STK chips

Hello Vivek,

Modern STK amplifier modules are indeed quite good.
The specs are good but the modern high power models are not rugged - ie they are usually rated for 8 ohm load only, and if running full rated supply rails, a 4 ohm load will likely kill them.
(I get to repair plenty of such amps that have failed because youngsters have connected a extra pair of cabinets !.)

For economy, repeatability, servicability, availability and ease of implementation they are probably hard to beat.
Some have handy features such as muting and limiting also.
STK modules are pretty much standard equipment in a lot of mid-fi audio gear.

Typical applications use an under rated transformer to give high transient ability, and power rails that sag somewhat according to output, thus giving a level of protection to the module in domestic audio gear.
Adequate heatsinking is important, fan cooling even better, or place the heatsink in the air path of the tuning port tube.

I have repaired subs using stereo modules configured in bridge mono mode, and performance/price work quite well.

Although I have not tried the experiment ( I have 2 STK4048mk2 modules in stock to try out one day) I expect/suspect with good layout, good power supply and good input drive stage they ought to perform rather well.

Another approach is to use STK DRIVER modules driving seperate output transistors for higher power and/or lower impedence loads.
The specs for these modules is rather good also.

For a bi-amp/tri-amp loudspeaker, these modules would provide a quick, easy, economical and easily serviced soloution.
I believe some modern active cabinets are indeed configured this way.

If you build such an amplifier, please report back your experience.

Regards, Eric.

PS - such an amplifier maybe the basis of another 'Gain Card' ?.
Hi Eric and thanks for the detailed reply.
I bought a STK 4171 Mark II amp kit. I bought it because I needed an amp to test a pair of bookshelf loudspeakers I built (my other audio system is too high powered for the new speakers and I am taking no chances). It is very cheap too. Of course, I do not think it is for serious hi-fi.
I am building the 60 W (Project 3A) amp by Rod Elliott.
The STK kit is almost done and I shall be more than glad to report the results to you.

It is interesting that you should make this post, as I was about to start a thread on these hybrid op amps.

A few months ago, I made a pcb for the STK4048XI. [As an aside, this device has full complementary outputs; the 'V' versions, as opposed to the 'XI' versions, are quasi-complementary. A smaller module, the STK4044XI, has 4 output devices and 15 pins; the STK4048XI, 6 output devices and 18 pins. You can ignore the last 3 pins on the '48 if you put a '44 into the same circuit, so design for the 18 pin versions and you can go with either unit.]

Anyhow, I had so many problems that I became frustrated and put them away: horrible oscillation that I could not stop despite trying all sorts of things with the grounding, etc. However, after a friend of mine fried one of my amps, I decided to attempt making a couple of STK modules to put into the damaged unit. This weekend, I finally figured out the problems, one of which is hinted at on the application docs by Sanyo, and another which I accidentally discovered and yet seems very important.

The main secret in stopping oscillation is to keep the gain at 40dB. I don't like amps with this much gain as the input sensitivity is around 0.4 V for full power. This kept me from getting rid of the oscillation as I had initially changed the input resistors from 1K/50K to 1K/20K, and never looked back.

Once you get the gain up, to the suggested 40 dB range, almost all oscillation stops. The amp then looks very good on the scope, except for a little bit of something on the sine waves. (BTW, I used all of the parts suggested on the app docs, various 100pF caps, etc to avoid oscillation as Sanyo warns.) This where the application docs stop. What I found was that there was still oscillation, but at VERY low levels, like 1 mW output into 8 ohms. [And I'm now wondering if this low level oscillation is what has given these guys a back name, particularly with respect to longevity. Any amp that oscillates will overheat and eventually burn-out the output devices.]

At these low levels, the oscillation took off again. I tried all sorts of stuff to fix it: bypassing the feedback resistor with small caps like 10 to 200 pF, changing the input blocking caps, but nothing helped--until the other evening. I was looking at th scope ready to go to bed when I started holding a 200 pF cap across the feedback resistor when the oscillation completely disappeared. But I'd made a mistake, it turns out I'd actually put this capacitor from the amp output side of the feedback resistor to pin 3 going to ground.

In other words, by placing a 200 pF cap from the amp output to ground (but at low current level location; I didn't use the output terminals), the remaining oscillation stopped. The amp then looked absolutely perfect and all distortion dropped 10 dB, especially at low levels. The heatsinks were now cold when idling; before they were warm but this was due to overlooked oscillation.

This STK4048XI produces, with a +/-55 V supply that sags to 50V at full output, 140 W into 8 ohms and 220 W into 4 ohms. Freq response is -0.4 dB at 10 Hz to -2 dB at 100 kHz (limits of my equipment). Distortion (THD) is 0.8mW is 0.08% at 1kHz, 0.025% at 1W, 0.007% at 10W. At 20 kHz and 10W it is 0.035% and only 0.1% at 10W at 50 kHz.

