Meanwell Modules - AH ML5000 PSU Replacement

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Hi All,
Parts and PCBs have arrived and it's pretty much ready to go, voltages all look good. About to load it and have a look with the scope!

I have been thinking about implementing a protect mode. Say one of the modules fail, It would be great if I could implement a way to switch them all off (cut mains). I have had a few ideas but would love to hear if anyone has any recommendations or have done something similar?

Once again any feedback welcome

(Just Powered up the console!! Everything Looks and Sounds good. Did a very quick comparison against the existing supply. console was always very quiet if anything now it is even quieter. Runs much cooler!

Thanks
Blissy
 

Attachments

  • ML5000 PSU Rebuild 10.jpg
    ML5000 PSU Rebuild 10.jpg
    1.7 MB · Views: 48
UPDATE

Ok I have just discovered something. Using a audio analyzer I did a quick comparison of the noise floor of the two PSUs 
I have discovered my build seems to have a fundamental at 125hz that oscillates in amplitude in contrast to the old supply that does not. I remember that squarewave mentioned the Q of the inductors in an earlier post and a possible bump in the noise floor.

looking back on this thread this seems similar to the issue Ian is having with an increase in noise at 100 cycles
however I am hearing a slow oscillation in amplitude at 125hz

Ok having a look at Squarewaves schematic. It does not show each SMPS module earthed, Currently Each of the modules is connected to chassis earth via screws and therefore AGND. is this where I have gone wrong

(These are at Extremely high beyond usable Gian levels however would be good if I could achieve the same performance of the existing supply)

Thanks
Blissy

 

Attachments

  • Noise Floor.pdf
    237.1 KB · Views: 12
Blissy said:
UPDATE

Ok I have just discovered something. Using a audio analyzer I did a quick comparison of the noise floor of the two PSUs 
I have discovered my build seems to have a fundamental at 125hz that oscillates in amplitude in contrast to the old supply that does not. I remember that squarewave mentioned the Q of the inductors in an earlier post and a possible bump in the noise floor.

looking back on this thread this seems similar to the issue Ian is having with an increase in noise at 100 cycles
however I am hearing a slow oscillation in amplitude at 125hz

Ok having a look at Squarewaves schematic. It does not show each SMPS module earthed, Currently Each of the modules is connected to chassis earth via screws and therefore AGND. is this where I have gone wrong

(These are at Extremely high beyond usable Gian levels however would be good if I could achieve the same performance of the existing supply)
Yeah, that's no good. Definitely have to fix that.

But your build looks very nice. It would not be hard to get something miswired but your build pic shows good construction technique so that is encouraging. I was worried it would be a mess of tangled wires. Not so. Good work.

When you say "SMPS module earthed", I assume you mean the mesh metal chassis of each SMPS? I don't think that would account for mains feed-through which is what you're seeing. Earthing the SMPS chassis' is pretty much just for safety.

But there should be a wire from the ground of the PCB to the chassis bolt. I thought my schematic had that but it doesn't. Maybe it was in my original schem and got lost in the update. Otherwise, how is the PCB ground (which I assume we're calling "AGND") connected to the chassis? It should be. ALL of the grounds converge on the PCB with the one exception of the wire between the earth terminal of the mains plug and the chassis bolt (which I can see in your pic looks correct). Do you measure continuity between AGND and the chassis? But you don't want grounds to ALSO be connected elsewhere (meaning you want branches of a tree with NO loops).

Incidentally is the chassis terminal of each SMPS connected to the chassis of the SMPS and therefore to the chassis of the whole PS through the bolts? Note that you might need some toothed washers to make good contact. Meaning what sort of resistance are you measuring between the chassis term of each SMPS and AGND?

Rather than backtrack step-by-step, I would actually remove one LRS-150-15 module and load test it by itself (no PCB). Specifically, make a load of ~4 ohms 100W somehow and put a scope probe on there, power it up using a power strip for it's switch just long enough to zoom in on the 125Hz / 250Hz, freeze the scope display and power it off. What's the amplitude? You shouldn't be able to see 125Hz at all on the scope really. Your graphic looks like that peak is about 5dB above everything else which you should be able to see on a scope (although your graphic cut off the dB and frequency graticule numbers so correct me if I'm wrong). If it looks "fuzzy", that's high frequency hash (which we don't care about right now) that you can get rid of using the bandwidth limiting feature of your scope. That will help you see the LF. If you can distinguish the 125Hz / 250Hz by load testing one SMPS by itself, that would be discouraging because that would mean that those modules are simply not suitable for this application (maybe like what Ian found). Obviously if you load test the one by itself and don't see 125Hz at all, that would be a completely different story and we can proceed incrementally with trying to diagnose the issue.
 
squarewave said:
Yeah, that's no good. Definitely have to fix that.

