Another Sytek MPX-4Aii. This one came in because of a “buzz” in channel 4. Turned out the buzz was actually a 150Hz oscillation in the differential output stage, which seemingly came from nowhere. Every component tested fine (there aren’t that many of them).
The answer didn’t appear until I traced the output stage circuit into a schematic (there are no Sytek schematics available online). Turns out, each half of the output stage feeds back into the other, in a kind of figure-8 NFB arrangement. Presumably, this lowers differential-mode distortion (and actually even allows for a certain amount of clipping correction), but more importantly, it floats both sides of the output, like an electronic transformer. That is, the differential mode output impedance is very low but the common-mode output impedance is nearly infinite. It took some digging, but the earliest version of this circuit I can find appeared in the HP 8903A audio analyzer (test signal output stage) around 1980.
The problem is that the gain of each half has to be PRECISELY matched for it to work. I had never really understood why the output stages of these preamps use 0.2% tolerance resistors, until now (the rest of the circuit uses 1%). Turns out even a 2% drift in ANY single resistor’s value will cause the output stage to go into oscillation. If this happens, the only solution is to buy all-new <0.2% 20k resistors and replace all 8 of them. In this unit, I did the caps and power supply decouplers too, for good measure.
With most of these units coming up on 20 years old, component drift is starting to set in. This is going to be an increasingly common problem.
These got famous because Steve Albini likes them, and they were designed by the same person as some of the better-known Neotek analog consoles. They’re super quiet and clean, with a minimal feature set. The “ii” revision (of which this is a very early example) has detented pots for gain matching, power supply updates for higher current, layout revisions for lower noise, input protection diodes, opamp sockets, and DC servos to eliminate coupling caps.
This one needed some new capacitors and new opamps in channels 1-2. I went ahead and replaced all of the local decouplers in those channels, and also replaced the bipolar coupling caps between the gain block and output stage on all channels.
Some opamp notes below...
People get weird about the opamps they find in here. Most Syteks that you see around have the “Burr-Brown mod” in channels 3 & 4. All that means is that the NE5532s in the gain block and output stage are replaced with Burr-Brown (now owned by TI) OPA2134s. The 2134 is a JFET input opamp that’s quite popular in audiophile circles, probably because it’s fairly forgiving of being plopped into circuits not designed for it. You’ll also see people talking about a Sytek “JFET mod” which is the same thing... or maybe different if Sytek couldn’t source 2134s in the early 2000s and used some other FET-input opamp instead.
In reality, the 2134 has higher distortion, higher noise, and poorer drive capability in this circuit than the NE5532 it was designed around. If anyone heard a difference (which is honestly not that likely since we’re talking about differences that are ~95dB down from full scale), it’s likely it was just more distortion, not “rounded tube-like transients” or whatever the mod was supposed to produce.
Then there’s the whole “why are there TL072s in here?!?!” thing that some people do when they open one of these. The TL072s are there to drive the peak LED, that’s it. That’s their only function. No point using something fancy in those positions. Some later models have NE5532s in these slots. Doesn’t matter.
It’s actually quite difficult to improve on the performance of the NE5532 in this circuit. This particular preamp now has LM4562s in the gain block for channels 1 & 2, which reduced 3rd and 5th distortion harmonics by about 3dB versus the original 5532 (not much when those harmonics were already 95dB down). The 4562 does not like this preamp’s output stage (its noise is higher in that position), so there are still 5532s in those slots for channels 1 & 2.
To keep the “Burr Brown” thing on channels 3 & 4 but improve performance over the 2134, I used OPA1642s (FET input) and OPA1612s (bipolar input). Specifically, 1642 > 1612 in each gain block and a 1642 in each output stage produced the lowest noise and distortion. These are modern SMT opamps that have to be mounted on small adapter boards, but they do well in this circuit. They’re rail-to-rail and have higher headroom than the 2134, with even lower noise than the 5532. Channels 3-4 now have only slightly higher distortion than channels 1-2, and the increase is all 2nd harmonic -- totally unobjectionable.
So now, channels 1-2 produce the “stock” sound but with slightly reduced distortion, and channels 3-4 produce slightly more distortion but have slightly lower noise. But I seriously bet nobody could tell them apart in a blind test.
by accident i have stumbled upon my favorite DI clean sound... the sytek MPX-4A mic pre! tons of character. thank you steve albini...!! #sytek #orangesqueezer #jazzmaster #curtisnovak #bossGE10 (at Los Angeles, California)