As for sound, it initially sounds very good, not as open or airy as my Rowland amp, but very good. The input needs adjustment: my initially listening was done with an input transformer, which I think may be a mistake on this amp (I was still trying to cut that incredible gain...). Once I get it going, and properly set up, I'll compare it to a Tripath 104, a Rowland, and a homebrew Aleph.
Thanks rljones,

for the words of your experience so far.
In my experience with domestic gear using these is that they are completely reliable, except for when loaded with too low impedence.
Typical amps strangle them with underrated power supplies - perhaps this is part or the trick to get them to run successfully.

And I'm interested to learn of your final results.

Regards, Eric.
Additional Thoughts

Hello rljones again.
my initially listening was done with an input transformer,
Are you loading the secondary correctly to avoid peaking response ?.
Shunt load and build out resistors are required IMO.
I have had bad experience with a transformer input coupled amp until I did this.
Maybe this helps you, maybe not.

Regards, Eric.
These amp modules work very good. I made once a 5 channel surround system with these. Oscillations can also start from bad grounding because of the high gain. Keep input ground, output ground and feedback ground in a star formation to avoid any loops or just keep your gain low.
The recent STK modules do "function" well and are relatively easy to use, once you get the hang of them. However, I've never been a big fan of the resulting amps. I've suspected the lack of degeneration resistors in the diff stage is a big contributor to my displeasure with the sound.

By the way, here is a nice link to a bunch of data sheets for the STK modules as well as other discrete transistors from Sanyo.

I wanna' be like Harry when I grow up :)


Hello rljones, promitheus and mlloyd1,

Thanks for the good info,advice and links.

If you are really good you can crack/punch through near to the edge, and with flat long nose pliers lift up and remove the thin 'lid' of the module
without disturbing the circuit.

I suggest you cadge a couple of blown ones from an audio repair center (the guys will throw them at you !) to practice this.
With the hybrid circuit exposed you are then free to do just about any mods you want.
Indeed I have had to repair some now obsolete ones this way.

One note - the ceramic/aluminium substrate SUCKS heat like crazy, so you need a hot fine tip varitemp iron.
If you pre-heat the base plate with a hair dryer, and then place it on cardboard, you can rework them effectively.
I wanna' be like Harry when I grow up
Hey, maybe that can inspire Hoary to give an opinion on how he might resonably modify say, an STK4048XI.
Hey Harry, ..... come on down !. :D

Maybe this module, or another model could be the basis of a new
'Gain Card' setup - cheap and easy and modular.
Maybe smallish external power supply boxes and smallish monoblock amplifier modules could be the basis of pretty neat reasonably high power surround or bi-amp/tri-amp systems.

You can expand according to budget.
For AUS$25.00 or so per STK presents good value for money IMO,
and pretty easy to implement.
Last night I was inspired to re-work the two modules. What I found is that the gain can be lowered to around 28 dB (619 ohm/15 kOhm) as long as that 200 pF cap described above is in place. Noise is now around 70 microvolts; DC offset is 1.5 mV and the remainder of the specs are the same, except the freq response is slightly down at -3 dB at 90 kHz and -1 dB at 10 Hz. One of the two modules did start to oscillate, but only at full power into 4 or 8 ohms.

I'm letting the amp 'burn-in' on the test bench for another 24 hours, and I'll do some listening tests this weekend. I'll post my findings soon after.

As for the comments of star grounding; it was done. I made 2 pcb myself and a close friend of mine, who is VERY experienced in the industry, having his own high end audio company after being an aerospace engineer, contributed his own pcb (Julius Siksnius of Audire, Inc). All oscillated as I described in the earlier post at lowered gain or at lower output levels. (Julius thougth his board clean, but had checked it out from 250 mW to full output; as I said, the amp oscillates at much lower levels, which is probably where it is used most often playing musical signals at modest volumes, not sine waves at 10 to 100 watts.)

I did take one destroyed module apart and it does look as though it could be modified--although I think it's beyond me...

regards, Robert
hi robert,

it sounds like what you might actually need is a Zobel network on the output of the amp. have you tried that? just put a .5uH inductor in series with the output, bypassed by a 10 ohm resistor. while you may not like having to put an extra component in series with the signal, i've heard it actually works wonders on sound quality even with amps that don't oscillate due to the isolation from load capacitance it provides... makes the speaker more managable for the amp. but if the amp is oscillating with no load, then another option is an RC network at the output, also a good fix for most any amp... instead of using a 220pF cap to shunt the output use a 10 ohm resistor in series with a .1uF film cap to ground. my speaker designer actually supplies external RC networks like this with his speakers (i have a Merlin TSM) and they do clean up the sound quality of many amps, presumably by providing a stable HF impedance and maybe shunting ultrasonics to ground. i'd be surpriced if STK doesn't show either of these options in their app notes (the National LM3886 does specify an optional Zobel network) but maybe STK's documentation is rather poor.


thanks for the suggestion, but I already use both a Zobel RC filter (aka Boucherot cell) and a series LR filter. As mentioned in the first post, I used all items suggested by Sanyo on their app note. To be specific, this includes 0.22 ohm resistors (non-inductive) in series with each output pin, a zobel RC filter (0.1 microfarad in series with 4.7 ohms, both to ground) and the LR filter (4.7 ohms in parallel with 3 microhenry inductor, both in series with the load). These values are those suggested on their app notes.

