Electronic States

What mystified me about the WA12 is the bizarre input impedance spec. Warm claims that it's 600 ohms, switchable to 150 ohms with the 'tone' switch on. What?? These are ridiculous numbers -- the vast majority of mics want to see a load of 1k or higher, and most mics will suffer pretty severe signal loss with loads under 600 ohms. And as it turns out, these numbers aren't even correct. It's pretty obvious that Warm just quoted the datasheet numbers from their OEM Cinemag input transformer. Cinemag specs their transformers by the expected source impedance on the primary winding, and translates that by the square of the turns ratio to the reflected source impedance on the secondary winding. So unless the secondary of this transformer is loaded with a 10k resistor, 600/150 ohms is NOT the input impedance of the circuit. Transformers don't have a fundamental impedance. They have an impedance ratio, which is the square of the turns ratio. The load you choose for the secondary is what determines the input impedance on the primary.

The stock input impedance I measured on the WA12 mkII is 1.05k, or 600 ohms with the tone switch in. That's actually fine, much more reasonable than the quoted specs. But it still doesn't add up -- if pushing the tone switch doubles the transformer turns ratio from 1:4 to 1:8 (and increases the gain by 6dB), then I should be seeing an impedance 4x lower when it's pushed, not slightly more than half. The other thing I noticed is that the input impedance measured 1.2k with the pad on, regardless of the tone switch position. A ha -- there's a parallel load on the primary side of the transformer that's loading down the input regardless of how the transformer itself is configured (beyond just the 6.81k phantom feed resistors, which have to be there).

Turns out, Warm wired the pad switch so that the pad resistors (a balanced voltage divider of three resistors: 620-140-620 ohms, R35/37/36) are strung across the input at all times -- whether the pad is on or not. The switch just connects the input transformer to either the outside of the network (off), or to the ends of the 140-ohm resistor in the middle (on). There's no need for that. It's worth looking at the way John Hardy wires their pad switches on the M2, since it's basically the same circuit topology. Without any extra parts, and by just changing how the DPDT switch connects to the two series resistors, the Hardy ensures that the pad is completely out of circuit when it's off, and presents no additional load to the mic.

I rewired the pad to according to the Hardy schematic. Boom, the input impedance was then 3.6k with the tone switch off, and 950 ohms with it on. Those numbers actually make sense. But it revealed a new problem -- the input impedance changed dramatically depending on what frequency I used to measure it (I had only used 1kHz up to this point). Specifically, it dropped to under 200 ohms at 10kHz! I felt that was unacceptable, so I started looking at the load on the transformer secondary.

(For our purposes, the load presented by the discrete opamp can be disregarded, since it's quite high, in the megohm range. I did all my measurements with the opamp removed from the socket.)

API 312s have a "zobel" network on the transformer secondary, which is there to damp the transformer's tendency to ring at high frequencies. On the 'canonical' schematic, the zobel network is a 5.1k resistor off the signal line, in series with a 220pf cap to ground. The zobel is in parallel with the secondary's load resistor -- not shown on the canonical schematic but typically found in the vast majority of old 312s -- which is typically 150k. Note that a 150k load on the secondary of a 1:8 transformer would reflect back to the primary as 150k/8^2 = 2.34k, or 150k/4^2 = 9.37k for 1:4! And in fact, 150k is what the WA12 mkII has (R13). So why was my input impedance still so much lower? The zobel network.

Turns out, Warm has a totally insane zobel network of 910 ohms (R14) in series with a 1.5nF (1500pf) capacitor (C19). That's not a zobel, that's just a straight up dump of treble frequencies. The zobel's turnover is so low that it's affecting the impedance across the entire audio band. It's honestly probably the single biggest factor in the "dark" sound of this preamp, and it's way worse with the tone switch in -- maybe that's why they called it a "tone" switch instead of an impedance switch or gain switch. The extreme values are just not necessary to get the (pretty decent) input transformer to behave.

I experimented extensively to find out what the optimal zobel network actually is for the WA12 mkII, and came up with 4.99k in series with 330pf. That's pretty similar to the canonical API spec. In fact, if there were no tone switch and this transformer could only be used at 1:8, the API values of 5.1k-220pf would be ideal. But the transformer performs significantly worse at 1:4 (again, odd that this is the default mode here), so something a little heavier was necessary. I also ended up reducing the load resistor from 150k to 100k, which helped damp ringing quite a bit. After these changes, there's near perfect square-wave performance in 1:8 (tone in) with a gentle roll-off above 20kHz and no peaking at all. At 1:4 (tone out), there's about 10% overshoot on square waves, but no ringing, and an innocuous +2dB peak at 27kHz that just adds a little bit of 'air' to the sound.

Now, the input impedance is 3.5k @ 1:4 (tone out) up to the top octave, where it drops to about 3k. At 1:8 (tone in), it's 1.05k and drops to 650 ohms at 20kHz. With the pad in, it's 1.2k in all modes and at all frequencies. Very nice! I now basically always leave the tone switch in, using the vintage-correct 1:8 input ratio, which has the best frequency and time-domain response, plus better SNR. If I want a little bit of extra 'air' in exchange for slightly worse time-domain performance and lower SNR, I turn the tone switch off and use 1:4.

