|'); } // End -->|
Well they showed up and got a quick look. First overall impression is quite acceptable. But in the quest to improve a $250 (CAN) power amp there are a lot of things that could be tweaked. A quick peep under the steel lid reveals a center mounted ~600VA toroid, two extremely puny pots, and a pair of cards mounted to the side heat sinks. A pair of 3300 mF/100V caps per card.
Listening in a very tiny control room with a giant pair of old JBL Studio monitor, revealed little.
First the noise floor was excellent,
Gain is a little higher than practical (Bal&Un)
Ice cold in idle and didn't warm up much during the test.
Bridged output ran without a burp and wasn’t strained. You got the impression it wanted to do play but wanted directions.
I will be getting into this more, but I already have some replacement pots on the way. What did they sound like... "veiled" is a good term. They seemed to have a dull sound, nothing distinct could be heard, be at the same time nothing offensive either.
My fear seeing “ultra linear” means too much overall feedback. I didn’t have my nose inside to long enough to draw the schematic in my head. But looked like two pairs of power transistors, a pair of driver transistors. I also noticed three tweak pots.
Anyway I'll update this post as I buffalo these beasts into civilized front row spankers.
I wonder if the rails are anywhere near 100V… I hope not…
What follows (for me) is a break thru in understanding. In an analog amplifier the “valve” permits a charge to form a pulse in the output wire. So if you match the impedance across the valve, maximum power transfer occurs. But more importantly nothing is lost.
The two new ideas here are: The capacitor has to be able to deliver energy at all frequencies equally and: The output wire has to convert all of the energies into a pulse. That’s why a little piece of silver wire makes such a Hugh difference to the signal accuracy. Once a pulse is formed it will tend to keep its shape, with predictable losses based on the total network. But the critical spot is the first inch* of wire off the "valve" which converts the charge into a pulse. The "valve" dose not produce a pulse, it allows a charge to pass into a wire to the load. Therefore the capacitor and the wire, on either side of the "valve" are equally important to the resultant pulse.
The old rule of thumb was a cables characteristic impedance showed up at 1-3ft… I’ve found 6-8 inches of silver wire produced the intended effect.
Other factors include the number of strands, overall gauge, and the tightness of the twist. From my small experience, 22 gauge or better, made up of fine to very fine gauge. I think mixing gauges is a mistake, because it messes up the midrange accuracy, but that is only from my experience. Why not copper. Well it should be possible to do in copper, and certainly some very great sounding amplifiers use copper, and its way more available and way cheaper.
Technically the wire doesn’t convert the energy into a pulse, it gives the electrons a path to flow and thereby produce a field. This continues to the load as a pulse. The important ideal is that the wire welcome all comers, in an orderly fashions, and speeds them on their way.
are op-amps involved as gain stages? - for such low cost sounds like good lil amp - if will 'come alive' - now might it compare to stock servo 260 or old NAD 2100? (I've got stock ep 2500)
First I will say again that what follows is overly critical for an amp in this price range and I didn’t need to take brain killers after hours of intense listening!
So before doing any major MOD, it is good practice to kill off any runts first. This also gives time to evaluate OEM performance. I ran them in their final positions (surround channels) and they sounded different than the retiring thirty-year-old MODed Harmon Kardon 730 receiver Amps.
So with night falling it is time to pull out the B&W 805s. With four sets of short 12 gauge cables BiAmped. Check my math here… running left and right program into channel one of both amps in bridged mode with both channel 2 outputs phase revered. Next I hooked up an external 47K dual gang pot and cranked the internal channel one pots wide open. That’s four speakers connected to 4 amplifiers with minimal cable coloration and effectively bypasses the internal pots.
So with this set up connected directly to one of two different CD players: I will endeavor to ascertain the character of these amps.
The top end is muted and splattered (no air). Flute (fuzzy), Piano(wrong), and Xylophone (chalky), all sounded distorted. Material in general sounds busy, and everything is popping/pumping out slammy and dry over dampened. Makes the attack on drums interesting, at the expense of subtle content which is lost completely.
I will hold off speculating what is causing it…but I’m hope more BIAS will warm up more than the heat sinks.
As I laser hop thru the piles of reference media the first MOD is formulating. I think it should be possible to shaft extent the volume knobs to PCB mounted replacement pots and lose the interconnect cables… Possibly use the OEM pots as bushings for the shaft extensions and just leave the knobs on them as is.
Well I stuck my nose back inside far enough to get a shock…maybe it sounds the way it dose because it’s a cap coupled bridge amp…let’s just add up the clues…no center tap off the toroid (single 52 volt rail), the big caps +/- connect to the positive output post, and there is no signal path for bridge mode (in mono channel 2 runs as a current mirror), four power devices.
