I tried the eBay MX50 DIY Kit, Class AB 100W+100W; Worked but problem

So I quick needed a 100W or more amp for a repair project. I ordered a set of 2ea. MX50's from China. This was my Plan B in case Plan A didn't pan out. Here's the link to the ebay listing:

2psc MX50 Class AB 100W+100W power Amplifier kit - eBay (item 200563911019 end time Jan-16-11 21:50:27 PST)

Plan A was to build the Cheap150 originally described by John Fisher here:

http://www.diyaudio.com/forums/solid-state/103632-cheap-100-150-watt-amp.html

Plan A worked great and my comments on the amp performance are on Page 9 of the above thread.

It's a good thing Plan A worked, because Plan B is suspect. The quality of the kit I received is very high. The PCB is one of the best boards I've worked with. I work on many instrument amps, and the OEM PCBs from the major brands I've worked on seem to be significantly worse than these Chinese board. The kit was missing one heatsink screw. All of the other parts were there. I've got plenty of these metric screws in my shop so that was not a problem. All the electronic parts seemed to be of high quality. The semiconductors are not US or Japan part numbers so they're probably not counterfeits of anything. As for the component specs, I haven't a clue as I haven't even tried to find datasheets for any of the parts.

The kits were very easy to assemble and the solder flowed very nicely on the two-layer PCBs. Here are a couple of pics of the finished board (1 of 2):

IMG_0992.JPG

IMG_0994.JPG

The heatsinks don't come with the kit. I used some angle steel to test. Also, no instructions, no schematic, no list of parts (BOM), no adjustment instructions, and no documentation of any sort came with the kit. All you get are the boards and the parts, and that's it. You have to work from the silk screen. If you are new to building electronic kits, this might be a challenge for you.

I powered the boards up with my trusty +/-18V supply. Both boards powered up nicely and seemed to work just fine. Here is a picture of my test rig:

IMG_0988.JPG

MX50 002.jpg

As you can see from the above instruments, I was running a 204Hz signal into the amp, the load was 4-Ohm resistive, and the output looks fine with a 6.35V signal out. That's about 10 Watts of power. I quickly found that the bias trim resistor had absolutely no affect on any voltage in the circuit. It did not change the offset voltage, it did not alter the voltage across the emitter resistors, it did nothing. But I'm not sure anything needed to be adjusted. Everything looked pretty good. The emitter resistors ran cool and there was no crossover distortion.

Here's what happened when I turned up the signal to 14 Watts:

MX50 004.JPG

I start seeing some fuzz forming on the waveform, and immediately I suspect parasitic oscillation. So I drive the amp harder to just before clipping at 18W:

MX50 003.JPG

More of the same. Starting to look a little funky now. It's 21W at clipping. About what I'd expect from my antique power supply. You can clearly see the noise is still there:

MX50 005.JPG

And finally when driven into saturation, the noise disappears:

MX50 006.JPG

So I cranked up the sweep rate to 0.5us/div and adjusted the trigger just so... and here's what I saw:

MX50 001.JPG

So that looks like just under 1us per cycle making the parasitic oscillation just over 1MHz.

Both boards do this exact same thing. They also behave the same way when driving an 8-Ohm speaker load. The Cheap150 didn't have this problem. Funny thing is the MX50 incorporates a 4.7-Ohm/0.22uF snubber circuit (Boucherot / Zobel is the nouveau term).

So I haven't had a chance to find the cause yet, but does anyone have any ideas why these amps behave this way? Here is the schematic.

MX50 Schematic-2.jpg
 
Last edited:
PROBLEM SOLVED

As the schematic shows, the input stage supply is nicely isolated from the rails with an RC network. But notice there are no caps on the rails themselves. So I added a 100uF filter and a 0.1uF bypass cap to each rail. The board isn't quite as pretty as it was but at least it works...

IMG_0999.jpg

Here's the ouptut at clipping. It's around 23W and nice and clean. There was no parasitic oscillation from zero signal to clipping. Looks like lack of caps was the problem.

