Sync Stripper Circuit Design Flaws and Easy Fixes

[NewSchoolBoxer]

Overview

The main flaw is proper 75 ohm termination on input and output. Most chips have extremely high input impedance, maybe 8 kohm for a BJT and gigaohm range and above for FETs. The output is very low impedance in the 0.3 to 3 ohm range. The reason is for maximum voltage transfer. Can do the Rload / (Rsource + Rload) calcs yourself. If the load is at least 50x greater than the input, you get at least 98% voltage transfer. The two downsides are reflections from the impedance mismatch and power loss.

In the MHz range, the reflections matter and degrade the video. If you read over CRT mod discussions, leaving out the 75 ohm termination is the difference between video displaying and not displaying. If the impedance from source to load match, there is no reflection. The worse the mismatch the greater the reflection and more damage it causes.

Input Impedance

Most designs are good about putting a 75 ohm resistor in parallel to ground before the composite video or luma input. The formula for resistance/impedance with two components in series is (R1*R2) / (R1 + R2). If R1 is 75 ohm and R2 is 50x greater, the effective resistance is just under 75 ohms and that's close enough. When you match a 75 ohm cable, i.e. video coax, with 75 ohm input and output impedance, you get no reflections, the expected 50% voltage transfer and other benefits.

The much smaller flaw I haven't seen addressed for Composite video input is using the ~500 kHz low pass filter that the LM1881, EL1883 and ISL59885 datasheets all tell you to use to get a higher quality sync output. It's a 620 ohm series resistor followed by a 510pF capacitor parallel to ground before the AC coupling capacitor. 5% tolerance components are fine here. You give up 75 ohm matching on the input but get a higher signal to noise ratio in the end. I think it's debatable if the filter gives a cleaner signal for Luma input but I doubt it would be worse.

In other words, add the 75 ohm parallel resistor or add the filter. Your choice. Either option is valid.

Output Impedance

On the output, the 75 ohm resistor goes in series instead to reach 75 ohm impedance since 0.3 to 3 ohms output impedance is effectively 0. Thus there is a problem when you add a different resistor in series. The design flaw I've only seen handled once by a Japanese designer is with the 330 or 470 ohm resistor on the 3-5V output to drop it to 300-600mV for SCART televisions. Leaving that as is a big mismatch. Can say 15kHz is too small a wavelength to cause damaging reflections but the typical +5V rail powering the sync chip is the dirty power from the console and the csync output isn't necessarily free of Composite/Luma artifacts or RGB crosstalk.

So if you have a 470 ohm resistor in series, what parallel resistor to ground do place after it to get effective 75 ohms? 89 ohms and the closest common value is 91 ohms that sells for 4 or 5 cents. If you only had 100 ohms, is close enough. Ideal pairing with 330 ohms is 97 ohms so round up to 100 there as well.

The impedance mismatch was hiding in plain sight. Cause everyone did it, I didn't think about it being wrong.

Related Discussion

Not limited to sync stripping is the colossal mistake of leaving out a bypass capacitor on the +5V rail, nor is there 75 ohm termination on the input (!) or output. Don't buy any of these piece of crap stripper circuits from Retro Gaming Cables! Three design flaws in one and the worst (read: cheapest) sync stripper, LM1881. I use ISL59885 on PS2 myself.

One last point, I did see sync output impedance discussed here from Amiga's TTL sync where the R1 resistor drops TTL to 75 ohm voltage. I don't know Amiga circuity so I have to side with viletim on the extra current draw from R4 current dividing potentially being a problem. In our sync stripper case, the stripper itself already consumes far more current than the impact of the extra resistor and a +5V rail isn't gimped on current sourcing like the video chip or a microprocessor.