W4S STP-SE DIY mods

Replacing the input/output op amps will change the sonic signature of the device. Choice is dependent on what you’re looking for. I have tried the dexa Special Edition op amps, which have a very fast 60V/us slew rate. While better than some other choices, these had somewhat of a shouty midrange and the sound was somewhat flat. Also, it has somewhat of a lean bass (just a little bit lean). I don’t know if the normal 20V/us version is any better.

Burson’s are FET devices and have that somewhat rich flavor. I didn’t really like them. While they had good resolution, they just didn’t "sing" and I found myself losing interest in a song very fast with these. I did put 175 hours on them, so they were definitely burned in. Others love the Burson. Personally, I think the best FET opamps are the OPA827. They are slow enough where you get bass/midbass, but fast enough where it’s not smearing the sound (like opa2134).

Sparkos was the best discrete I have found, but be careful. Too many Sparkos in a circuit and the sound could become too bright/rich or it could become too slow (if not enough power supply). They are awesome sounding Class A op amps if used correctly and convervatively. Two dual-channel op amps in the STP should be fine.  The Sparkos definitely have that "Class A" goodness/magic.

I have no experience with Sonic Imagery. That being said, these discrete Class A op amps require a huge power supply. The STP has more than enough to handle these discretes, so go ahead and experiement!

The idea of the STP is to be very transparent. It is meant to act more as a passive preamp, but with the ability to add voltage/gain if required on the output impedance side. If you want to keep this passive/transparent character without adding any sonic signature, the best "transparent" op amp I have used is the AD797. You can get a dual-AD797 on an adapter from Brown Dog. Make sure to solder a 47pf ceramic between pins 6 and 8 of each AD797 mono op amp for the high frequency compensation (reduces high frequency distortion).

I have also had benefits when soldering a 47uf Nichicon Muse + 0.1uf to the +/- power pins of an opamp.

Another thought is you could try replacing those green Nichicon Muse bipolar opamps with Elna Silmic. These would make a better DC blocking signal capacitor.

My initial thought would be that it wouldn’t make sense to have only one output buffer because W4S indicates this preamp is "fully balanced from input to output".

One theory would be that you mixed it up and there is only one input buffer with two output buffers. The one input buffer would only be used to create the "inverted" balanced signal when a single-ended RCA input is used. Then it goes to the discrete FET stage, which should be fully balanced/differential. Then there are two output buffers for both the normal and inverted signals of the XLR connection.

All op amps will have their own sonic signature and introduce some sort of sonic character to the signal. In my experience, the AD797 introduces the "least" amount of sonic change. It is a bipolar opamp and bipolar devices are usually less distortion and cleaner. The Sparkos is also bipolar, but it is biased heavily into Class A. Best way I can explain it is it introduces a "liveness" or "reverberation effect" to the music. However, if used too much you can overdo it and the sound can become harsh/bright (like if used for both input and output stages). The AD797 is more dry/neutral sounding in comparison.

FET opamps (such as OPA827, Burson, etc.) have more of a rich type of sonic signature. However, the richness/warmth creates somewhat of a smeared sound throughout the range. The smeared high frequencies could come across as bright/harsh. Depends on system synergy. Even the ultra-fast OPA627 has this type of sonic signature, and is more lean on bass/midbass (because the slew rate is so fast, it tends to translate waveforms to be faster then they should).

If I am right in my assumption above that there is only 1 input buffer, I would use a dual-AD797 on adapter for the input buffer to keep the input sonic signature as pure as possible. Then you could experiment with the output buffer. Use AD797 if you want the most transparency. Or try some Sparkos if you want to turn this into a Class A preamp. Or something else entirely.