Finally I got around to figuring this out. Had to poke around a little bit.
Ran across this schematic in the land of diy forums. It’s for the G2 Rokit 6 amps, which had the notorious black goo of death problem, so totally different design (well, probably not that different… these budget companies never really redesign their stuff if they don’t have to, unless they can make it simpler/stupider), but it struck me that their limiter was using just one bipolar transistor right after the input buffer, so I went looking for lone transistors near the input, and lo and behold, there is only one (Q5, just above the right-most large ceramic capacitor).
Oscilloscope revealed bits of signal being leached off, so it seemed I had found the culprit. Removed Q5, but both amplifiers suddenly muted… hm.
Traced signal from top-most pin of Q5, found it running to the ‘mystery chip’ just above, U5. Looks like it’s some kind of gate circuit: when it receives enough input voltage at pin 5 (lower left pin), seems like it opens pin 4 (upper left) and passes 5V, which is what the amp circuits need to unmute and shut off standby (datasheet for TDA7296 actually specifies 3.5V minimum to disengage mute/standby function).
This reveals a couple of things about KRK’s design. First off, the limiter is running all the time. Even at relatively low levels it seems to be leaching signal, and globally for both treble and bass. Secondly, the mute circuit depends on the output of the limiter in order to engage, so the limiter needs to be operating rather aggressively in order to unmute the amplifiers. This is why, when I removed Q5, the mute circuit wasn’t seeing any input. I’ll say again: IF YOU HEAR ANY SOUND, THE LIMITER IS LIMITING, A LOT. 🤦♂️
So… remove the transistor (hot air is easiest), remove the IC, and short pins 3 and 4 as shown (I used a little piece of wire clipping).
Now the amp engages immediately, without any lag. There are electrolytic caps decoupling the input to the mute/standby pins in the amplifier, so there won’t be a big pop. Can see the voltage at pins 9/10 of TDA7296 rising and falling gently with on/off.
Things sound ‘snappier’, more precise––these are some real speakers now! This project has officially reached it’s zenith. EDIT: I’m shitting my pants at how amazing this sounds. All my previous mods really seem to have come to life now, the low end is incredibly clear and integrated (it’s like all the overtones just lock), I find I’m listening with the volume a bit lower, and both speakers sound more balanced (the right speaker always felt just a little softer, and they’re ‘matched’). My reference recordings sound uncanny. And just over a little transistor…
*PROCEED AT YER OWN RISK* I listen primarily to Classical stuff, I’m not pounding EDM or anything. I assume this circuit is here for a reason.
If you want someone to do this for you, check out this guy: https://www.danieldialatone.com/gallery. Used to be a head engineer for Black Lion Audio, studio monitor ‘enhancements’ are his specialty. I just use the TI SoundPlus series of opamps (which I think sound great, enough of an improvement), based on their functional equivalency to what’s on the board (FET/bipolar, comparing input impedance, etc) he actually hand selects different opamps for their unique flavor. If my amateurish experimentation got me this far, just think what a real pro can do!
This was not my first major modification project (my first was an old Presonus MP4 headphone amp, AMAZING, which I might write about another time), but it was significant insofar as these are my studio monitors, and I basically wasn’t using them. They’re just not clear enough. Muddy and incoherent, kinda bland, can’t hear anything clearly, tried some corrective EQ, only mitigates. I primarily work with classical and acoustic stuff, so quick and snappy good. I know the Rokits are favorites of EDM folks, I don’t really care about that stuff (sorry), but these were my first monitors and I didn’t know any better…
Anyway, I’ve been interested in audio modification for a while, and the arguments that swirl around them––rather than constantly buying new stuff, buy used stuff on its way to the garbage bin, pimp it out, and end up with a device that rivals high-end gear. It makes complete sense to me that a company with a profit margin, like KRK or Presonus or M-Audio or RME, would cut corners where they think it’ll be least noticed by 95% of folks in most normal-use situations (i.e., a singer-songwriter at home on their bed crooning into some crappy Chinese mic). It shouldn’t be wrong to “tune” your equipment to sound the way you want.
To me the mindset is very much like what people think about in the instrumental world. I grew up with a Yamaha G2 grand piano, a rather petite instrument, about 5’7″ or so. Yamaha makes spectacularly reliable actions, but the sound usually leaves something to be desired. The hammers were rock solid (not in a good way), so when I was in college we had a piano technician replace them with a set custom made for the piano by Ari Issac in Toronto. The new set completely transformed the sound––there was color, finally, shading, resonance, a dynamic range (especially soft)––and over a decade later those hammers continue to break in, revealing new sounds I never thought I’d hear from such a small instrument. Recently I had the bass strings measured for replacements. Yamaha is not known for making great bass strings––they sound a bit like surgical tubing drawn taut. I haven’t yet heard the results, but I will when I return home to visit my family for the holidays. Anyway, modification is a very musical thing to do, so I don’t think people who are interested in modifying their audio equipment are somehow phoolish to think they can eke out every last drop of good sound. Your milage may vary, and that’s ultimately all that matters.
Okay, if you pay $50 for one capacitor made with raw sheep’s milk or whatever, you are an idiot.
Anyway, I wasn’t using these Rokits much, plus the warrantee was out. What could go wrong?
Nothing wrong with them. They work just fine. This is the internal amp I was going to modify. The actual amplifiers are two integrated circuits (TDA7296, one for tweeter, one for woofer) pasted to the heat sink in back, the rest of this circuit is power supply, input buffer, and crossover. Power supply in the back (large capacitors and rectifier), input section in the front right using bipolar caps and two NJM4580s, crossover section in the middle front left uses three TL074 (and the green mylar caps). The yellow caps up front appeared to be for the corrective filters on the back, and I never use that function, so I figured I’d leave them alone.
