will scan a device's I2C bus for a TDA9887 chip, and upon finding one it will attach to the bus and receive commands which it will interpret into configuration changes for the chip.
One complication involved in the configuration of the TDA9887 is that it has several binary configuration bits - "port1", "port2", and "qss" - whose settings depend apparently on the type of tuner being used and the surrounding circuitry on the device. This information unfortunately cannot be easily learned just by monitoring the parent driver (i.e. pvrusb2) for state changes, nor can the parent driver always know how these settings should be controlled. To get around this problem, the tda9887 module has three module options that may be manually specified when the module is loaded. They are:
In each case <x> must be either 0 or 1. I don't entirely understand what these all are supposed to do, however they all default to 1 if not otherwise controlled and I am told that port1 and port2 at least must be set to 0 in order to get sound when using PAL. (In other cases it apparently doesn't do any harm to have them be zero and without these set you may still at least have tuning problems.)
A side note... One other handy module option for tda9887 is debug. If you set debug=2 as an option for tda9887, i.e. if you do this:
modprobe -r tda9887; modprobe tda9887 debug=2
Then the tda9887 module will dump a verbose interpretation of the commands it is sending to the chip, every time it configures the chip. You'll be able to tell from that what the port1 and port2 settings currently are.
The take-away here is that if you encounter a situation where sound does not work, or tuning in general seems flakey, you may want to try adjusting these settings. This situation exists for any device which has a TDA9887 inside of it; this is not specific to the pvrusb2 driver.
Unfortunately, with the pvrusb2 driver, the situation gets yet more interesting...
There are two complications involving the pvrusb2 driver. One is that some PVR USB2 devices don't actually need the tda9887 module to operate. The other is that the pvrusb2 driver has had in it all along some old hacked up code which had been trying to manipulate the TDA9887 anyway. These two complications just serve to further muddy up any understanding of TDA9887 issues within this driver.
The first complication has to do with even having a TDA9887 present at all. It turns out that the PVR USB2 device I had been testing against for the better part of a year doesn't need this help at all. I think there may still be a TDA9887 inside of it, but I am in NTSC territory here and my reading of the TDA9887 datasheet suggests that it is possible to hardwire the TDA9887's configuration for NTSC (thus no need for it to even be connected to the I2C bus). So, there are PVR USB2 devices out there for which this TDA9887 nonsense is just, well, a non-issue. Unfortunately that fact was true for your driver author here and therefore I never even knew to suspect issues here. That all changed when I acquired a second PVR USB2 device, which does have a live TDA9887 in it.
The second complication has to do with some old logic in the driver. As I had stated at the top of the main driver web page, I didn't write this driver. The original version of this driver directly coded for all those I2C chips within the driver itself and did not depend on any outside I2C modules to assist in its operation. All that hardwired code unfortunately wasn't very flexible and couldn't account for all the various situations that people can find themselves in while using the device. One of the first things I did therefore was to rip out all that old redundant code in favor of using the already-debugged I2C modules in V4L. During my initial work, I did this for everything I could identify, and that resulted in the pvrusb2 driver using tuner, saa7115, tveeprom, and msp3400. However at the time there was one block of wierd I2C commands that I could not identify. At the time - not wishing to mess with what apparently wasn't harming anything - I decided to just leave that logic intact. It turns out that this unidentified logic was in fact the hardwired stand-in for the tda9887 module. Upon finally understanding this I recently ripped out that old code from the pvrusb2 driver so that tda9887 could do its job properly.
What the above means is that people who didn't think they needed tda9887 nearby may now suddenly find that they need it.
The pvrusb2 driver as it stands for the 20051231 snapshot and onward should have everything correct now. With that snapshot, all the old hack code is gone, and the driver attempts to correctly operate the tda9887 module. But for clarity, here is a summary of the snapshots over time and how each interacted with the TDA9887 chip:
For everything before the 20051206 snapshot, the pvrusb2 driver made no attempt to interact with the tda9887 module and that old TDA9887-controlling hack code was present.
In late November 2005, I began learning the wonders of TDA9887 (said with tongue in cheek), and thus with the 20051206 snapshot, the pvrusb2 driver now recognized the tda9887 module and attempted to operate it. But this was very minimal - I still had yet to learn about the port1 / port2 nonsense and I hadn't yet stumbled back on that old hacked up TDA9887-controlling code that was still lurking in the driver. So what worked here at all probably worked badly.
Unfortunately in parallel to the TDA9887 craziness I was also in the middle of a large round of driver changes that involved how the driver interacts with all of the I2C modules. But in the middle of all that I ran into the old hacked TDA9887 code and suddenly realized that it was probably clashing with the tda9887 module. The purpose of this snapshot was to remove that hacked code. At this point there was still no attempt in the pvrusb2 driver to deal with the port1 / port2 configuration issues. Anyone using this snapshot therefore still has to possibly specify "port1=0 port2=0" as options to the tda9887 module. The exact workaround for this snapshot should be this:
The last step is needed because this driver snapshot did not know to reconfigure the tda9887 module when it reappears in the system. Re-plugging the hardware forces the entire driver to re-initialize itself and thus tda9887 along with it. (The big change in the next snapshot solves that issue.)
From the New Years Eve snapshot onward, the pvrusb2 driver includes new logic to handle all the I2C modules. With this change comes a new ability implemented for tda9887. Now when the pvrusb2 driver detects tuner types 38 or 43, AND if tda9887 is attached, it will issue a command to the tda9887 module to set port1=0 and port2=0. The new I2C control logic is also smart enough to autonomously (re)do this change when circumstances change, e.g. if the tda9887 module is unloaded then reloaded.
At this point, everything should be correct.
The decision to only set port1=0 and port2=0 on tuner types 38 and 43 follows behavior in the ivtv driver (well, for type 38, where this seems to work). I have no idea if this is good enough however. It's possible that other tuner types with tda9887 loaded may require this tweak. In any case however the settings for port1, port2 (and qss) can still always be overridden using the usual tda9887 module options.