On Wed, 2010-06-09 at 19:48 -0400, Dale Pontius wrote: > On 06/04/10 21:42, Andy Walls wrote: > <snip> > >> cx18-0 843: Video signal: present > >> cx18-0 843: Detected format: NTSC-M > >> cx18-0 843: Specified standard: NTSC-M > >> cx18-0 843: Specified video input: Composite 7 > >> cx18-0 843: Specified audioclock freq: 48000 Hz > >> cx18-0 843: Detected audio mode: mono > >> cx18-0 843: Detected audio standard: no detected audio standard > > ^^^^^^^^^^^^^^^^^^^^^^^^^^ > > | > > And this is what makes me pull my hair out. ----+ > > > > Same card, same chipset, same firmware image, in the same PC and mobo, > > same TV channel with the same RF signal, same analog tuner assembly, and > > one '843 core detects the audio standard and the other '843 core > > doesn't. :P > > > > If you have controlled the signal levels properly and used good cable > > and grounding, the difference has to be in the analog tuner can. Either > > it has gone bad, or it is picking up a lot of noise from somewhere. > > > > Come to think of it, a bad analog tuner assembly could explain the red > > screen. > > > > Anyway, if *everthing* is the same, and one card works and one card > > doesn't, there's nothing to be done in software - except implement > > workaround that may not work. I *speculate* that you have a dying card. > > > To continue contributing to sales of Rogaine... > > This evening I finally got a few minutes to rub together, so I've done > an experiment. I shut down mythbackend, then started rmmoding as much > of the cx18 stuff as I could positively identify using modules.dep. The > idea was to "shut off" the cards with nothing else significant going on > with the system, like a power-down or reboot.
Hmmm. Once some of those components are powered and enabled by software, they are never disabled. The digital tuner and demodulator chips come to mind. I think the analog ones are always consuming power (with slightly more power used when left tuned to higher channel frequencies due to running the oscillators faster). BTW, if from the v4l-dvb source code tree, as root you do # killall pulseaudio; modprobe -r cx18-alsa (repeat that until it removes the cx18-alsa module) # make unload; make unload That should unload all the modules for the video and dvb stuff. > Then I modprobed cx18 and > checked the status of both video devices. Both came up properly, both > had sound. Tempting fate, I decided to try a reboot. Both cards came > up properly, good sound, no red screens. > > Which makes me think back just a little bit... > > A few weeks back we had a bit of a heat wave here in Vermont - 80s even. > The heat broke, and everything started cooling down after Friday, June > 5th, in quite a marked way. (some 20 degrees cooler) Looking back I > see that I reported that the card "came back" on June 6. It came up OK > just now, and it's still cool. > > I'm wondering if there's a temperature effect here. Temperature is > quite a significant factor in semiconductor performance. I'm going to > keep an eye on this, and if sound goes away when it gets hot, I'll turn > on the A/C for a few hours to cool the room and let the system thermally > soak, then try again. > > To put this back into the "dying card" mindset, there are many failure > and degradation modes in semiconductors, but one of those mechanisms is > a threshold shift (Vt-shift) that tends to slow things down. Higher > temperatures generally slow things down. I'm wondering about putting > them both together. The hottest part of the card dealing with the analog signal is the analog tuner can. It has two chips in it: a mixer/oscillator and an IF demodulator. I'm betting the IF demodulator is failing when getting too hot. AFC and AGC do rely on thresholds, so maybe threshold shift matters there. Just put more fans in the PC chassis and be done with it. :) > How is audio initialization done? Well, there are a lot of things. Let me write out the RF-aduio chain: Antenna cable Tuner preselector band filters (VHF low, FM radio, VHF high, UHF) Tuner 1st stage: oscillator. mixer, IF amplifier, and AGC Tuner SAW IF filter Tuner 2nd stage: AGC, AFC, sound IF output (and sound demod to mono AF - not used by driver but wired up) External capacitors and resistors (block DC, meet impedance requirments, etc). CX23418(CX25843) Analog front end: clamping, droop compensation, variable gain analog amp with AGC, and ADC SIF measurement and decoding: 8051 microcontroller assisted/automated Dematrix (stereo L R, SAP) Basedband digital audio routing & processing: volume, balance, equalization, clamping, etc. Sample rate conversion Audio Input Mux 1: selects between '843 digital audio and external I2S digital audio CX23418 APU input In your failure case, the 8051 microcontroller had failed to detect BTSC in the SIF. What should be really easy to find in the SIF for NTSC? A BTSC signal - especially the FM modulated mono L+R channel. Working backwards, that means either 1. the SIF measurement/analysis was bad (bad 8051 cx23418.dig firmware load?) 2. SIF analog to digital conversion was bad 3. SIF analog signal conditioning by the CX23418 analog front end and AGC was bad 4. external components between the tuner can and CX23418 charged up close to the rails, killing the SIF signal 5. the tuner IF demodulator output a useless SIF signal or failed to lock on to the sound carrier 6. the IF freq out of the mixer oscillator was off center, video passed through, but the 4.5 MHz sound carrier was filtered out 7. the mixer, oscillator was at the wrong freq, or locked on to a nearby intermodulation product (spur) as the pciture carrier instead of the actual picture carrier 8. Gremlins It's easy enough to clip a scope probe on the leg of the tuner that has the SIF signal on it to check at #5 above. Regards, Andy > I know I've seen you give parameters > before to "slow things down", but don't know if that's relevant in this > case. The weather is warming up toward next week, so it might be a good > time to try some of these experiments. > > Dale Pontius > > _______________________________________________ > ivtv-users mailing list > [email protected] > http://ivtvdriver.org/mailman/listinfo/ivtv-users _______________________________________________ ivtv-users mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-users
