What is so disappointing about this kit, besides the obvious errors in the schematic and that the schematic they suggested it was is not what was delivered, is they removed the darlington VAS. That increases the distortion by about 100 times. I expected not to have matched transistors, high offset and other issues. I only got it to play with on the bench and it looked easier to drill a few more holes than have my own boards build. After playing with it in Spice and my learning, I may not bother with it.
You say it does not need bias adjustment. What is idle current as you have it set?
I was looking at MJL1302/MJL3281 as outputs.
You say it does not need bias adjustment. What is idle current as you have it set?
I was looking at MJL1302/MJL3281 as outputs.
That does tell you something about some of the $1000 amps on the market!
Worth every penny, absolutely if you already have a old chassis and transformer. Most any throw-away mid size receiver will do, or numerous e-bay broken amps.
One more 35 cent transistor in the VAS would have made it a lot better. I would have tweaked a few resistor values. No output network is asking for serious trouble. They should have suggested an "outboard" solution.
Worth every penny, absolutely if you already have a old chassis and transformer. Most any throw-away mid size receiver will do, or numerous e-bay broken amps.
One more 35 cent transistor in the VAS would have made it a lot better. I would have tweaked a few resistor values. No output network is asking for serious trouble. They should have suggested an "outboard" solution.
Being still unfamiliar with some of the abbreviations used, could you please point out in the schematic or explain otherwise to me what you mean by VAS?
Should I look to Q7 or any other transistor?
Hi baskiria: I agree with your opinion, I use many amplifiers MX50 SE in a multi amplified in mid and high frequencies(MTM system) and its result is very good. Sounds realistic, fast anyway very nice.
Last edited:
"Stereo MX50 power amp DIY kit X 1 (include 2 channel PCB and components)."
$17 is for two channels delivered.
http://www.ebay.com/sch/i.html?_sacat=0&_from=R40&_nkw=MX50+kit+amplifier&_sop=15
$17 is for two channels delivered.
http://www.ebay.com/sch/i.html?_sacat=0&_from=R40&_nkw=MX50+kit+amplifier&_sop=15
Last edited:
Turn PCB-s 90º, no need for cooper wires.
Dimension of PCB is 76.5mm x 73.4mm, it fits good.
Can you please help us and share the info you discovered? Also I'd be interested in what you'd change, you mentioned resistors and adding a capacitor somewhere in VAS.
This is just from my beginners experience. I am sure an experienced pro ( of whom many are on the forum) would give far more instructions. I want to build this one as well as I can, without drastically changing it or spending much money. I have another project to build one my way.
Put it in SPICE. That should always be step one. You may have some difficulty finding some of the transistor models, but you can find pretty close ones. Here is where the learning starts.
Gain is a bit high. If you plan on it being 50W into 8 Ohms, you can increase the feedback to 512 Ohms. With single outputs, I might even reduce it further. Your choice as this is DIY.
I am a fan of TMC so I would add a 1K branch and a pair of 100p. This may allow deletion of the shunt across the input.
The 47 Ohm stoppers on the drivers seem a bit high unless you find you have issues with parasitic oscillation.
Play with your bias. I found about 100mA to be a good starting point as I don't trust the amount of servo temp comp until I test it on the bench.
I have not done a thermal analysis. I suspect the VAS current to be a bit low once warmed up.
It needs a standard DC, temp, mute add on. The circuit/card from ESP would do just fine.
I would add flying diodes across the bias servo to prevent excessive current into low loads. Protection is actually more important in small amps as they are more likely to be stressed. I doubt I have ever clipped my 250W Parasound, but I did occasionally clip my 60W Rotel. ( It was really LOUD)
I would change the input electrolytic to as large as fits on the board. 10u minimum.
When you simulate, you find you run out of current to the drivers clipping at about 16v into 2 Ohms. 4 Ohms is marginal. I think I would call this as a nice 25W amp with good overhead than to call it a 50W.
It is a design choice they used a CFP output. You will have to live with the crossover issues, but gain the output stage feedback. Some find this to be a very positive trade-off. Amps of this topology have been well received over the more typical EF.
Not having the output coil is just plain dangerous as a great many speakers are very nasty loads. If you want to omit it, you need to test on the speakers you will be using with the cables you will be using.