But your build looks very nice. It would not be hard to get something miswired but your build pic shows good construction technique so that is encouraging. I was worried it would be a mess of tangled wires. Not so. Good work.

When you say "SMPS module earthed", I assume you mean the mesh metal chassis of each SMPS? I don't think that would account for mains feed-through which is what you're seeing. Earthing the SMPS chassis' is pretty much just for safety.

But there should be a wire from the ground of the PCB to the chassis bolt. I thought my schematic had that but it doesn't. Maybe it was in my original schem and got lost in the update. Otherwise, how is the PCB ground (which I assume we're calling "AGND") connected to the chassis? It should be. ALL of the grounds converge on the PCB with the one exception of the wire between the earth terminal of the mains plug and the chassis bolt (which I can see in your pic looks correct). Do you measure continuity between AGND and the chassis? But you don't want grounds to ALSO be connected elsewhere (meaning you want branches of a tree with NO loops).

Incidentally is the chassis terminal of each SMPS connected to the chassis of the SMPS and therefore to the chassis of the whole PS through the bolts? Note that you might need some toothed washers to make good contact. Meaning what sort of resistance are you measuring between the chassis term of each SMPS and AGND?

Rather than backtrack step-by-step, I would actually remove one LRS-150-15 module and load test it by itself (no PCB). Specifically, make a load of ~4 ohms 100W somehow and put a scope probe on there, power it up using a power strip for it's switch just long enough to zoom in on the 125Hz / 250Hz, freeze the scope display and power it off. What's the amplitude? You shouldn't be able to see 125Hz at all on the scope really. Your graphic looks like that peak is about 5dB above everything else which you should be able to see on a scope (although your graphic cut off the dB and frequency graticule numbers so correct me if I'm wrong). If it looks "fuzzy", that's high frequency hash (which we don't care about right now) that you can get rid of using the bandwidth limiting feature of your scope. That will help you see the LF. If you can distinguish the 125Hz / 250Hz by load testing one SMPS by itself, that would be discouraging because that would mean that those modules are simply not suitable for this application (maybe like what Ian found). Obviously if you load test the one by itself and don't see 125Hz at all, that would be a completely different story and we can proceed incrementally with trying to diagnose the issue.

Thanks Squarewave,

Yes each modules metal chassis is connected to the case through its mounting which is also connected to the Earth terminal block on the Module.

I am measuring about 0.4ohm from AGND to the modules chassis.
As the 12v Supply is also Earthed to the chassis I am getting the same 0.4ohm of the Chassis of the 12v digital Supply to AGND

You are correct I do have a Spade terminal for the PCB to Chassis ground that I left Unpopulated that I will connect to the Star Point. I did this as in the bottom right hand conner of the image of my PSU there is a chassis connection to the connector PCB that was existing.
I shall populate this terminal.

Whats interesting is that the 125hz tone/noise is oscillating in amplitude. It's amplitude is modulated.

Good idea to test the module individually on the scope under load! I shall see what I makeup for a load and post the results!
 
Ok I have just taken out one of the modules tested it on its own under Load 3.5Ohm on the scope here are the results at 200mv per division. Scope Screen shots in drive below

https://drive.google.com/drive/folders/1Dp_4JAiCEoobhb72Us4U056dN83dUNpd?usp=sharing

 
Blissy said:
Yes each modules metal chassis is connected to the case through its mounting which is also connected to the Earth terminal block on the Module.
Ok. That's fine.

Blissy said:
I am measuring about 0.4ohm from AGND to the modules chassis.
As the 12v Supply is also Earthed to the chassis I am getting the same 0.4ohm of the Chassis of the 12v digital Supply to AGND
Ok. So just to be crystal clear, is V- of the 12V supply (aka DGND) still separated from AGND by the 100 ohm resistor? Meaning what sort of resistance are you measuring between AGND and DGND? Should be 100R. If not, figure out why.

Blissy said:
You are correct I do have a Spade terminal for the PCB to Chassis ground that I left Unpopulated that I will connect to the Star Point. I did this as in the bottom right hand conner of the image of my PSU there is a chassis connection to the connector PCB that was existing.
I shall populate this terminal.
Ok but just so that I understand, what path is connecting AGND to the chassis right now? You said it's 0.4 ohms between chassis and AGND but if you didn't connect AGND to chassis explicitly, then how is it connected now?

