I did work on the ivtv driver to help get the 150 supported so I feel I've got some expertise on the subject. I also know a lot about the MythTV source, but mainly as a tinkerer. This page compares image quality between the Hauppague PVR-250/350 model (which use a Philips saa7115) and PVR-150/500 model (which use a Conexant cx25840 chip).
Yes, there are differences between the quality of both cards.
After reading the specs on the 150, I thought it would be better considering it has a 3 line
comb filter (as opposed to the 250's 2-line). When I first got it, I was less than impressed
(first is from the tuner, second on composite fed from a DVD):
Tuner Composite In
The blurriness however is caused by noise on the channel surpassing the luma comb
error threshold, which disables the 2 and 3 line filters and falls back to "notch mode".
This being Linux, those values are tweakable (sometimes too far as in the second photo):
The 250 also has a "luma gain filter" which is in essence a sharpness control
(boosting the luma signal by 5.5dB at peak). The 150 also has this filter, so enabling
that got me even closer to identical output. I actually like the sharpness turned down,
+5.5dB makes some noise in the image really stand out. Here's a sample 150 shot with I think a 2.5dB luma gain:
The only real problem I see with it is that the hue is slightly more red than the 250. I can adjust this in mythtv, but it affects both cards so I just modify the driver to push the hue a little more green and they match pretty much dead on. This apparently is only my card, since I borrowed a friend's for comparison and his was a much closer match to the 250.
My 250 isn't without it's problems though. It has an automatic gain control for brightness that sometimes will suddenly flip the picture noticably darker. I also fix this by disabling the AGC at startup and forcing it to a specific gain. Some ivtv driver versions also have some frame tearing (where the bottom half of the picture is shifted right by half the screen).
# Set the PVR250 AGC control to a fixed gain of 0x120 ivtvctl -d /dev/video0 -g reg=0x03,val=0x26 ivtvctl -d /dev/video0 -g reg=0x05,val=0x20
Both seem to exhibit generation of some pretty ugly MPEG blocks. I don't know if this
is related to a buggy fDCT and quantization process or if the recording bitrate is just too low.
4x magnification of capture
Which do I like better? Right out of the box, the 250. After tweaking, I the 150's quality is pretty close, I think it is slightly noisier (due to me having to crank up the luma comb threshold to prevent notch fallback). A friend of mine has a 150MCE that looks as good without tweaking as my 150 does with tweaking. I've also seen some pretty ugly 250 capture sent to me when I was working on ivtv, so your mileage may vary. My PVR-250 is the primary recording device in MythTV.
If you're considering the purchase of one or the other, I'd suggest the 150 simply because it is cheaper and the picture quality varies on both cards anyway.
My MythTV box is 2.4GHz P4 with 768MB RAM (256 didn't seem to be enough when I had only software encoding cards and Linux 2.4, lots of thrashing). GeForce FX 5200 outputting over DVI to an HDTV, I use a 720p modeline for gui, then a generic 640x480 modeline for playback.
Brightness: PVR250=115 PVR150=90
Contrast: PVR250=66 PVR150=70
Saturation: PVR250=62 PVR150=90
Hue: PVR250=0 PVR150=23
720x480, 4Mbit - 9.6Mbit, dnr_temporal=4
The 150 is still slightly oversaturated and a touch darker.