I test in spice at 8 ( spec reference) 4 , 2 Ohms, and with the Cordell simple speaker model. Test at 20K reference full power. Test at just a bit over clipping. Use the AC analysis to find the zero gain point and test at that frequency, looking across the IPS for phase margin. ( Or if you can get a TIAN probe to work. I have not yet. ) Test with square waves. 1ns rise time, 1V , just under clipping. Bounce around looking at the voltages and currents everywhere. You will see the crossover glitch issues. Glitches you have to live with, watch for oscillations. Set the temps at 40 to see how much drop in IPS and VAS current. Test output temps 25 to 80 and watch the bias current. Be sure it does not go into runaway or drop too far.
They give no recommendations on heat sinks. Do the calculations, or go "obviously big enough" When on the bench, playing half power into 8 Ohms, they should be no more than warm.
As the transistors are not matched, watch the output DC offset. You may want to put in a little circuit across the emitters of the current mirror to null the offset. This does wonders for distortion too. ( A 10K pot seems to work pretty well. Center tied to the negative rail. ) I don't want more than a couple of mV DC. Some say 50.
Build it and test just like you did in SPICE. Spice lies, so you need to test in the real world. You can bet the kit manufacturer never did either.
I have a thing about power and grounds. I star all of them. So, I would lift the 68 Ohm resistors and supply them directly from the stub off the main bank where you take power to the output stage. I fly all grounds back to the main bank, not sharing any traces on the board. I found HEXFRED rectifiers to reduce the noise a bunch.
My model went from .07% @ 20K, 50W into 8 Ohms to .02% That is probably respectable keeping the advantages of the CFP output. I have not done the noise simulation. If the real amp does .05, that would be a real success.
Put all this into context. Analysis and learning is cheap, but when you start changing parts, it costs more than the kit. Sure a couple of 100p caps are 20 cents, but postage is $12. Building the ESP boards will cost a lot more. Hopefully you have a surplus chassis, transformer and heat sinks, or suddenly this gets darn right expensive. You may find a decent output relay in the old chassis too.
On fuses, I fuse each channel rail independently as well as the AC. It is my choice and I don't risk liability to anyone but myself, so I don't use a speaker fuse.
Please, I am not bashing this product. Consider the price! Build it on surplus parts. If your goal is a state of the art high end amp, cost no object, look elsewhere. This little puppy should sound considerably better than the typical receiver if you pay attention to the physical construction. Some big name brands you pay a lot for have less inside.
Put it in SPICE. That should always be step one. You may have some difficulty finding some of the transistor models, but you can find pretty close ones. Here is where the learning starts.
Gain is a bit high. If you plan on it being 50W into 8 Ohms, you can increase the feedback to 512 Ohms. With single outputs, I might even reduce it further. Your choice as this is DIY.
I am a fan of TMC so I would add a 1K branch and a pair of 100p. This may allow deletion of the shunt across the input.
The 47 Ohm stoppers on the drivers seem a bit high unless you find you have issues with parasitic oscillation.
Play with your bias. I found about 100mA to be a good starting point as I don't trust the amount of servo temp comp until I test it on the bench.
I have not done a thermal analysis. I suspect the VAS current to be a bit low once warmed up.
It needs a standard DC, temp, mute add on. The circuit/card from ESP would do just fine.
I would add flying diodes across the bias servo to prevent excessive current into low loads. Protection is actually more important in small amps as they are more likely to be stressed. I doubt I have ever clipped my 250W Parasound, but I did occasionally clip my 60W Rotel. ( It was really LOUD)
I would change the input electrolytic to as large as fits on the board. 10u minimum.
When you simulate, you find you run out of current to the drivers clipping at about 16v into 2 Ohms. 4 Ohms is marginal. I think I would call this as a nice 25W amp with good overhead than to call it a 50W.
It is a design choice they used a CFP output. You will have to live with the crossover issues, but gain the output stage feedback. Some find this to be a very positive trade-off. Amps of this topology have been well received over the more typical EF.
Not having the output coil is just plain dangerous as a great many speakers are very nasty loads. If you want to omit it, you need to test on the speakers you will be using with the cables you will be using.
I test in spice at 8 ( spec reference) 4 , 2 Ohms, and with the Cordell simple speaker model. Test at 20K reference full power. Test at just a bit over clipping. Use the AC analysis to find the zero gain point and test at that frequency, looking across the IPS for phase margin. ( Or if you can get a TIAN probe to work. I have not yet. ) Test with square waves. 1ns rise time, 1V , just under clipping. Bounce around looking at the voltages and currents everywhere. You will see the crossover glitch issues. Glitches you have to live with, watch for oscillations. Set the temps at 40 to see how much drop in IPS and VAS current. Test output temps 25 to 80 and watch the bias current. Be sure it does not go into runaway or drop too far.