MX50 006.JPG

Moral of the story... don't underestimate the need for on-board filter and bypass capacitors on the rails.

BTW, soldering and unsoldering was very easy and the boards held up very well There was no trace separation or board burn. It's a very high quality board.

I don't have a +/-50V supply handy right now, but as soon as I get one, I'll post some more MX50 performance info.
 
I'm contemplating a summer project of building a 5 channel amp using these. Have sort of made my plans and am going to try putting together a schematic together tonight.

Buckeye, can you tell me how hot the output semi's got when you were testing? I'm trying to establish how much heat sinkadge I'm going to have to build into the amp.
 
Unfortunately, I don't have any hard data about temperature for the output transistors. I tested the amp with the transistors bolted to a piece of angle steel. I have no specs for the thermal resistance. What I know for sure is that what I used was totally inadequate. The transistors got HOT!

You probably already know, the heat sink requirement will be a function of the power supply you're using and the average power delivered to the load, and load impedance.

Looking at the 2SD1047 transistor specs, http://www.audiolabga.com/pdf/2SB817.pdf, says you can take the case all the way to 120 deg. C. I wouldn't. I'd say no more than 100. The total temperature drop to ambient is 100 - 25 = 75. Divide 75 by the number of watts dissipated by each transistor. Case to mica (with thermal compound) to heat sink is about 1 deg/W. So subtract 1 from the number above and that's what you'll need for each transistor. If you use one big heatsink, you'll need a fourth of that thermal resistance.

Notice that you don't have to worry about junction to case thermal resistance because the transistor specs are given at case temperature.

Hope that helps. Here's a good reasource for heat sink design: ESP - Heatsink design and transistor mounting

About the other question... The amp sounded very good. I tested it for sound reinforcement with less than audiophile=quality speakers. It was clean and low noise.
 
Buckeye....

you might not like the tone but please read the post ...

anyone the realizes that bias pot has no effect over the bias or the emmiter resistors and keep the amp playing makes a mistake

obviously either there is a problem caused by you or the designer arround the vbe multiplier area or simply you are not measuring a few millivolts and look for much higher voltage in the wrong resistor

you choise about amplifiers is dramatically poor a quasi complementary amplifier with 4 output transitors and each and everyone of them has its own driver will be a nightmare to stabilize if not designed properly

as about the choise of the cheap 150W amplifier all your results are may correct but not enough ....

your testing procedures are not even close to real life conditions and its expected your cheap 150W to only survive just a few minutes before blowing an expensive speaker ..

i will not bother to explain why once more ( Jhon fisher actually is a guy that i like very much ,exchange emails with ... he is a very cheerfull guy , quitar player and music composer ...he has not much to do with electronics and this a circuit given from a friend to a friend as a quick crappy solution . ) ..... obviously if the amplifier is not allready returned to u in a very smoky enclosure ....dont worrie ....it will any minute now ....


the idea is quickly get back the cheap 150w amplifier ....replace it with a working circuit

then since you dont like me any more ( cause of my tone ) ask other forum members to advice you on bias setups and real life resistive /inductive/capacitive tests on amplifiers


kind regards sakis
 
......... notice there are no caps on the rails themselves. So I added a 100uF filter and a 0.1uF bypass cap to each rail. .................... There was no parasitic oscillation from zero signal to clipping. Looks like lack of caps was the problem..............don't underestimate the need for on-board filter and bypass capacitors on the rails.
called decoupling.

Absolutely mandatory to meet the fast changing current demands of devices on the PCB.

It might be even better to locate the decoupling caps right across the devices that have this fast changing current demand. 4 output devices = 4 decoupling caps.
 
The output transistor quiescent current is not monitored.
The bias knob is only having a direct effect upon drivers.
Who knows how hot the outputs will get???
0.22R in the collector circuit is useless.

How to fix:
Cut path from upper driver emitters and lower driver diode par 100R
direct to the load, and rewire these to the output collectors instead.

Now sum of each driver + output current is watched over by 0.22R
Your bias knob will now work properly, and maybe you can
tweak away the parasitic thing....
 
Last edited: