Re: [PATCH] media: cx231xx: fix si2157 inversion for Hauppauge WinTV-HVR-935C

Hi Brad,

I extracted the 4.0.11 firmware from your
linux-firmware-hauppauge package (md5 matches yours:
b2670d8ae5e3369fc71edbb98cdd8f6e), restored the stock unpatched
module (inversion = true), and the result is an immediate, clean lock:

Lock (0x1f) Signal= -49 dBm C/N= 37 dB UCB= 0 postBER= 0

dmesg confirms: firmware version: B 4.0.11

So the mainline driver is correct as-is, and the problem was entirely
the firmware I had installed: the 4.0.25 build distributed in the
LibreELEC dvb-firmware repository does not lock with inversion = true
on this device — my patch was unknowingly compensating for that.
Please consider the patch withdrawn.

Hans, apologies for the noise, and thanks for routing this to the right person.

One thought: since the LibreELEC repo is one of the first search hits
for this firmware file, other users may hit the same silent failure.
If there is a canonical location you'd like people directed to for
the 4.0.11 file, I'd be happy to report the incompatibility to the
LibreELEC dvb-firmware maintainers as well.

Thanks again for taking the time to set up the generator and test —
much appreciated.

Kind regards,
Dimitris

On Sat, Jul 25, 2026 at 1:33 AM Bradford Love <[email protected]> wrote:
>
> Hi Dimitris,
>
> The 975 is the "full" build with two demods. The other two devices
> only have one demod populated, if you open yours you'll see a spot for
> the atsc demod.
>
> I configured my generator for your exact parameters, I find channels
> whether I set my generator to invert RF or not. I tested 6, 7, and 8
> MHz bandwidths to be sure.
>
> What I do notice is that your firmware file is not the same version as
> what we supply.
>
> $ md5sum /lib/firmware/dvb-demod-si2168-b40-01.fw
> b2670d8ae5e3369fc71edbb98cdd8f6e  /lib/firmware/dvb-demod-si2168-b40-01.fw
>
> dmesg reports:
>
> si2168 14-0064: firmware version: B 4.0.11
>
> I would suggest finding and trying firmware 4.0.11 and reporting back.
> I'm user ~b-rad on launchpad.net, you're welcome to extract it from my
> linux-firmware-hauppauge package. I have not tested against 4.0.25 and
> cannot verify that firmware works with our devices.
>
> Regards,
>
> Bradford
>
>
>
>
>
>
> On Fri, Jul 24, 2026 at 4:11 AM Dimitris Kerasiotis
> <[email protected]> wrote:
> >
> > Hi Bradford,
> >
> > Thanks a lot for testing — and for the 975 ATSC data point, that's very
> > useful.
> >
> > Here are the exact parameters of the muxes I'm receiving (Athens, Greece,
> > Hymettus transmitter). All of them are plain DVB-T, no DVB-T2 in use here
> > for the main national muxes:
> >
> > Bandwidth: 8 MHz
> > Constellation: 64-QAM
> > Transmission mode: 8K
> > Guard interval: 1/4
> > Hierarchy: none
> > Code rate (HP): 3/4 on the DIGEA muxes, 2/3 on the ERT muxes
> >
> > Frequencies (all locked with inversion = false):
> > 474 MHz (ERT) 482 MHz (DIGEA) 498 MHz (DIGEA)
> > 522 MHz (DIGEA) 530 MHz (ERT) 546 MHz (DIGEA)
> > 554 MHz (DIGEA) 578 MHz (DIGEA)
> >
> > Signal on the strongest mux (522 MHz), rooftop antenna:
> > -48 dBm, C/N 33-38 dB, UCB 0, postBER 0, PER 0.
> >
> > With inversion = true the same mux never gets past
> > FE_HAS_SIGNAL | FE_HAS_CARRIER (status 0x03) at that same signal level.
> >
> > One thing that may be relevant to reproducing this: my unit's Si2168-B40
> > ships with ROM firmware B 4.0.2, and with that firmware the demod reports
> > 0.00 dBm on every frequency — no carrier detection at all. I had to add
> > dvb-demod-si2168-b40-01.fw (which brings it to B 4.0.25) before I got any
> > signal readings at all. The inversion issue only became visible after
> > that. If your test units load a different demod firmware, that might
> > explain part of the difference.
> >
> > On the 975 result: looking at the driver, the 975 attaches both lgdt3306a