Also, what does your meter read when you just connect the probes together (maybe rub them together a little first to scrape off some oxidation)? Is it less than 0.4 ohms?

Blissy said:
Whats interesting is that the 125hz tone/noise is oscillating in amplitude. It's amplitude is modulated.
That is kinda interesting. That sounds like inter-modulation distortion. Although with the frequency of harmonics landing on the 2nd and 3rd harmonics of mains, that would be a hell of a coincidence. Otherise, it's mains feed-through.

But ... does the modulation change at all if you adjust the voltage trim on one of the LRS-150-15? Meaning it could be the two SMPS are closely matched but slightly different and so you're seeing "beating" (like a synth with two oscillators slightly detuned). If that is the case, there might be a hack to sync them.

Incidentally, can you post the schem of the original AH supply? I know you posted it in another thread but it would be good to post it here. I don't recall if it had any kind of regulation or capacitance multiplier circuitry or anything like that but if it does, that would be noteworthy.

Also, your analyzer plots were a little crude. It might be worthwhile to compare the noise floor levels. It might be, albeit unlikely, that the new supply with mains harmonics is actually lower than the noise floor of the old supply. That would create a real pickle since it would be technically better and yet still imperfect.

Finally, if the harmonics are really only 5 dB above the noise floor, it should be stated out-loud that you're almost certainly not going to actually hear that anywhere. There's multiple levels of filtering throughout the console. To prove it, terminate one of the mic pres with 500-1K or so, crank the gain 100%, take out any low-cut and look at the noise floor in your analyzer (through the direct-out if pos.). Do you see the mains harmonics? If not, ... eh.
 
Blissy said:
Ok I have just taken out one of the modules tested it on its own under Load 3.5Ohm on the scope here are the results at 200mv per division. Scope Screen shots in drive below
I can't make heads or tails of that. What's time per division? Is the 50nS some sort of trigger threshold? That would make the ringing of the waveform in the MHz. Do another capture with multiple periods of 125Hz. So use like 10 ms / division.
 
squarewave said:
Ok. That's fine.
Ok. So just to be crystal clear, is V- of the 12V supply (aka DGND) still separated from AGND by the 100 ohm resistor? Meaning what sort of resistance are you measuring between AGND and DGND? Should be 100R. If not, figure out why.

Yes Measuring 100R between AGND and DGND

Ok but just so that I understand, what path is connecting AGND to the chassis right now? You said it's 0.4 ohms between chassis and AGND but if you didn't connect AGND to chassis explicitly, then how is it connected now?

Ok so at the Connector PCB (reused from old supply) there is a Chassis Connection from Pin 3 straight to the Chassis which is tied to AGND. Which is why I left the PCB Chassis terminal unpopulated however I can very easily connect it to the chassis and see if it makes any difference.

Also, what does your meter read when you just connect the probes together (maybe rub them together a little first to scrape off some oxidation)? Is it less than 0.4 ohms?

Yep, once I cleaned the probes I measured 0.2 between AGND and Chassis and 0.2 with probs shorted.  So essentially 0

That is kinda interesting. That sounds like inter-modulation distortion. Although with the frequency of harmonics landing on the 2nd and 3rd harmonics of mains, that would be a hell of a coincidence. Otherise, it's mains feed-through.

But ... does the modulation change at all if you adjust the voltage trim on one of the LRS-150-15? Meaning it could be the two SMPS are closely matched but slightly different and so you're seeing "beating" (like a synth with two oscillators slightly detuned). If that is the case, there might be a hack to sync them.

Ahhh thats interesting I shall try it!

Incidentally, can you post the schem of the original AH supply? I know you posted it in another thread but it would be good to post it here. I don't recall if it had any kind of regulation or capacitance multiplier circuitry or anything like that but if it does, that would be noteworthy.

Also, your analyzer plots were a little crude. It might be worthwhile to compare the noise floor levels. It might be, albeit unlikely, that the new supply with mains harmonics is actually lower than the noise floor of the old supply. That would create a real pickle since it would be technically better and yet still imperfect.

Finally, if the harmonics are really only 5 dB above the noise floor, it should be stated out-loud that you're almost certainly not going to actually hear that anywhere. There's multiple levels of filtering throughout the console. To prove it, terminate one of the mic pres with 500-1K or so, crank the gain 100%, take out any low-cut and look at the noise floor in your analyzer (through the direct-out if pos.). Do you see the mains harmonics? If not, ... eh.

Very true !! I did check under crazy circumstances with Monitors Cranked, Master faders at 10 and control room pot at 10 you can very faintly here the modulation within the noise floor.