The 200 pF is in addition to the above, and for whatever reason, seems absolutely essential. I would encourage someone with functioning STK modules and a test bench, to try to verify these findings by adding a 200 pF cap as described. (I did not play with the amount of capacitance; perhaps some other value is optimum, but I cannot fathom how: the freq response is as expected and distortion is below specified.)

In any event, on the test bench, I have a simple 8 ohm load (capable of withstanding 1 kW out output and with a switch convertable to 4 ohms at 2 kW) without any inductance, except for whatever is in the cables. Thus, the test load is rather benign.
oh whoops, sorry, did not know you tried those already. i guess the STK app notes were not as bad as i thought either. :p
hmm that is really odd, i have not heard of an amp requiring additional capacitive load at the output to stabilize an oscillation. and a small one too - strange. i wonder if this is just compensationg for some other problem in the circuit?
Hi Eric,
I completed the STK 4171 stereo module and as you said, it sounds pretty good.
But there is one problem. There is a hum if I switch on without a source connected. The hum disappeared when I connected my CD player. Is this a grounding problem?
The other thing is that the output is very high. Can I connect 10K pots at the input to be used as a volume control.
I finished the burn-in and decided to run a few tests then listen. I looked around and couldn't find any slew readings, so I measured this using a 10 kHz square wave just below full output. The results are about 20 V/microsecond (10-90% read-out). This again is using 619 ohm/15 kOhm resistors instead of the original 1k/56k values. Noise is now 65 microvolts; and after adjusting the power supply, the amp now produces 150 W into 8 ohms (35 V output), 270 W into 4 ohms (33 V), and 300 W into 2 ohms (24.5 V). I believe the 2 ohm results are only limited by my power supply. These modules are dynamic! The 35 V output means at 8 ohms that the unweighed S/N is -125 dB (amazingly good, look at various specs; it is rare to find production amps so quiet). Freq response again is -3 dB at 90 kHz and -1 at 10 Hz.

The power supply is simple: single bridge rectifier followed by 0.1 ohm in each rail going to one 12,000 microfarad cap (with bypass 0.1 microfarad and 4.5 kOhm) per rail, in-line 7 A fuse per rail and then onto each channel's STK module, supplying +/- 55 VDC, which sags now to +/- 52 VDC. Thus, this is not a dual power supply; a potential point of improvement.

Initial listening was done with several sources, primarily SACD (esp. the Mark Levinson sampler with his double bass and Kim's spoken poem--forget the title). I removed the x/o and subs from my system and ran the Quad 988s full range. I only compared the STK modules to a Rowland 10 amp (150 W at 8 ohms, and which has a separate PS). I run a fully balanced system, so while my measurements are all single ended, I listened in balanced mode. I send the (-) signal into the (-) input on the STK module rather than have it grounded. (I can later elaborate on this in more detail if need be.) The STK did not have an input cap in the (+) input (unlike the app note), but did have a 50 microfarad cap in the (-) input side (like the app note).

My first impressions were that the STK had better depth, but a slightly harder high end and a less controlled and detailed bass. To attempt to improve this, the coupling cap that blocks DC in the feedback loop was removed (the 619 ohm resistor goes through this cap to ground in single ended mode). The DC offset with load was still essentially zero, but the amp is now fully DC coupled amp and one must be careful to use a preamp with no DC offset. The amp then measured flat at 10 Hz (limit of my test equipment).

Once this mod was done, the amp became phenomenal. The bass is now the equal of the Rowland 10 and the highs are no longer slightly harsh, but as smooth and rounded as the Rowland. Both my wife and I felt the depth was better with the STK modules, with the voice projecting in front of the speakers and the recesses being so slightly deeper. The air that originally was missing when I had greater gain and the input transformer was now present.

This STK4048XI module is a true gem, once slightly refined with lower gain. Lower gain also means greater bandwidth (and I suppose greater slewing). But it must have that 200 pF cap. (BTW, the slight residual oscillation I mentioned that remained on one channel, is now completely gone: I used a better quality 12 microfarad/63 cap on the PS rails entering the modules, so each is now better de-coupled.)

I look forward to hearing the results of others.

Regards, Robert