> > and si2168 to a single si2157, so the inversion flag is shared between
> > the ATSC and DVB-T frontends. The 935C is si2168-only. It may be worth
> > noting that the only other si2168-only board in cx231xx-dvb.c, the
> > Evromedia Full Hybrid Full HD, already uses inversion = false, while the
> > lgdt3306a and si2165 boards all use true. That made me wonder whether the
> > correct value is really demod-dependent rather than board-dependent.
> >
> > If that's the case, a question that would settle it: on your 975, does
> > DVB-T lock reliably with inversion = true? If DVB-T is also marginal
> > there, the 975 may need the flag set per frontend rather than per board.
> >
> > Caveats on my side: I only have one unit (model 111519, rev G2I6) and I
> > have only tested DVB-T — no DVB-C or DVB-T2 here to try.
> >
> > I'm happy to run any test you'd like on this unit, or to respin the patch
> > in whatever form you and Hans prefer — gated on the eeprom revision, or
> > reworked so the 975 keeps true for ATSC while si2168 gets false.
> >
> > Kind regards,
> > Dimitris
> >
> > On Fri, Jul 24, 2026 at 1:48 AM Bradford Love <[email protected]> wrote:
> > >
> > > Hi Dimitris,
> > >
> > > I am testing this and investigating. I don't have a 935, but I have 955 and 975 of all revisions. The "family" consists of those three. Your finding would affect both 935 and 975.
> > >
> > > I will confirm that your patch does break atsc when using dvbv5-scan on a 975, so while as is I imagine it's ok for 935's setup, there is something else going on. The device you're using has been in production a very long time, and you're the first to report such issue. It's just odd. I'm going to talk with hardware lead about exact revision changes. I need to reach out to our European support as well to have them test.
> > >
> > > Can you by chance give me your exact parameters so I can set my generator up as close as possible to your Greek stations? Granted I don't have a dvbt2 generator, but there may be one in the main office. When I set my dvbt generator up the rf output setting does not matter, both normal and inverted work without issue when switching between for dvbt.
> > >
> > > Kind regards,
> > >
> > > Bradford
> > >
> > >
> > >
> > > On Tue, Jul 21, 2026, 12:30 Dimitris Kerasiotis <[email protected]> wrote:
> > >>
> > >> Hi Hans, Brad,
> > >>
> > >> Thanks for looking at this.
> > >>
> > >> I can only confirm the behaviour on my own unit: Hauppauge model
> > >> 111519, rev G2I6 (USB ID 2040:b151), tuner reported as Si2157-A30.
> > >> With inversion = true it never locks on any DVB-T mux, briefly showing
> > >> FE_HAS_SIGNAL | FE_HAS_CARRIER but never FE_HAS_LOCK, even at -48 dBm
> > >> / ~33 dB C/N. With inversion = false it locks immediately and scans
> > >> all 8 DVB-T muxes here in Athens, Greece cleanly (postBER 0). The same
> > >> unit works under the Windows driver.
> > >>
> > >> I don't have a second 935C to test other revisions, so I can't tell
> > >> whether this is universal or specific to this revision. I based the
> > >> change on the fact that the Evromedia Full Hybrid Full HD, which uses
> > >> the same cx231xx + si2168 + si2157 combination, already uses inversion
> > >> = false.
> > >>
> > >> If it turns out to be revision-specific, I'm happy to respin the patch
> > >> to gate it on the eeprom model/revision instead — just let me know
> > >> what you'd prefer.
> > >>
> > >> Regards,
> > >> Dimitris
> > >> On Tue, Jul 21, 2026 at 8:14 PM Dimitris Kerasiotis
> > >> <[email protected]> wrote:
> > >> >
> > >> > Hi Hans, Brad,
> > >> >
> > >> > Thanks for looking at this.
> > >> >
> > >> > I can only confirm the behaviour on my own unit: Hauppauge model 111519, rev G2I6 (USB ID 2040:b151), tuner reported as Si2157-A30. With inversion = true it never locks on any DVB-T mux, briefly showing FE_HAS_SIGNAL | FE_HAS_CARRIER but never FE_HAS_LOCK, even at -48 dBm / ~33 dB C/N. With inversion = false it locks immediately and scans all 8 DVB-T muxes here in Athens, Greece cleanly (postBER 0). The same unit works under the Windows driver.
> > >> >
> > >> > I don't have a second 935C to test other revisions, so I can't tell whether this is universal or specific to this revision. I based the change on the fact that the Evromedia Full Hybrid Full HD, which uses the same cx231xx + si2168 + si2157 combination, already uses inversion = false.
> > >> >
> > >> > If it turns out to be revision-specific, I'm happy to respin the patch to gate it on the eeprom model/revision instead — just let me know what you'd prefer.
> > >> >
> > >> > Regards,
> > >> > Dimitris
> > >> >
> > >> >
> > >> > On Thu, Jul 16, 2026 at 7:34 AM Bradford Love <[email protected]> wrote:
> > >> >>
> > >> >> Hi Hans et al,
> > >> >>
> > >> >> I'm out of town currently but can run tests Sunday. I have models of every generation, so I can test for inversion changes. I will run dvbt tests and reply again.
> > >> >>
> > >> >> Regards,
> > >> >>
> > >> >> Bradford
> > >> >>
> > >> >>
> > >> >>
> > >> >>
> > >> >> On Wed, Jul 15, 2026, 06:18 Hans Verkuil <[email protected]> wrote:
> > >> >>>
> > >> >>> Brad,
> > >> >>>
> > >> >>> Can you review this patch? I'm not sure if this 'inversion = false' is valid for just
> > >> >>> one specific model, or if it is valid for all 935C models.
> > >> >>>
> > >> >>> I'm hoping you have some info about that, or are perhaps able to test this with a
> > >> >>> 935C.
> > >> >>>
> > >> >>> Regards,
> > >> >>>
> > >> >>>         Hans
> > >> >>>
> > >> >>> On 09/07/2026 23:19, Dimitris Kerasiotis wrote:
> > >> >>> > On the Hauppauge WinTV-HVR-935C (USB ID 2040:b151) the si2157 tuner
> > >> >>> > is configured with inversion = true. On at least some hardware
> > >> >>> > revisions (tested: Hauppauge model 111519, rev G2I6) this is wrong:
> > >> >>> > tuner and demod probe successfully, firmware loads (si2157 patch
> > >> >>> > 3.0.5, si2168 B 4.0.25), signal strength is reported correctly and
> > >> >>> > the Si2168 briefly reports FE_HAS_SIGNAL | FE_HAS_CARRIER, but it
> > >> >>> > never achieves lock on any DVB-T mux, on any frequency, regardless
> > >> >>> > of signal quality (tested down to -48 dBm, C/N 38 dB). The same
> > >> >>> > unit works under the vendor's Windows driver.
> > >> >>> >
> > >> >>> > Setting inversion = false results in immediate lock (FE_HAS_LOCK,
> > >> >>> > postBER 0) and a successful scan of all eight DVB-T muxes receivable
> > >> >>> > in Attica, Greece. This matches the configuration already used for
> > >> >>> > the Evromedia Full Hybrid Full HD, which uses the same
> > >> >>> > cx231xx + si2168 + si2157 chip combination.
> > >> >>> >
> > >> >>> > Tested on kernels 6.1 and 6.12 with identical results.
> > >> >>> >
> > >> >>> > Signed-off-by: Dimitris Kerasiotis <[email protected]>
> > >> >>> > ---
> > >> >>> >  drivers/media/usb/cx231xx/cx231xx-dvb.c | 2 +-
> > >> >>> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >> >>> >
> > >> >>> > diff --git a/drivers/media/usb/cx231xx/cx231xx-dvb.c b/drivers/media/usb/cx231xx/cx231xx-dvb.c
> > >> >>> > index 472a4405a..a5486bcca 100644
> > >> >>> > --- a/drivers/media/usb/cx231xx/cx231xx-dvb.c
> > >> >>> > +++ b/drivers/media/usb/cx231xx/cx231xx-dvb.c
> > >> >>> > @@ -1014,7 +1014,7 @@ static int dvb_init(struct cx231xx *dev)
> > >> >>> >               si2157_config.mdev = dev->media_dev;
> > >> >>> >  #endif
> > >> >>> >               si2157_config.if_port = 1;
> > >> >>> > -             si2157_config.inversion = true;
> > >> >>> > +             si2157_config.inversion = false;
> > >> >>> >
> > >> >>> >               /* perform probe/init/attach */
> > >> >>> >               client = dvb_module_probe("si2157", NULL, tuner_i2c,
> > >> >>>