I have attached original Schematics and updated drive with new measurements not sure how helpful they are. Not seeing any low freq noise I think...

https://drive.google.com/drive/folders/1Dp_4JAiCEoobhb72Us4U056dN83dUNpd?usp=sharing
 

Attachments

  • MPS 14 -2.pdf
    240.6 KB · Views: 8
Blissy said:
Ok so at the Connector PCB (reused from old supply) there is a Chassis Connection from Pin 3 straight to the Chassis which is tied to AGND. Which is why I left the PCB Chassis terminal unpopulated however I can very easily connect it to the chassis and see if it makes any difference.
Ah, the connector PCB is connecting the grounds together. So that's not ideal. That's sort of defeating the whole "star" ground scheme. Like I said before, you don't want grounds connected together in multiple places. It's ok (and actually good) that the chassis is connected there (so you don't need the PCB chassis terminal if you use the one from the connector PCB). But not AGND or any of the other grounds. You want them all to converge on your PCB with the caps.

I'm not sure it will make a difference but since you have two connectors, maybe you might consider desoldering one (braid and flux or hot air station?) and making something that doesn't use that connector PCB so that it's not defeating the whole "star ground" principle.

Although I don't think it will make a difference actually. I'm starting to think the SMPS are just feeding through some mains.

Blissy said:
Very true !! I did check under crazy circumstances with Monitors Cranked, Master faders at 10 and control room pot at 10 you can very faintly here the modulation within the noise floor.
That would still drive me nuts. If I were doing this it would have to work at least as well as the old supply. Otherwise, what's the point?

Blissy said:
I have attached original Schematics and updated drive with new measurements not sure how helpful they are. Not seeing any low freq noise I think...
Mmm, so there is regulation. And current limiting? There's MOSFETs on the +-17 and a 783 on the 48V.

Errr, I'm not sure if you're going to get silence without filtering down to mains frequencies with an active circuit. I'm starting to feel like the larger SMPS just are not designed to eliminate mains feed-through. Even though it is kind impressive that it's only +5dB above the noise floor without any LF filtering. But when you crank everything to the point where you can hear the noise floor, it will stick out.

You could try to just put larger caps on the outputs. Right now you have 100uF. You've got lot's of space and the pitch is probably about the same. If you replace the outer ones with 1000u, does it help? Although one thing to watch out for is that SMPS don't like too much capacitance on the output. I think it's because the extra surge in current will put it into overload. But after a few "hiccups", it might startup ok.

Otherwise, you might get away with just capacitance multipliers. The SMPS regulate well enough. But unfortunately you can't use MOSFETs or the 783 like the original PS because the voltage drops would be too much. I haven't done high power stuff but if I were doing it, I would probably just make another board to go over (or under?) the current one (stacked with stand-offs) and get some 20A SMD (or something that bolts flat) bipolar transistors and make huge copper pours above and below with lots of vias as heat sinks and then do just three basic CMs. They would literally be like 4 parts each (resistor, cap, transistor and protection diode).

Your scope plots are still not nearly zoomed in enough. The noise is going to be more like 2mV and not 200mV. You need to enable bandwidth limiting (down to 1KHz if it can do it). That looks like a fancier scope than your usual Rigol junk so there's almost certainly something to remove the HF jazz. Then you can zoom in another 10x and see the 125Hz.
 
Ok so I just tested Trimming the voltages to see if this made a difference, it did not. I have attached a sample of the noise.

https://drive.google.com/file/d/1LlQFa4wtP0To9KraxjXGMKg_gThakfHb/view?usp=sharing

( If was to pass any audio through at this level not only would I probably have a heart attack, but the drivers of my monitors would be blown across the road!)
 
Ok so my last noise measurements where pretty crude. So this time using a Focusrite interface line level input and SMAART I took 3 measurements.

Direct Output of Preamp (terminated with 1k) at 60db of Gain

Preamp to Mix Bus with +60db of Gain +10 channel fader

And all channels assigned to the Mixbus @ +10  channel faders

I have made it this far with the project, would be great if I could achieve same specs as the original. So here it goes.
So if I where to use CM's one on the +17 -17 and +48? following the CLC filter?
I have uploaded a rough circuit of a capacitor multiplier I drew, am I at all on the right track...?
With 1k and 10uf cap the filter should roll off around 15hz the BJT is;

PHPT60415NY
https://au.element14.com/nexperia/phpt60415nyx/transistor-npn-40v-sot-669/dp/2498525?st=bjt

Any help with component selection would be greatly appreciated not familiar with High Power BJT's, What's common SMD part ? (I have left out the diode in the circuit whoops)

Huge thank you to squarewave! your advise has be extremely helpful!

Noise Measurements and CM circuit in drive below.
https://drive.google.com/drive/folders/1Dp_4JAiCEoobhb72Us4U056dN83dUNpd?usp=sharing
 
Blissy said:
I have made it this far with the project, would be great if I could achieve same specs as the original. So here it goes.
So if I where to use CM's one on the +17 -17 and +48? following the CLC filter?
Yeah ... but there are actually some issues.

First, the simple one transistor CM isn't going to do it. At 5A collector current, that's 5 / hFE which is maybe 50mA if you're lucky and the 1K resistor cannot supply that without a huge voltage loss.

So you need two transistors. There's really no other way AFAIK. Unfortunately that means a two diode drops loss. Meaning if you put in 17V (what are you getting out of your SMPS after trimming up?) you're only going to get at most 15.8V. Probably more like just over 15V. But honestly that's not really a big deal. You're talking about literally a 1dB reduction in headroom. I would not even blink at that. The whole thing will run cooler and your caps will last longer. So ...

The two transistor CM is simple. It uses a Sziklai pairs like this:

  https://www.eevblog.com/forum/blog/eevblog-1116-the-capacitance-multiplier/

I haven't watched that video but I can see from the picture that's the circuit you want.

However, transistor selection is delicate. Power transistor gain tends to drop off at high current. So when you're looking at datasheets you want to go straight away to the hFE vs collector current plot and make sure it's not dropping off at 5A.

I don't love the transistor you picked because it's wattage is a little low (and the package is a little exotic). You have to account for at least 2 volts (let's say 4) across the pass transistor * 5A = 20W but double that for good measure. So a little searching on mouser turns up parts like:

  D44H8/D45H8
  MJB44H11/MJB45H11
  BD909/BD910
  2SC5200
  FJP1943

The ones advertised as for power amplifiers seem to have nice characteristics. Personally I kinda like D44H8/D45H8 because it seems to be a super common part and, perhaps more important, the TO-220 is easy to solder well (very important for high power dissopation parts) to a plated hole. It's got a giant hole in the metal tab for your soldering iron tip. Just make a part with an extra pad (connected to the collector) but with a hole that's the same diameter and location as the hole in the tab. Then put your iron in there with a little solder and lots of flux and after 30 seconds of heating feed a boat load of solder in there and let it wick all around. Or use a bolt like "normal" people.

But again, at hFE 100 your other transistors have to be a little chunky too. For those I recommend something like:

  FZT751QTA/FZT651QTA

Or something in that SOT223 package which is the beefy SMD part. Might as well just keep going with surface mount. When you're picking parts like this, all of the good ones are going to be surface mount. Get used to it. Honestly I'm starting to prefer surface mount. They're really not that hard to solder. I do 0.5mm pitch parts without too much trouble. These are huge by comparison.

Blissy said:
Huge thank you to squarewave! your advise has be extremely helpful!
No worries mate :) I've been pumping up SMPS for years now. I want to find out if I was right. A little mains feed-through at 5A shouldn't really surprise anyone. I think a CM could solve it (albeit at a cost of a few volts).
 
squarewave said:
Yeah ... but there are actually some issues.

First, the simple one transistor CM isn't going to do it. At 5A collector current, that's 5 / hFE which is maybe 50mA if you're lucky and the 1K resistor cannot supply that without a huge voltage loss.

So you need two transistors. There's really no other way AFAIK. Unfortunately that means a two diode drops loss. Meaning if you put in 17V (what are you getting out of your SMPS after trimming up?) you're only going to get at most 15.8V. Probably more like just over 15V. But honestly that's not really a big deal. You're talking about literally a 1dB reduction in headroom. I would not even blink at that. The whole thing will run cooler and your caps will last longer. So ...

The two transistor CM is simple. It uses a Sziklai pairs like this:

  https://www.eevblog.com/forum/blog/eevblog-1116-the-capacitance-multiplier/

I haven't watched that video but I can see from the picture that's the circuit you want.

However, transistor selection is delicate. Power transistor gain tends to drop off at high current. So when you're looking at datasheets you want to go straight away to the hFE vs collector current plot and make sure it's not dropping off at 5A.