Only way to remove the amplifier board was to desolder it from the input PCB. Definitely ripped a pad here, but it wasn’t a problem to repair later (just used a jumper or something).
[There’s also a “mystery circuit,” seen just to the top left of the input board that appears to supply 5V for a digital… something. Wasn’t connected to anything, so I wonder if KRK has another line of bluetooth products that use this amplifier board? I couldn’t find any. I left it alone. EDIT: if you check out my next post about this, I discovered this is part of the mute/standby and limiter circuit, both of which I figured out how to disconnect!]
The electrolytic caps are all easy to replace. They all have capacitance and voltage ratings written on the case––you can go bigger on both, just not smaller, but remember you’ll have to deal with larger cans. The mylar caps are another matter. They didn’t seem to have much of anything written on them, plus everything was bathed in a disgusting sticky glue which seemed to obscure any writing that may have been there. I used a capacitance meter to measure each cap after I removed it.
(apologies for the chicken scratch, and no offense to chickens)
Couldn’t be exactly sure what the value was of each cap (they have a tolerance that usually is within 20% of the stated value), but rounding to the closest available value on Mouser or DigiKey usually revealed what they must have been.
Another view of the mylar caps (green).
To remove the op amps I very gently grasped them with a pair of pliers and used a hot air rework station to melt the solder––add some flux paste around the pins to lower the melting point of the solder. There’s no need to pull, the chips weigh nothing, they basically float up off the pads the moment the solder melts (which usually takes about 5-10 seconds).
Clean board (all the SMD resistors, diodes, capacitors are left in place, of course):
Now had to remove all that glue:
Added extra supply rail decoupling, 0.1uF ceramic capacitors from each power rail to ground. Seemed there was one ceramic cap between the rails for both input opamps, extra won’t hurt, and would certainly help prevent stability problems when I put in the faster opamps. I used OPA1642/1644––FET input opamps, safe replacements for NJM4580. The OPA164x family is, according to TI, descended from OPA2134 (a popular chip among modders, since it can pretty much replace anything without causing issues), with lower self-noise and current consumption.
The basic guideline for choosing replacement opamps is pretty simple: you can generally use a FET input opamp to replace anything. The quietest opamps out there, speaking in terms of their inherent self-noise, are bipolar opamps (OPA1612 is TI’s flagship, but the NE5532 is the ‘classic’ example), but they tend to be a little fussier and will oscillate if they’re not situated in a complimentary circuit. A lot of this has to do with opamp’s inherent resistance to current, or input impedance, which relates to the bias current of the input transistors in the chip. FET opamps use FET transistors at the input, which require very little bias current, so the input impedance is very high, 10s of megaohms. Bipolar opamps require more current flowing across the input transistors, so their input impedance is necessarily lower, 20kohms or so (some bipolar opamps like MC33078 or NJM4580 have higher input impedances, 200kohm to 5megaohm range, but these are are chips I tend to replace…). I could have used the very quiet OPA1612 (20kohm input impedance) at the input here, since the input impedance for this circuit is only 10kohms. Provided the power supplies are properly decoupled (as I’ve done), it would probably be stable, but the problem is that this chip only has low self-noise at relatively low impedances. The higher the input impedance, the more noise at the input, and extra expense becomes self-defeating, as you can see in this graph.
At 2kohms and below, the bipolar chip (in this case, the OPA160x series) has lower self-noise than the FET input OPA164x, but above that, as the current across the input transistors drops, the chip becomes noisier. So using a fancy chip like OPA1612 in this kind of setting, at $7/ea, doesn’t really make sense.
So, onwards, I replaced all polarized electrolytic caps with Panasonic FR, and bipolars with Panasonic SU. I kept all the values the same, although, come to think of it, I might have done well to at least double all the signal path caps, especially because I bypassed every one with a polypropylene film capacitor (also Panasonic, though I can’t remember which type). Mylar crossover caps were also replaced with Panasonic polypropylene. A couple values weren’t available on Mouser, so I used a Vishay polypropylene cap. Were I to do this again, I’d use WIMA MKP/FKP, I just didn’t know where to look when I did this.
So I plugged it in and… no smoke! So how does it sound? I placed both speakers side by side (as pictured at the top of the page), one modified and the other stock, plugged them into the two outputs of my interface and routed an identical mono mix to both (I think it was of Mitsuko Uchida playing a Mozart concerto). “Panning” back and forth between the channels, the difference was NOT subtle. One gave the impression of having a wet towel wrapped around your head, the other didn’t and sounded like music. In fact, more than that, in the modded speaker I could perceive a sense of horizontal depth even with a mono signal, like distance front-to-back between instruments on the stage. That surprised me. So yeah, it was worth it. I modded the other one promptly.
These are the 5-inch speakers, so they still have a limited low end. I’m not sure increasing signal path cap sizes would have helped that much. I understand there is a limiter circuit built into this amplifier somewhere that could be disconnected (EDIT: I did it). Again, I don’t listen to EDM or anything with super heavy bass that could blow out the voice coils, so I’m not worried about damaging the speakers, plus I’ve heard of folks doing this in the Yamaha HS8s to great effect. Finding it would be a matter of poking around with an oscilloscope, so maybe I’ll do this at some point.
So how do they compare to other more expensive speakers? I don’t know, I don’t own anything else (I compared them to some Mackie 828s, which got smoked by my KRKs, but that was unfair because the Mackies are inherently slow and noisy), haven’t done a side-by-side with someone who does. [EDIT: you know who has is this guy, who used to be a head modification designer at Black Lion Audio. He mods these speakers professionally, and others, has done comparisons with hi-end stuff like Genelec and says these modded KRKs hang right in there, and represent the absolute best value in terms of the sound you get!]