They give no recommendations on heat sinks. Do the calculations, or go "obviously big enough" When on the bench, playing half power into 8 Ohms, they should be no more than warm.
As the transistors are not matched, watch the output DC offset. You may want to put in a little circuit across the emitters of the current mirror to null the offset. This does wonders for distortion too. ( A 10K pot seems to work pretty well. Center tied to the negative rail. ) I don't want more than a couple of mV DC. Some say 50.
Build it and test just like you did in SPICE. Spice lies, so you need to test in the real world. You can bet the kit manufacturer never did either.
I have a thing about power and grounds. I star all of them. So, I would lift the 68 Ohm resistors and supply them directly from the stub off the main bank where you take power to the output stage. I fly all grounds back to the main bank, not sharing any traces on the board. I found HEXFRED rectifiers to reduce the noise a bunch.
My model went from .07% @ 20K, 50W into 8 Ohms to .02% That is probably respectable keeping the advantages of the CFP output. I have not done the noise simulation. If the real amp does .05, that would be a real success.
Put all this into context. Analysis and learning is cheap, but when you start changing parts, it costs more than the kit. Sure a couple of 100p caps are 20 cents, but postage is $12. Building the ESP boards will cost a lot more. Hopefully you have a surplus chassis, transformer and heat sinks, or suddenly this gets darn right expensive. You may find a decent output relay in the old chassis too.
On fuses, I fuse each channel rail independently as well as the AC. It is my choice and I don't risk liability to anyone but myself, so I don't use a speaker fuse.
Please, I am not bashing this product. Consider the price! Build it on surplus parts. If your goal is a state of the art high end amp, cost no object, look elsewhere. This little puppy should sound considerably better than the typical receiver if you pay attention to the physical construction. Some big name brands you pay a lot for have less inside.
Been beating this thing to death with all I have learned.
Sim went from .02% to .007% Not too bad for not really changing the design and not hacking up the board and not spending more than it cost in different parts.
Changed to TMC, 100p/100p. Stable at 4 Ohms.
Reduced gain slightly. 1V RMS = 44W into 8 Ohms. No way this was ever a 60W amp. I have a 32V power supply.
Reduced IPS degen to 22.1 Ohms
Increased the IPS shunt to 220P
Set bias way down as per SELF. 15.3 mA Yea read right 15.3 mA. (Page 183)
Added 1u to reference point in IPS ccs
Add an output off-board RL.
Replaced the 68 Ohm resistors in the VAS power line with CRCRC 100u/10 Ohm series which will go right back to the main bank.
Now this is about ready to get built. I think I have enough parts to make up a DC/mute/therm protection from parts in the old integrated amp this is going into. Basically the circuit ESP sells on a nifty little board. ( Buying a few for my other old amps that don't have these "modern" innovations)
Sim went from .02% to .007% Not too bad for not really changing the design and not hacking up the board and not spending more than it cost in different parts.
Changed to TMC, 100p/100p. Stable at 4 Ohms.
Reduced gain slightly. 1V RMS = 44W into 8 Ohms. No way this was ever a 60W amp. I have a 32V power supply.
Reduced IPS degen to 22.1 Ohms
Increased the IPS shunt to 220P
Set bias way down as per SELF. 15.3 mA Yea read right 15.3 mA. (Page 183)
Added 1u to reference point in IPS ccs
Add an output off-board RL.
Replaced the 68 Ohm resistors in the VAS power line with CRCRC 100u/10 Ohm series which will go right back to the main bank.
Now this is about ready to get built. I think I have enough parts to make up a DC/mute/therm protection from parts in the old integrated amp this is going into. Basically the circuit ESP sells on a nifty little board. ( Buying a few for my other old amps that don't have these "modern" innovations)
Reading your last 2 posts, you earn my respect for your work. Great job!
Indeed, for $17,- you can't expect a non-compromise High-End amp. But it has a lot of potential, as you've proved. It is a candidate for a project for my daughter. The 25 Watts is power enough and the reserve it has at that rating makes it A-OK.
I'd like to go the same way you did: put it in Spice (mine is LTspice) and fiddle with the components before building. This step however will give me some headache...