This changes my listening test setup, as I will use a Y-Cord to direct drive each amp.
Other tid bits: the input pots are 20K, the three tweak pots are 2K2, 100R, 2K2 (probably BIAS, Balance, BIAS…?), the Toroid has three legs on the line side(half the primary unused in 120V).
Going forward what does this mean. Well cap coupling an 8 ohm load with 6600mF will get you down to ~16Hz which is fine. Getting a flat response into a speaker(dynamic load) is a different story. Stacking springs doesn’t give you good cornering. What to do… well to clean up the top end you put in bypass capacitors. These little guys are faster on their feet and will pass the top end. But what about the midrange? Well if I had an Ongagu Amp, I could just rip out the hand rolled pure silver caps. Or you could just roll your own with aluminum foil and wax paper. Ideally the XL,XC dynamics of your output chain would add up to zero or resistive across the usable bandwidth. Reality is that ain’t going to happen, but you can tweak things to your ear. I’m going to stew about it for a while. I will speculate that adding eight caps per stereo amp should do the trick, or just swapping the four out….
Beyond the vail.
Well some time Tuesday, the sound noticeably cleaned up, I’d say 50% better. Still excessively punchy, but the top and the mids improved; still smeary but more is getting through...
Wednesday: The image is a little more involving but not front row. Having considered the assumption of the topography of this amp; Differential Darlington front end(voltage gain)driver: Complimenty Transistor Split Bridge(current gain) Final, Dual Polar Capacitor Coupled Output.
The choices are: Adjust Bias, Add Bypass Coupling Capacitors, Replace Coupling Capacitors, Remove Coupling Capacitors.
Adjust Bias. Measure (and record) idle (DC) voltages across the four current limit resistors. From this you can calculate the idle power dissipation of the transistors. Typical(wimpy) idle bias current in an AB amp is about 6mA. So if the rail is 52V and the resistor is R22 you get about 1.32mV across the resistor, or with two resistor, and two transistors across 52V you get ~26V x 6mA=150mW of power dissipation per transistor. Let's crank that to 5watts of dissipation per device. 5W/26V=192mA or 42mV across each current limit resistor. This should have four effects. Each heat sink will have to dump 20watts, and will no longer be a tree hugger: Two the ripple on the supply will increase and may cause hum: three crossover distortion if present should be greatly reduced: Four low end should tighten up, and low-level signal distortion should be improved. It is more important to null the DC on the (pre Coupling Capacitors) then to match Bias Currents. DC pre Coupling Capacitor should be nulled after the transistors have warmed up and should be checked hot and cold. All adjustment done, with no signal applied, and output to dummy loaded. If your dummy load can take it you can scope sine wave looking for crossover distortion. If you like you can also look at square wave response with and without by-pass capacitor/s also checking for RF.
Recheck for DC on the outputs and with a headphone across the dummy load audition some music. I think it will be impractical to correct this capacitor coupled amp for “Truth”. But it maybe possible to get “Beauty”. A capacitor-coupled amp is always going to punch out faster than a direct-coupled amp, so the objective is to get the mids and the highs out just as fast and cleaner. The choices are: Swap them out with say Lo ESR caps at 100KHz, or pick some by-pass capacitors to stack on the existing capacitors. Finding the best ones for your ears/speakers will be loads of trial and error. There are a few other variables which already exist on the way out. These come mostly from the output hook up wire, and the series inductor. The purpose of the series inductor is to prevent external RF from getting in the feedback loop, and thus will be left alone. The hook up wire represents a small amount of series resistance, which can be manipulated to conceal the response curves of the output coupling capacitor array. For example each added capacitor could have it’s own current path (match) wire type/gauge to band to be enhanced. Low frequency flows thru wire cross sectional area, and high frequency flows thru wire surface area. Therefore solid wire for low frequencies, and fine stranded for Hi frequencies. Overall gauge effect resistance, therefore level, and Q (width of band). Three really skinny pairs would hide the effects the best, but too skinny and dampening is going to suffer on reactive loads or at high power.
Picking values is going to be the real fun. One approach would be to fix 1Khz and 10Khz and call it a day. From my math worst-case scenario requires the capacitors to survive 4.5amps of ripple. From T=RC your looking for 100uF for 1.25KHz, and 10uf for 12.5KHz. A long scowl down the Digi-Key list failed to excite. Maybe Digi key 565-1428, and 565-1473) at about a buck each this ain’t no wallet busting MOD. You would need four of each per stereo amp with at lease a 100V rating. Higher voltage caps may have higher ripple current ratings but will move the image back (deeper into the sound stage).
Get a JAL cut(less pointy) Philips #2 screw driver and try not to strip the heads on the cheap hardware. You may want to replace the lot of them. With the lid off note the polarity of the 3300/100 capacitors and with a real soldering iron remove one. Solder in three wires in each hole. Now terminate each pair of wires with its matching capacitor.
Do the same for the other 3300/100. Activate your handy universal mounting system. (Cable ties, electrical tape or shrink tubing, RTV Silicone). Just prevent shorting which leads to crying. So you should have three capacitors with say eight inches of wire attached to each leg. All the positive legs tied together at the PCB positive coupling capacitor hole, and the same for the negative legs> negative PCB hole. Choosing a combined gauge which actuately fits in the PCB hole will make it easier… Do the same for the other Coupling capacitors. So you end up with a ball of six capacitors a side connected with twelve wires to four holes on the PCB. I arbitrarily picked eight inches, but you can use what you like. Longer should flatten out the response changes, shorter will enhance them. For than matter they don’t have to be the same length either. You may refine your “mix” by adjusting the length of each pair. You can get as crazy as you like. Copper wire is far more forgiving (resistive) for this type of tuning than pure silver. The trick is in the strand size and count. You could try solid 22 gauge (Digi Key A3051B-100-ND), 7/30 gauge (Digi-key A1855B-100-ND), 19/34 gauge (Digi-Key A185519B-100-ND). But these will set you back over $100 CAN so you might want to just rip some out of something else… Where is that OnGaku...
The other option fits this forum to a tee. It is the simplest, cheapest and possibly the best MOD. It could deliver truth, but without a schematic or a complete understanding of the bias circuit could be devastating! You simply jumper out the each coupling capacitor and see what happens. In theory a classic bridge amp isn’t capacitor coupled. But it is almost a certainly that some; however small amount of DC will be present on the speaker terminals and that DC may increase as the transistors get hot, age, or freak out when confronted with real life stresses. Again the first step is to measure and null the DC on the output (precoupling capacitors) in idle into dummy load. Jumper both coupling capacitors. Then remove the fuse on the PCB and solder a 1/4watt > 1ohm resistor across the fuse holder. With a dummy load and no input roll the dice and turn it on. Measure the DC on the output…or post a picture of your black hole generator! Now with coupling capacitors in series with headphones connected across the dummy load audition some low level music. Let it run for an hour this way. Re measures the output for DC. No DC, remove the fuse resistor and try again. No DC… bring up the signal level let run and re measure. Continue till you run out of amp or dummy load. Re measure for DC. Well now it’s time to commit. A word of caution, 6.5 Amperes of DC will light up most speakers like a light bulb without the bulb. A proper electronic crossbar protection circuit would be adviceable if you like your speakers. Unlike a relay, it consists of solid-state cross point, which shorts out the amp if DC is detected. So it has no delay, no contacts in the signal path to corrode, and no mercy for a misbehaving power amp. A pair of suitable MOSFETs connected in series back to back driven of the protect LED should do the trick if you can get this far. Cranking up the Bias before trying this may result in less DC offset/drift. If any one has the balls to get this far you should have what most would have considered what you thought you where buying to begin with. The bass should come out stiff, the sound stage should larger (you should hear natural decay), and the top should have some air. Not letting the Coupling Capacitors go to waist you could hang them in parallel across the diode bridge. One other consideration, if you direct couple the output. I would highly recommend NEVER bridging the two channels; as this will almost certainly up set the bias and eat some power transistors! Crazy glue the bridge switch in the stereo position and NEVER connect a speaker across the positive output posts!
What am I going to do? Well I think I will let the burn in run another week. Then I will decide if they are fit as is to be surround channels. If not. I will look at the cross over distortion on a scope. If it stinks I will crank the Bias, and try a couple of by-pass capacitors. I am still stewing on the values/type. If the bass continues to be too punchy I will try the jumper bypass. But it is hazardess at best!
I didn’t proof read this so I hope it makes sense. Any of these things requires some past experience in electronics, and are not for the chickens. Risk all to win all or Q&D fudge, it’s up to you. Don’t freak on me if you build a black hole generator!
If someone out there succeeds first let me know.
Done…success! No Doubt Front Row! The 2.2uF caps, and some thin silver wire completed the journey. I would have preferred a coarser gauge of silver wire but it’s all there. Everything I play on it sounds different, even cuts on the same disc have more diversity than I remember. Bad mixes sound bad, good mixes are no longer busy, everything has the space it got in the track. The focus is perfect, and the sound stage extends beyond the speaker width. The bouncy bottom only pokes out ever so slightly. The midrange is rich and the top is very acceptable and extended greatly. I think the reason for this is two fold in respect to the by-pass network. One: there are more paths for the frequencies to escape (less electron competition on the way out) , and two: these added paths improve the negative feed back accuracy.
It was a hard decision to turf the five way binding posts but I’m not going miss screwing around in the back that’s for sure. Whoever decided(50years ago) to use ¼” connectors on high-level signals should have been fired. So I tossed the whole output connector block and opened up the left and right ¼” jack holes for Speak-On connectors. You will have to rebuild the four-resister network for the meter bridge if you toss the connector block.
I wasn’t getting quality sex from the; internal power connectors or speaker connectors so they where chopped off and soldered directly.
I also noticed one of the power transistor mounting screws was loose.
All the added caps where 100V electrolytes from Digi-Key accept the 2.2uF/250V poly.
I doubled up the silver wire which was recycled from the core of Teflon foam coax this gave me separate leads to the four pole Speak-On jacks.
I put the twelve cap prefabbed bypass network across the top two bridge diodes and the output inductor. The 2.2uF caps where connected directly to the 3300uF caps.
Once again I would like to say that this MOD is not for the squeamish. Behringer has put this amp out with half of its potential hiding laden. Which is a boon to finally hear! I’m sure a lot of audiopiles would dig the Sonics.
I’m listening right now and it sounds relaxed, revealing and envoling. I’ll give it a couple weeks and then drag them down to the high-end stereo shop for a professional second opinion.
So that’s it for this thread. The MOD set me back about $50 CAN bucks an amp, and if you can decipher, my notes and have the balls to follow them, you will be rewarded by the results.
Beyond the veil.
Well it was time to get in there and F with things. I built a twelve-capacitor bypass network soldered on to three strips of copper sheet. Consisting of two each 100uf, 68uf, 47uF, 22uf, 10uF. 4.7uF. Removing the original 3300uF caps I traced back the most practical point of connection; which turned out to be off the bridge diodes and the output inductor. After a couple hours (it’s still early) I’d say the change/improvement is the bass is way less punch drunk, and the midrange has more range. But since the added network consisted of electrolytic caps and only targeted the midrange (635Hz, 919Hz, 1K33Hz, 2K84Hz, 6K25Hz, 13K3Hz) I find I’m still missing the top end (air). What appears to have happened is the hollow bass has been fortified (stiffer slam) and the sound stage has become more evolving. But I relies now that I have not actually by passed the coupling caps in the classic sense. So the next Mod will be to put a pair of 1uF directly across each 3300uF. I was hoping to use ceramic caps but they are pricey and I have some poly caps at hand. If they fit between the circuit board and the heat sink I will use a pair of 2.2uF poly caps.
Unfortunately a camera was not available when I had it ripped apart but I did make a few new observations. T9, T10 are Toshiba 2SC5200, T11, T12 are 2SC1943 (power transistors), T7 is a MJE15033, T8 is a MJE15032. Two 4580 dual op amps where also observed near the input. So with this new information I have adjusted my assumption of the circuit. Looks like an op amp front end, Complimenty AB driver, Complimenty Bridge Final cap coupled… So since most of the (voltage) gain is probably accomplished in the op amp, which has good overall specs except for a 5 volt/microsecond slew…most of the sound character is coming from the op amp. This pretty much ends the story, unless you want to start from scratch. Looks like bias thermal tracking uses: a thyristor for driver section, and a diode for the power section.
Another observation that was made: was in an effort to force the top out I swapped out copper speaker cables for pure silver. I was surprised how sensitive the amp was to this change, but in hindsight it make sense with a reactive (cap coupled) amp.
I would add that the “Touchless Speaker Terminals” are “Touchy Speaker Terminals” because I had one fell apart for no particular reason. Speak-On connectors on the way.
It was also observed that some of the 4Mx10M screws where factory stripped, and anyone disassembling the amp should replace them all.
So despite what appears to be an endless slag of this amp I still think they were a good deal, and after a few more Mods, I look forward to enjoying them.
The added electrolyte cap network took some edge off the bouncy bottom and turned it into tighter slam. There was a (very short) moment I had a Krell flashback…
Everything is still rushing out too fast but at least the midrange is keeping up now. The sound stage strangely is as deep as ever… go figure.
So I have:
To chop the Speak-On connectors in,
Replace the volume pots and add drive shafts, chop sticks
Add the 2.2uF poly by pass caps
Might add AC line filter,
Replace internal speaker wire (last bit of tuning)
Maybe crank bias.
I hope someone finds this usefull...
This post is made possible by the generous support of people like you and our sponsors: