I want to write down also my minds to this topic.
THS7314 vs. THS7374The main two differences were already mentioned here. First, the THS7374 is a four channel amplifier and the THS7314 a three channel amp. Second, the THS7374 has a bypass mode for the internal low pass filters.
The main similarities are that both chips have 6dB amplifiers and a low pass filter (which is designed for 480i content) on each channel.
I switched to the THS7374 due to the following reason:
- With the four channel I have the opportunity to buffer the /CSYNC, which could be designed for TTL syncing with high Z termination or standardized sync signal with 75ohm termination.
The example in the spoiler shows the THS7374 in a N64 with /CSYNC buffering designed for standardized sync.
- The low pass filter has a bypass mode (discussed below)
The drawback is that the 0.65mm pitch of the THS7374 is not that easy to solder in a DIY project compared to the 1.27mm pitch of the THS7314.
I don't want to go into detail, but also the filters are implemented in a different way in both chips. However, what is more important is to understand what the purpose of the filter is.
- The first one is the reconstruction of the filter, behind a R2R ladder output, e.g. However, most people are interested in sharp pixels.
I want to mention at this point that in my point of view the 'shimmer' at the 'Link'-picture from @retrorgb on the 'filter on' side is not related to the attenuation near to the cut off frequency! It is more related to the phase shift of high frequencies due to the filter. - The second one is to band limit noise (signal processing / information theory). I think this does not come into effect because we discuss closed system with short transmission channels.
To keep it short. Putting the THS7374 filters into the bypass mode is a good idea for retro consoles. I want to admit that my actual boards are designed such that the THS7374 is in filter mode and can be putted into bypass mode by closing the solder jumper. I will change that to have it the other way around in future.
By the way, in bypass mode there is also a low pass filter present in each channel. The cut off is something around 150MHz and this filter has a appr. constant phase shift for 480i content.
Brightness issue and Ghosting on 1Chip consolesI want to clarify what I mean with which word. So I try to make a brief definition first to separate especially ghosting and ringing:
- brightness issue: if the RGB signals, S-Video signal, Composite Video signal going into the TV (or what ever device comes first at the end of the cable) has a peak to peak voltage (Upp) over the standardized Upp the picture is too bright. This is the case for all stock / unmodified SNES consoles with the S-CPUN, i.e. for so called 1Chip models and Mini/Jr.
- ringing: A video signal with to high Upp (over the standard one) and also a signal with in proper termination for the cable (in our case we need a 75ohm in serial inside the console (output impedance) and a 75ohm resistor to ground inside the TV (or what ever device comes first) (load impedance)) may cause reflections of the signal.
- ghosting: Normally this term can be used similar to ringing, but for the SNES with S-CPUNs I want to keep it separately. The S-CPUN overshoots the outputs for RGB. This has an ghosting effect. Ingame-pictures to that issue come later. Here is an oscilloscope measurement of a clear white image made by rama (user at circuit-board.de)
The birghtness issue can corrected by either the 'one-resistor-method': lifting S-CPUN pin 155 and reconnect this pin back to Vcc (+5V) over a 20 ohm resistor.
Personally I can only recommend to do so for very experienced modders. The risk to damage something is quite large for this pin! (due to the orientation of this pin to the cartridge connector). Not practical for most DIY modders!
This issue can also be resolved using the 'three-resistor-method' as outlined
on assembler games. This method is quite easy to implement and recommended for most DIY modders. That's why I have pads for these resistors on my designs, too.
Never implement both methods inside one console - always implement either one or the other one!
With the brightness correction eliminates at least one reason for the rining. If someone uses modding boards or follow DIY modding guides carefully as well as using proper RGB cables, a proper termination is also not an issue here

Finally going into the ghosting effect. I just want to put a few pictures into the spoiler. All pictures there are made by rama. He used inverted colors show the effect more obvious. All pictures are taken with the S-Video output with proper termination.
A clear white image (colors inverted):

Super Mario World 2 (colors inverted):

Super Mario World (colors inverted) - notice the 'second' Mario right of Mario:

I want to point out that this effect is related to the overshoots on steep slopes and cannot be resolved by 'three-resistor-method'! However, the 'one-resistor-method' reduces these overshoots such that the ghosting is reduced to a non-noticeable effect!
retrorgb wrote:
BuckoA51 wrote:
While I have your attention too, have you ever tried Borti's method for fixing the brightness on the Mini?
To be honest, I think that's just out of my skill level. I tried doing a SuperCIC mod and thought I did everything right - I was super careful and delicate...and the pin still snapped off the chip when I lifted it; They're extremely brittle.
If one of the more talented modders wouldn't mind trying it, I'd love to test one and do a comparison.
I don't know where it was (it was in another thread here) and don't know if you realized it: I would like to send you my SNES for testing if you want to.
Woozle wrote:
The ghosting (at least, what I think is ghosting) is really noticeable to me on my SNES mini when playing super Mario world with the bright blue background in the first few levels (using OSSC or xrgb-mini on an lcd). It's irritating enough that I'm going to try the pin lifting mod on my spare SNES when I go in to fix the power jack.
It would be nice if there were an easier fix for the overshoot, I really don't like lifting IC pins :/
What I never tested so far is what happens if one add an additional load capacitor (something around 120pF - 270pF) at the S-CPUN RGB outputs. I do have the caps here but still haven't found the time to test and measure it.
Guspaz wrote:
Seems very difficult to do, and I do believe that the recent RGB boards like Voultar's take care of attenuating the brightness for you already?
As mentioned: also my open sourced designs has places for these resistor.