If you're going to the trouble of doing this work (really just moving two resistors and replacing two other resistors and one capacitor), it's also a good idea to hand-match a new pair of phantom feed resistors for best common-mode noise rejection. I found the stock 6.8k parts in the WA12 mkII to be about 1.2% apart from each other, when the IEC phantom specification says that it should be less than 0.3%. I make it a habit to match pairs of 6.81k metal film resistors to within 1 ohm (<0.03%) when I'm working on mic preamps. No need to buy parts at those tolerances -- the absolute value isn't important, just find any two in a pile that are super close to each other in value.

Unlike the larger models in this series, there are no solid-state signal components here. The boost is just a switchable bypass cap on the first triode stage. (From what I understand, the 'gain' circuit in the larger models is a Pigtronix distortion pedal circuit, but I don't have one of those so I'm not going to trace it.)

I would not recommend this amp as a platform for mods. It's pretty troublesome to get the PCB out of the chassis. If you do pull it, be aware that you need a Torx screwdriver to remove the line out/footswitch jack panel before you can get to the PCB.

The amp is underfiltered, and hums noticeably. It benefits from having a small choke (3-4H and 50mA is plenty) ahead of the main reservoir, and a smaller filter cap before and after the choke (maybe 22uF in each spot). This would make the power transformer run happier and kill all of the output stage hum.

Another problem is the way the volume control (really a gain control) uses a second gang to shunt a bypass cap on V1B's cathode. Because it's a linear 1M pot in parallel with a 22k resistor, the second gang basically doesn't do anything until the last 5% of rotation. When it finally wakes up, the suddenly fully-bypassed V1B drops its output impedance by 50%, which raises the corner frequency of the tone control by an octave. In other words: the amp will get suddenly, noticeably brighter in the last 5% of the volume knob's rotation, and turning down the tone control won't really compensate. I cut the wiper trace from that pot gang and replaced the 22k with a 10k resistor, so the tone is consistent across the whole volume knob. The preamp overdrive is really pretty nice on this amp.

It's ludicrous to have a 47k grid stopper on the power stage -- paralleled triodes have twice the Miller capacitance of a single one, so this is creating a treble roll-off down in the audible range. It should really be 10k or less. Shunting treble before a tube stage that's distorting almost never sounds good, IME. It just reduces articulation. The high value may have been an attempt to reduce blocking distortion by starving the 12AU7 grids of current when they go into conduction, but blocking distortion isn't a problem on this amp!

With stock biasing, the 12AU7 triodes go into grid conduction WAY before cutoff. You can cool the bias to get marginally more symmetrical clipping, and power output is largely unchanged or even slightly increased -- 680 or 750 ohms for Rk works well if you go that route. The reason power doesn't drop is that plate voltage goes up -- the PT secondary is poorly regulated and sags a lot under load. You might see 280V B+ with the stock biasing, but 295V with a 750-ohm Rk.

The biggest disappointment here is that the power tube distortion doesn't sound very good. It has a ratty, crackly breakup that seems to ride on top of the note (this is not the speaker or tubes, because it sounds the same into other cabs and with all 12AU7s). Looking at the scope, this happens when the 12AU7 goes into grid conduction and the positive peaks get chopped off. But that shouldn't sound this bad! From what I can tell, the output transformer is probably responsible -- it's very likely undamped ringing of the high harmonics produced by clipping. Obviously the designers were aware of this, since they added a 4.7nF bypass cap across the OT primary, but unfortunately this doesn't help much.

I may try installing a Hammond OT at some point to see if it cures the nasties. Update: I installed a 1750C, which is a 5W Champ transformer, 5k + 8k primary taps and a 3.2-ohm secondary. This allows using the 5k tap with an 8-ohm speaker to get a 12.5k primary Z, or using the 8k tap with a 4-ohm speaker to get a 10k primary Z. I found that power output went down slightly (maybe 5%, not enough for a noticeable volume difference), but the overdriven tone was much, much nicer.

Another thing you can do is drop in a 12BH7 instead of the 12AU7. It's larger, and requires more heater current, but the power transformer seems to handle it just fine. Power output goes up maybe 10% with no other changes. I think it sounds a little nicer when it breaks up, too.

The stock speaker is actually pretty good! It looks like it was OEMed by Eminence (in Dongguan, not KY, no "67" EIA code), almost certainly an 8-ohm paper-cone version of the 820H. It's far beefier than an amp like this would require. It's pretty dark and throaty sounding, and has a good amount of excursion with a fairly heavy cone and a large spider. Zero cone breakup, and makes the amp sound larger than it is. The problem is that it's super stiff out of the box, and in a 1W amp like this, it will literally never break in. I pulled it and thrashed it with a 25Hz tone generator for 10 hours at 10W, and it loosened up quite a lot. Resonant frequency went down almost an octave and it sounds more extended in the treble now.

I also tried a Weber Alnico Signature 8 (well broken-in). Versus the stock speaker it's more 'crunchy' or 'raspy' with noticeably truncated low end and more 'honk' in the low mids. Much more like you'd expect a small amp to sound, and far more compressed than the stock speaker. Equally good, but very different. I'd say the Weber records a little better, but the stock speaker sounds more balanced in the room.

Here's a comparison of the two speakers on a re-amped guitar track, recorded with an R-121 and API 312 (same gain on the mic, so you can see actual sensitivity differences):