I don't love the transistor you picked because it's wattage is a little low (and the package is a little exotic). You have to account for at least 2 volts (let's say 4) across the pass transistor * 5A = 20W but double that for good measure. So a little searching on mouser turns up parts like:

  D44H8/D45H8
  MJB44H11/MJB45H11
  BD909/BD910
  2SC5200
  FJP1943

The ones advertised as for power amplifiers seem to have nice characteristics. Personally I kinda like D44H8/D45H8 because it seems to be a super common part and, perhaps more important, the TO-220 is easy to solder well (very important for high power dissopation parts) to a plated hole. It's got a giant hole in the metal tab for your soldering iron tip. Just make a part with an extra pad (connected to the collector) but with a hole that's the same diameter and location as the hole in the tab. Then put your iron in there with a little solder and lots of flux and after 30 seconds of heating feed a boat load of solder in there and let it wick all around. Or use a bolt like "normal" people.

But again, at hFE 100 your other transistors have to be a little chunky too. For those I recommend something like:

  FZT751QTA/FZT651QTA

Or something in that SOT223 package which is the beefy SMD part. Might as well just keep going with surface mount. When you're picking parts like this, all of the good ones are going to be surface mount. Get used to it. Honestly I'm starting to prefer surface mount. They're really not that hard to solder. I do 0.5mm pitch parts without too much trouble. These are huge by comparison.

Definitely!!

No worries mate :) I've been pumping up SMPS for years now. I want to find out if I was right. A little mains feed-through at 5A shouldn't really surprise anyone. I think a CM could solve it (albeit at a cost of a few volts).

Agreed  :)

Excellent! So with the LRS-15-150 Trimmed all the way I measure 18.5v which means I shouldn't drop to far below +17! I shall Draw up a schematic! and take a closer look at components, Thanks Again Squarewave.
Now I just need to come up with a way of Interconnecting the CM board to the existing board keeping in mind high current...
 
squarewave said:
Yeah ... but there are actually some issues.

First, the simple one transistor CM isn't going to do it. At 5A collector current, that's 5 / hFE which is maybe 50mA if you're lucky and the 1K resistor cannot supply that without a huge voltage loss.

So you need two transistors. There's really no other way AFAIK. Unfortunately that means a two diode drops loss. Meaning if you put in 17V (what are you getting out of your SMPS after trimming up?) you're only going to get at most 15.8V. Probably more like just over 15V. But honestly that's not really a big deal. You're talking about literally a 1dB reduction in headroom. I would not even blink at that. The whole thing will run cooler and your caps will last longer. So ...

The two transistor CM is simple. It uses a Sziklai pairs like this:

  https://www.eevblog.com/forum/blog/eevblog-1116-the-capacitance-multiplier/

I haven't watched that video but I can see from the picture that's the circuit you want.

However, transistor selection is delicate. Power transistor gain tends to drop off at high current. So when you're looking at datasheets you want to go straight away to the hFE vs collector current plot and make sure it's not dropping off at 5A.

I don't love the transistor you picked because it's wattage is a little low (and the package is a little exotic). You have to account for at least 2 volts (let's say 4) across the pass transistor * 5A = 20W but double that for good measure. So a little searching on mouser turns up parts like:

  D44H8/D45H8
  MJB44H11/MJB45H11
  BD909/BD910
  2SC5200
  FJP1943

The ones advertised as for power amplifiers seem to have nice characteristics. Personally I kinda like D44H8/D45H8 because it seems to be a super common part and, perhaps more important, the TO-220 is easy to solder well (very important for high power dissopation parts) to a plated hole. It's got a giant hole in the metal tab for your soldering iron tip. Just make a part with an extra pad (connected to the collector) but with a hole that's the same diameter and location as the hole in the tab. Then put your iron in there with a little solder and lots of flux and after 30 seconds of heating feed a boat load of solder in there and let it wick all around. Or use a bolt like "normal" people.

But again, at hFE 100 your other transistors have to be a little chunky too. For those I recommend something like:

  FZT751QTA/FZT651QTA

Or something in that SOT223 package which is the beefy SMD part. Might as well just keep going with surface mount. When you're picking parts like this, all of the good ones are going to be surface mount. Get used to it. Honestly I'm starting to prefer surface mount. They're really not that hard to solder. I do 0.5mm pitch parts without too much trouble. These are huge by comparison.
No worries mate :) I've been pumping up SMPS for years now. I want to find out if I was right. A little mains feed-through at 5A shouldn't really surprise anyone. I think a CM could solve it (albeit at a cost of a few volts).

I have attached a draft of the CM circuit so far. Do you think it is necessary to add one to the +48 Rail as the current draw is relatively low ?

Cheers
Blissy 
 

Attachments

  • ML5000 SMPS CM [for review] SCHEMATIC 201014.pdf
    15.3 KB · Views: 13
Yeah it's probably ok to leave out the CM for phantom. Phantom is pretty much guaranteed to have separate RC on each mic pre. What is the RC for phantom in the console? I assume you have a schematic? It's probably something like 100R / 47u which should provide decent filtering at 125 Hz and above. If you want to be absolutely certain you could load test it separately to see how much if any mains feed-through you're getting. Or at least hook up a condenser mic and listen for some kind of change in the noise. Have you tested 48V at all? Better check now rather than go through the trouble of making another board only to find out there's actually a problem with the 48V.

Regarding your schem, you have your transistors all messed up. Q2 is actually the pass transistor so thats the high power PNP D45H8. Q1 is the smaller NPN FZT651QTA. Then for the negative side, mirror the circuit like you did but invert PNP with NPN and visa versa. So Q4 is NPN D44H8 and Q3 is PNP FZT751QTA.

Incedentally a very common SMD diode package is SOD-123 (SOD stands for "small outline diode"). Like S1GFL. But it would be equally reasonable to use through-hole reasoning that most people will have lots of axial diodes laying around that would work there.

Also, I would make the cap bigger. The more you shift the corner frequency down, the more attenuation you get at 125 Hz. You could actually make those caps huge and it would just attenuate that much more. It's not going to trigger the SMPS to go into huccup because the 1K limits surge. I think the only caution would be that at some point the rails might be slow to come up which might cause weird behavior. You're already using 100uF on the LC board so that would be fine. I wouldn't go above 1000u.
 
squarewave said:
Yeah it's probably ok to leave out the CM for phantom. Phantom is pretty much guaranteed to have separate RC on each mic pre. What is the RC for phantom in the console? I assume you have a schematic? It's probably something like 100R / 47u which should provide decent filtering at 125 Hz and above. If you want to be absolutely certain you could load test it separately to see how much if any mains feed-through you're getting. Or at least hook up a condenser mic and listen for some kind of change in the noise. Have you tested 48V at all? Better check now rather than go through the trouble of making another board only to find out there's actually a problem with the 48V.

Yes, there is an RC on the phantom of each mic pre. A 1k and a 100nf cap so cut freq is only 1.5k. So I will test test under load and have listen aswell!

Regarding your schem, you have your transistors all messed up. Q2 is actually the pass transistor so thats the high power PNP D45H8. Q1 is the smaller NPN FZT651QTA. Then for the negative side, mirror the circuit like you did but invert PNP with NPN and visa versa. So Q4 is NPN D44H8 and Q3 is PNP FZT751QTA.

Thanks! will update with changes.

Incedentally a very common SMD diode package is SOD-123 (SOD stands for "small outline diode"). Like S1GFL. But it would be equally reasonable to use through-hole reasoning that most people will have lots of axial diodes laying around that would work there.



Also, I would make the cap bigger. The more you shift the corner frequency down, the more attenuation you get at 125 Hz. You could actually make those caps huge and it would just attenuate that much more. It's not going to trigger the SMPS to go into huccup because the 1K limits surge. I think the only caution would be that at some point the rails might be slow to come up which might cause weird behavior. You're already using 100uF on the LC board so that would be fine. I wouldn't go above 1000u.

Will do, Thanks

I am thinking the best option to integrate this PCB with the CLC filter PCB would be to mount it directly using Female PCB spade terminals on top of the existing. Shall Test 48v and then attempt a board layout!

Cheers
Blissy 
 
squarewave said:
Yeah it's probably ok to leave out the CM for phantom. Phantom is pretty much guaranteed to have separate RC on each mic pre. What is the RC for phantom in the console? I assume you have a schematic? It's probably something like 100R / 47u which should provide decent filtering at 125 Hz and above. If you want to be absolutely certain you could load test it separately to see how much if any mains feed-through you're getting. Or at least hook up a condenser mic and listen for some kind of change in the noise. Have you tested 48V at all? Better check now rather than go through the trouble of making another board only to find out there's actually a problem with the 48V.

Regarding your schem, you have your transistors all messed up. Q2 is actually the pass transistor so thats the high power PNP D45H8. Q1 is the smaller NPN FZT651QTA. Then for the negative side, mirror the circuit like you did but invert PNP with NPN and visa versa. So Q4 is NPN D44H8 and Q3 is PNP FZT751QTA.

Incedentally a very common SMD diode package is SOD-123 (SOD stands for "small outline diode"). Like S1GFL. But it would be equally reasonable to use through-hole reasoning that most people will have lots of axial diodes laying around that would work there.

Also, I would make the cap bigger. The more you shift the corner frequency down, the more attenuation you get at 125 Hz. You could actually make those caps huge and it would just attenuate that much more. It's not going to trigger the SMPS to go into huccup because the 1K limits surge. I think the only caution would be that at some point the rails might be slow to come up which might cause weird behavior. You're already using 100uF on the LC board so that would be fine. I wouldn't go above 1000u.


Been caught up in work things, I'm back  :D
I have fixed up the schematic and have done a rough layout of the new CM PCB.
Looking at the datasheet for the D44H8 I am not sure if the Tab is connected to pin 2.

I am a little concerned about the thermals. The large heat skink pad will be top and bottom layer via stitched 25mm x 25mm. Let me know what you think ?

[haven't added GND plane yet]

https://drive.google.com/drive/folders/1T99iWEIkpcDTJIBi-PSlBwC9T0iKvNX2?usp=sharing

Thanks
Blissy     
 
The tabs are almost always connected to the collector. It says in the datasheet figure 1 that it is.

You still have your Q4 transistor reversed. Swap pins 2 and 3. Think about which way the current is running.

Make the WHOLE board heat sink. Make the board as big as the other board (you're stacking somehow right?) and then literally just make four ground pours (actually collector pours) from the entire board with above and below stitched together with vias and split right down the middle with one side for positive and one side for negative. And make sure that you put the smaller base-driver transistor on the same side and over it's connector pour so that it's thermally coupled. There might be a trick to get the pours to dodge the other parts but I'm not familiar with the tool chain you're using.

It doesn't say in that particular datasheet but if you look at similar SMD power transistor datasheets you'll find that to stay on the SOA of the part, you need a certain amount of area of 2 oz copper that you almost certainly don't even have enough space for but the parts are conservative and it's actually good that that board is separate from the other because it's sort of like a heat sink so I think you'll be ok.
 
squarewave said:
The tabs are almost always connected to the collector. It says in the datasheet figure 1 that it is.

You still have your Q4 transistor reversed. Swap pins 2 and 3. Think about which way the current is running.

Make the WHOLE board heat sink. Make the board as big as the other board (you're stacking somehow right?) and then literally just make four ground pours (actually collector pours) from the entire board with above and below stitched together with vias and split right down the middle with one side for positive and one side for negative. And make sure that you put the smaller base-driver transistor on the same side and over it's connector pour so that it's thermally coupled. There might be a trick to get the pours to dodge the other parts but I'm not familiar with the tool chain you're using.

It doesn't say in that particular datasheet but if you look at similar SMD power transistor datasheets you'll find that to stay on the SOA of the part, you need a certain amount of area of 2 oz copper that you almost certainly don't even have enough space for but the parts are conservative and it's actually good that that board is separate from the other because it's sort of like a heat sink so I think you'll be ok.

Thanks Squarewave! Updates made. I have Spit the board in 4 planes top and bottom with a thick GND down the centre. The connectors are placed to line up with the connectors below. The PCB is not as wide as the original PCB to allow for the input Spade connectors.

Have a little bit of cleaning up to do, any feed back greatly appreciated.

https://drive.google.com/drive/folders/1T99iWEIkpcDTJIBi-PSlBwC9T0iKvNX2?usp=sharing 
 
Hi Blissy,

You don't need heat sink pours for the smaller transistors. They should be thermally coupled to the larger transistors which are also handling literally 100x the power of the small ones. So do like I said before and just put them on top. You should create a nice collector pour for them but maybe only 1/4 of the area of one of the larger transistors.

So is there solderable copper exposed to the tab of the large transistors? Or are you just going to bolt them in? I would make that a plated hole so so that you can get a really good thermal bond with a bunch of solder. Otherwise, if you use a bolt, you might need some thermal paste or something like that which I always thought was messy.

Also, it now occurs to me that the big speech I made about all-things-ground converging at the same point is being thwarted by this separate board. This board is now providing the "star" ground point. So you really need to move all of those spades over the the center of the board. Even if it means bending a wires from the middle of this board to an offset location of the other. Then make a ground pour down the middle, above and below stitched with vias and get all of the spade terminals and caps close together.

Where the 3 wires (+17,-17 and AGND) all come off the board, where it goes out AND where it comes in from the other board, the wires need to be close together. The rule is that you want return currents in close physical proximity to source currents. We talk about that all the time here. It's very important. If you have a gap between the wires, it will emit electromagnetic radiation that will be picked up by nearby amplifiers. So you have to move those spades over.
 

Latest posts

Back
Top