#de-esser

Still, I was able to improve noise and distortion performance on this unit by changing the input amp to a FET input OPA1641 (which also raised the input Z a little), and the output amp to a bipolar OPA1611. I chose a FET amp for the input stage because the 1641 has only slightly higher voltage noise than the stock 5534, but exponentially lower current noise -- which matters quite a bit, given that the surrounding impedances of the stage are fairly high. At the time the 526A was designed, the 5534 was just about the quietest opamp available, and the TL072 was the FET-input alternative -- its voltage noise is 5x the 5534′s. The parts we have today make FET opamps a better choice in many older circuits that use bipolar-input amps, where the source impedance or feedback resistances are >2k. The voltage noise is comparable, but the current noise is exponentially lower, and this is where most of the benefit lies.

I replaced all the resistors in the signal path with low-noise metal film, which helps a lot since there are lots of high values and voltage dividers with large attenuation. And I replaced the tantalum feedback shunt caps with bipolar electrolytics (tantalums have pretty high distortion, and modern audio-specific bipolars are much better).

These pieces are really nice to work on -- the top and bottom plates of the chassis come off, so the PCB is easily accessible from both sides. At this age, every electrolytic cap should be replaced, and the supply caps should be bumped to at least 2x the stock values. Opamps have local decoupling caps stock, but they should be increased to at least 100nF everywhere. I used even higher values in positions that pull big current pulses, like the output amp and the sidechain rectifier. Quieting the supply is important because there are a couple places in the layout where a power rail is used as a shield trace to protect a signal trace from interference -- most notably, on the gain pot traces. All the opamps are socketed, so no soldering is required to change them out.

For comparison, after modification this preamp is quieter than my ISA Two. Hum/buzz is completely inaudible. That’s quite a feat, since the ISA’s noise floor is insanely low already, so low that a Cloudlifter actually increases the noisefloor when running ribbons or an SM7B into it. I wouldn’t hesitate to use an SM7B for speech or singing through the Orban, with no need for a mic booster (or the phantom supply it’d require). The de-esser is just an added bonus. If I needed a compressor, I’d be just as happy patching the Orban after the compressor on a channel insert. Either way, it’s not adding anything to the noise floor in any normal configuration.

The x63X series has a reputation for being “dirty” and “noisy” but I’ve not found that to be true. They’re really quite well-designed and still totally usable in their stock form. Input and output headroom is about +21dBu, full pro line level, and I suspect that when people encounter noise it’s because they’re driving them with tiny consumer- or instrument-level signals. Proper gainstaging reveals the true capabilities. But it’s also easy to improve them.

Most of these units are old enough now that they need re-capping. There are two positions in the 263X where polar electrolytics are used to couple signal, and these can end up slightly reverse-biased. It’s best to replace those with high-quality bipolar electros. I haven’t found that massively increasing the supply reservoirs has much benefit (though it doesn’t hurt either), given that the supply is regulated downstream and current draw isn’t enormous. I did bump the size of the caps after the regulators, though. And the 1uFs in parallel with the big reservoirs are there as RFI bypasses, since large caps tend to be more inductive. You can improve the performance in that role by replacing them with 1uF stacked non-inductive film caps, though most big modern electrolytics have much better decoupling characteristics than the original components (I use Panasonic FC series because of their low ESR, high ripple current, and long life + temperature tolerance).

All opamps have 10n ceramic local decoupling caps, a really nice thing that allows you to drop in most modern opamps without much worry. However, you can do better by replacing these with 100nF non-inductive film or ceramic caps, especially if you’re swapping in higher-bandwidth opamps.

For opamps, U1 needs to be a FET-input type. As usual, I find the OPA1642 works wonderfully here, because of its low noise and current draw. U7 is the output amp, which is a Signetics NE5534, stock. This is a great opamp and doesn’t need changing, though it is somewhat overcompensated -- best to reduce C24 to 33pf or 22pf, though you should increase the local decouplers to .1uF if you do this. U5 is the crossover opamp, and it does pass signal to the output when you’re in high-frequency mode. The stock 4558 is really too noisy and slews poorly for the application. I replaced this with an NE5532 and found that both noise and distortion performance improved by about 5-8dB.

The front and rear inputs are identical, except that the rear input shorts out the input gain control. The input impedance for both is 390k, which is not really optimal for either high or low-impedance applications (most DIs have a 1M input Z). I opted to DC-couple the input (shorting C2) and changed R2 to 100k to reduce input noise a little further. 100k is still easily drivable by the typical guitar pedal, for example. You can also move C1 to the other side of R1 (i.e. in parallel with R2), to create a better input RFI filter. This makes sense because most modern sources you’d drive this with have very low output impedance, so strapping a cap across the input doesn’t do anything because it’s just in parallel with the cable capacitance. Putting it after R1 creates a lowpass filter -- and no, it won’t cut into the audio range; the corner frequency is about 1MHz. Even if you were driving the unit from a passive guitar pickup, the corner frequency would still be up around 75kHz.

Lastly, all x63X models almost universally need calibration, just due to normal drift from age. I found that the VCA symmetry trim was off by quite a bit in this 263X. I don’t think that it’s necessary to replace the VCA -- the 1252, aka 2150, is fine -- but most people replace them with trimless THAT2180s, which explains why they see a performance increase (you don’t need to calibrate the symmetry on the 2180). You get most of the same benefit by just calibrating the unit and leaving the 1252 in there.