With regard to your changes:
I like your changes in the power-lines and grounding. The RL addition to the output is very useful I think. A lot of your other changes, I have yet to find out what they mean (lack of theoretical knowledge), but perhaps you could help me on that.
Cheers, Edwin
Great explanation tvrgeek, thanks a lot. My only question is, what sonic difference those make?
esgigt: what is the great difference between this amp, and let's say an Audio Analoge Puccini? The AA has an NE5532 driving TIP darlingtons. Nothing fancy or complicated, and it's a well regarded - we can say high end - amp. Or you meant some million dollar exotics by high end?
esgigt: what is the great difference between this amp, and let's say an Audio Analoge Puccini? The AA has an NE5532 driving TIP darlingtons. Nothing fancy or complicated, and it's a well regarded - we can say high end - amp. Or you meant some million dollar exotics by high end?
Never heard of the Audio Analoge Puccini until you mentioned it
Appearently these Audio Analoge's are prone to thermal problems..
What I understand to be an (almost) uncompromizing High End amp: e.g. the ML's, Krells and Pass's of this world. ...
Never heard of the Audio Analoge Puccini until you mentioned it
Appearently these Audio Analoge's are prone to thermal problems..
What I understand to be an (almost) uncompromizing High End amp: e.g. the ML's, Krells and Pass's of this world. ...
That's the SE, I was mentioning the original one. Anyways, it's a well regarded amp, even if you are not familiar with the brand.
On the Pass amps... Is this any much more sophisticated or complicated than what we're discussing in this topic? https://www.passdiy.com/project/amplifiers/son-of-zen
BTW Pass says measure less, listen more, and you guys are working on trimming distortion from 0.02% to 0.007%... That's why I asked. What sonic difference this does make? Does this put the amp in a higher league with easily perceivable difference?
The bottomline of the question is are the so called high end amps much more complicated in their circuitry? As the Pass example shows, not necessarily, am I right? What makes a high-end amp so high-end then? Exotic components? Overkill oversizing of just about everything? Being fully class A? Thick aluminium front panel?
If our guy LJM really reverse engineered a Musical Fidelity X-A50 as this kit is claimed to be a clone of that hence the name MX50, than we have a 500GBP per piece amp for $17 a PAIR, with nice build quality and no fake components, plus a really good starting ground to further experiment...
Not to mention it sounds fantastic out of the box with great dynamics and nice clarity. As I wrote here earlier, in the league of some $1000 amps that I listened to, and that tells all. Oh, those had a nice brand name, a thrilling desgined remote, gold RCAs at the back and a thick aluminium front pane, all very important aspects if you want to OWN something and don't really care about the sound @Baskiria:
Imho the issue is not "all the other amps", but the MX50 amp. I understand that there are 2 versions on the internet. One version with 2 output transistors, which is discussed in this topic.
So maybe the conclusion "... this kit is claimed to be a clone of that hence the name MX50, than we have a 500GBP per piece amp for $17 a PAIR, with nice build quality and no fake components, plus a really good starting ground to further experiment..." is justified, maybe not.. Being a clone of the Musical Fidelity X-A50 as you suggest, might not be a bad case...
Since I have not built the amp yet, I am still in the phase of data-collection and asking questions.
Imho the issue is not "all the other amps", but the MX50 amp. I understand that there are 2 versions on the internet. One version with 2 output transistors, which is discussed in this topic.
So maybe the conclusion "... this kit is claimed to be a clone of that hence the name MX50, than we have a 500GBP per piece amp for $17 a PAIR, with nice build quality and no fake components, plus a really good starting ground to further experiment..." is justified, maybe not.. Being a clone of the Musical Fidelity X-A50 as you suggest, might not be a bad case...
Since I have not built the amp yet, I am still in the phase of data-collection and asking questions.
@Baskiria:
Imho the issue is not "all the other amps", but the MX50 amp. I understand that there are 2 versions on the internet. One version with 2 output transistors, which is discussed in this topic.
So maybe the conclusion "... this kit is claimed to be a clone of that hence the name MX50, than we have a 500GBP per piece amp for $17 a PAIR, with nice build quality and no fake components, plus a really good starting ground to further experiment..." is justified, maybe not.. Being a clone of the Musical Fidelity X-A50 as you suggest, might not be a bad case...
Since I have not built the amp yet, I am still in the phase of data-collection and asking questions.
Please note there are two versions, the MX50 and the MX50X2, these are not identical topology, and not just because of the output transistors, the previous stages are also different.
I encourage you to build it, it's a cheap and nice project