Swapping class AB amp linear supply with a SMPS - incl measurements

[McFly]

Very nice, though I don't think it beats Benchmark

No regarding SINAD it will not beat Benchmark me think.

Im just gonna requote myself

measures better than any class AB-linear-power-supplied amp measured here at ASR?

But I dont disagree at all, 1khz measurements into fixed loads is not the be all end all. But its a nice quick overview, a bit like square waves are for fault finding.

 

[Waxx]

The best class AB amp i heared was a custom design by an retired Philips engineer. And he used an complete overbuild SMPS with mosfets and a crystal as clock and claimed that that was the most important part to get the noise down in the amplifier. For the rest he used the LM3886 IC as amp in a gainclone like circuit and NE5534 opamps to drive it back in the days (2006), so no esotheric parts or so. At that time i did not understand amps technically yet (he learned me most i know) but i do know that it was the first time i heared a chipamp and i really liked it. He claimed it had less than 0.01% THD, which i can't be sure off, but i do know that it was a very neutral and clear amp and sounded very similar to the Ncore amp (class d) i build shortly after that. I never heared an class AB amp that clean again after this one. But i like a bit of colour so...

But since then i know that an SMPS is often better (except maybe for tube amp, and even then) in being a low noise psu, at least when it's build right. So i'm a big supporter of using them everywhere, even on old style designs like tube amps or class A and AB amps. It's only that the audiophools insist on linear psu, and often even unregulated linear psu's.

I never heared the Benchmark nor the Perreaux amps so i can't judge those. But i was not surprised to see a smps in the Benchmark when i saw the test results and think that is a big factor in the low nosie design of those.

 

[sq225917]

It really isn't. A smps let's you put more power I'm a small space, you can build a linear supply just as quiet as a smps, infact much quieter at HF.

They just need more space for all the parts and the radiation from the traffo. My own Nuerochrome amps have linear supplies, and they're bettter than -110db thd,+n into 8 ohm as built.

 

[Tangband]

So my hypothesis, based on measurements seen on this site and others, is that (audio specific) Switch Mode Power Supplies produce less noise (all 3 ways - directly into the amps DC rails, and via radiated magnetic fields, and mechanically) within the audible 20-20khz band when installed in an amplifier. I set out to draw my own conclusions. And share them here for anyone that's interested.

Disclaimer: Now, as usual, the results are my own, others may vary, my measurement set up did change between the amp having a 'big iron' transformer supply vs the new SMPS. What does this mean? The noise floor and subsequently the THD+N measurements are not really valid between the before and after. Shock horror. But take a look at the 50hz and 100hz noise + harmonics themselves - I think the data there is still useful hence why I'm sharing. Take these with a teaspoon of salt. I am also no electronics engineer - I understand only a little about extra low voltage electronics, I'm much more in tune with speakers. But I am, as I have explained before, a high-er voltage commercial electrical engineer, so I know how to be safe when playing around with appliances, read diagrams and use electrical testing equipment.

The amplifier chosen was a 1982 Vintage class AB Perreaux PMF2150B - made right here in NZ. I believe my serial number puts it around 85-86 year. It has a heavy, noisy EI type transformer inside it. I wanted to find a nice classic piece of HIFI gear, especially after I heard @restorer-john singing this units praises in an older thread. This thing is heavy as hell and hummed like a fridge both mechanically, and through the speakers when connected. Especially the right channel. It was almost unusable as I picked it up on the second hand market. (it was also heavily oscillating which was easily fixed before any of this went down) I tried replacement smoothing caps, bridge rectifier and even soaking the transformer in varnish with very little success. There are other benefits to fitting a SMPS like seriously reduced weight and no mechanical hum. I think some poor SMPS designs can emit very high pitched whines, and some even block out AM/FM radio tuners from working.

Testing conditions;

Class AB amplifier with linear transformer/rectifier/smoothing caps supply. I have attached the amplifier schematic and specs for reference.
1, the right, channel tested - The right channel's input stage is mounted closer to the original transformer and had more power supply noise.
Initial test: - Focusrite Solo input and output - Right channel into 8 ohms - Factory 1200?VA iron transformer, 35A bridge rectifier, 2x 18000uf smoothing caps
After: SMPS test - input - E1DA ADC - output topping E30. Right channel into 8 ohms - Hypex SMPS1200A700 2x85VDC. 1200W supply.

Original Linear Power supply; Focusrite Solo, 5W/6.3V into 8R

View attachment 224420

Test with Hypex SMPS installed and E1DA ADC: 5W/6.3V 8R

View attachment 224422



Notable problems with my testing;

• I definitely didn't get enough output levels/tones/sweeps logged when I did the original measurements to compare different power levels etc
• Obviously, the aforementioned testing equipment
• I could have tried replacing the transformer with a Toroid type - supposed to have less magnetic field spray - but the cost of this to NZ was more than the hypex SMPS, much more.
• The (82V) rail voltages are still slightly high with the hypex SMPS (measured 84VDC) - yes but they were also higher with the original transformer, even with the 240V tap (our mains are 230V, but hover around 238V at my place) when I first got the amp I measured 92VDC on the rails with the 230V transformer tap.
• There could be much more out of audio band noise that I didn't test for/look at

Results;

• Notwithstanding the test equipment, the THD+N has gone from 0.3% or -50dB to 0.003% or -90DB, which is somewhat correct due to the 50hz noise being -50dB down in the initial testing. Thats a 40dB SINAD improvement just changing the power supply.
• The amp is now DEAD SILENT, everywhere. I can now use it. Cant hear anything with my ear 2cm from the cones of speakers or the amplifier chassis.
• The amp now measures well below factory specifications for THD, THD+N, hum and noise. I cant say for certain that it didn't before due to the test equipment limitations.
• The amp is probably saved from the landfill, should someone less enthusiastic have purchased it
• Subjectively (notwithstanding the original hum) the amp sounds exactly the same, if not better because there's no hum in silent passages. I actually enjoy using it now.
• There are numerous references to SMPS sounding 'flat' and lacking the 'transient current availability' of linear supplies. I hear none of this, although it is hard if not impossible to do an AB comparison. Would need another factory PMF2150b (and use my auto amp switcher I made.) This could be the case if I drove the amp hard
• What I will admit I have likely done, is limited the 'headroom' of the amplifier - the old linear supplies have been provably better in temporarily overloading which does translate to better headroom for dynamic peaks, however my peaks in my listening never go beyond the 200WPC output of this amplifier anyway. Sure, if you always run your amps hard you may want to stick with old iron, or, oversize the SMPS to something like a 2KW or even 3KW - then you'd be fine.
• The rail voltages don't sag as much under full load with the SMPS compared with the linear supply - could this be going against the design topology and affecting the sound somehow?
• I also admit, the amplifier will likely no longer meet the brochure specs with regards to the full FTC sections, as the power supply would overheat before getting there. I don't listen to full scale sine waves though so I'm happy to live with this.

Conclusion;

I can say for certain my testing method is flawed. I cant say for certain that SMPS installed into any class AB (or other class) will net an improvement as this is only one amplifier. There could be much more out of audio band noise that I didn't test for/look at. But the result here is a subjective and objective success, at least to me. First, it fixed the actual physical problems with the amp and second, appears to measure better at least in the areas the power supply affects the output. I didn't set out to say that old iron linear supplies are sh*t - not at all, there are amps with fantastic low noise iron supplies, but for sure the one in my amp here was either no good, or had deteriorated so much over time. Now I don't get a sore back lifting it.

Thoughts, was it worth it, should I stick the iron back in?

P.S. I got the matching SM2 preamp with it and am going to the E1DA to see if a good old fashion 'recap' does anything to the preamp.

Biggest issues with SMPS is that they (sometimes) contaminates the mains meaning they can make other hifi components sound worse . But maybe this issue was more common in the past than today? You could try to measure the SINAD of another hifi components connected on the same power block with first the linear power supply and then the SMPS. Is there any difference ?

Im not saying that this issue are serious in every SMPS ( linear power supplies also contaminates the mains ) so you have to measure to be sure.

 

[Waxx]

Biggest issues with SMPS is that they (sometimes) contaminates the mains meaning they can make other hifi components sound worse . But maybe this issue was more common in the past than today? You could try to measure the SINAD of another hifi components connected on the same power block with first the linear power supply and then the SMPS. Is there any difference ?

Im not saying that this issue are serious in every SMPS ( linear power supplies also contaminates the mains ) so you have to measure to be sure.

If it got a (mandatory in the EU) wel build EMI/RFI filter at the AC input that should not be the case. But it's true that in the past this was not the case and SMPS's had switching noise, both on the DC (because the lack of or not enough dc output filtering) to the amp and to the AC on the power grid. Modern certified SMPS's don't have those anymore.

 

[Tangband]

If it got a (mandatory in the EU) wel build EMI/RFI filter at the AC input that should not be the case. But it's true that in the past this was not the case and SMPS's had switching noise, both on the DC (because the lack of or not enough dc output filtering) to the amp and to the AC on the power grid. Modern certified SMPS's don't have those anymore.

Thanks for verifying this info .
A question : What year was this rule with EMI/RFI filter at the AC input in EU launched ?
Is a switched mode powersupply , 15 years old constructed with those filters ?.

 

[Waxx]

If it got a (mandatory in the EU) wel build EMI/RFI filter at the AC input that should not be the case. But it's true that in the past this was not the case and SMPS's had switching noise, both on the DC (because the lack of or not enough dc output filtering) to the amp and to the AC on the power grid. Modern certified SMPS's don't have those anymore.

I don't know when it was set mandatory, but it's part of the LVD regulations of the CE directives, manadatory for all electric equipment sold and used in the EU. But that directive is constant under review to keep up with modern tech, so it's hard to find out when that part was made. The CE directives were installed in 1993 as part of the Shengen agreement of free trade and travel without restrictions within the EU and some asociated countries like Norway and Switzerland.

 

[Piere]

A CE mark on the SMPSU only says that the manufacturer garantees the brick complies to the EU regulations. Nothing more. But even then it also largely depends on the use. The final use can still violate compliance. Therfore the whole finished product that contains a CE qualified SMPS part must be tested for compliance.

 

[Deleted member 48726]

I did.

I dumped 400W into 2x channels 4ohm simultaneously for 30 seconds, and measured 1300W from the wall. As the hypex had no pfc front end, it was actually drawing 2230VA. Good thing we only pay for Watts!!

LOL
That is actually not legal in my country if it was in the industry..

 

[Piere]

To my knowledge there is no EN-IEC directive about PF requirements. But maybe that has changed recently? Local energy suppliers however can ask for minimal PF requirements.

 

[Deleted member 48726]

To my knowledge there is no EN-IEC directive about PF requirements. But maybe that has changed recently? Local energy suppliers however can ask for minimal PF requirements.

It's a national legislation in my country that cos-phi must be between 0.9-1 in industrial installations.

 

[egellings]

So my hypothesis, based on measurements seen on this site and others, is that (audio specific) Switch Mode Power Supplies produce less noise (all 3 ways - directly into the amps DC rails, and via radiated magnetic fields, and mechanically) within the audible 20-20khz band when installed in an amplifier. I set out to draw my own conclusions. And share them here for anyone that's interested.

Disclaimer: Now, as usual, the results are my own, others may vary, my measurement set up did change between the amp having a 'big iron' transformer supply vs the new SMPS. What does this mean? The noise floor and subsequently the THD+N measurements are not really valid between the before and after. Shock horror. But take a look at the 50hz and 100hz noise + harmonics themselves - I think the data there is still useful hence why I'm sharing. Take these with a teaspoon of salt. I am also no electronics engineer - I understand only a little about extra low voltage electronics, I'm much more in tune with speakers. But I am, as I have explained before, a high-er voltage commercial electrical engineer, so I know how to be safe when playing around with appliances, read diagrams and use electrical testing equipment.

The amplifier chosen was a 1982 Vintage class AB Perreaux PMF2150B - made right here in NZ. I believe my serial number puts it around 85-86 year. It has a heavy, noisy EI type transformer inside it. I wanted to find a nice classic piece of HIFI gear, especially after I heard @restorer-john singing this units praises in an older thread. This thing is heavy as hell and hummed like a fridge both mechanically, and through the speakers when connected. Especially the right channel. It was almost unusable as I picked it up on the second hand market. (it was also heavily oscillating which was easily fixed before any of this went down) I tried replacement smoothing caps, bridge rectifier and even soaking the transformer in varnish with very little success. There are other benefits to fitting a SMPS like seriously reduced weight and no mechanical hum. I think some poor SMPS designs can emit very high pitched whines, and some even block out AM/FM radio tuners from working.

Testing conditions;

Class AB amplifier with linear transformer/rectifier/smoothing caps supply. I have attached the amplifier schematic and specs for reference.
1, the right, channel tested - The right channel's input stage is mounted closer to the original transformer and had more power supply noise.
Initial test: - Focusrite Solo input and output - Right channel into 8 ohms - Factory 1200?VA iron transformer, 35A bridge rectifier, 2x 18000uf smoothing caps
After: SMPS test - input - E1DA ADC - output topping E30. Right channel into 8 ohms - Hypex SMPS1200A700 2x85VDC. 1200W supply.

Original Linear Power supply; Focusrite Solo, 5W/6.3V into 8R

View attachment 224420

Test with Hypex SMPS installed and E1DA ADC: 5W/6.3V 8R

View attachment 224422



Notable problems with my testing;

• I definitely didn't get enough output levels/tones/sweeps logged when I did the original measurements to compare different power levels etc
• Obviously, the aforementioned testing equipment
• I could have tried replacing the transformer with a Toroid type - supposed to have less magnetic field spray - but the cost of this to NZ was more than the hypex SMPS, much more.
• The (82V) rail voltages are still slightly high with the hypex SMPS (measured 84VDC) - yes but they were also higher with the original transformer, even with the 240V tap (our mains are 230V, but hover around 238V at my place) when I first got the amp I measured 92VDC on the rails with the 230V transformer tap.
• There could be much more out of audio band noise that I didn't test for/look at

Results;

• Notwithstanding the test equipment, the THD+N has gone from 0.3% or -50dB to 0.003% or -90DB, which is somewhat correct due to the 50hz noise being -50dB down in the initial testing. Thats a 40dB SINAD improvement just changing the power supply.
• The amp is now DEAD SILENT, everywhere. I can now use it. Cant hear anything with my ear 2cm from the cones of speakers or the amplifier chassis.
• The amp now measures well below factory specifications for THD, THD+N, hum and noise. I cant say for certain that it didn't before due to the test equipment limitations.
• The amp is probably saved from the landfill, should someone less enthusiastic have purchased it
• Subjectively (notwithstanding the original hum) the amp sounds exactly the same, if not better because there's no hum in silent passages. I actually enjoy using it now.
• There are numerous references to SMPS sounding 'flat' and lacking the 'transient current availability' of linear supplies. I hear none of this, although it is hard if not impossible to do an AB comparison. Would need another factory PMF2150b (and use my auto amp switcher I made.) This could be the case if I drove the amp hard
• What I will admit I have likely done, is limited the 'headroom' of the amplifier - the old linear supplies have been provably better in temporarily overloading which does translate to better headroom for dynamic peaks, however my peaks in my listening never go beyond the 200WPC output of this amplifier anyway. Sure, if you always run your amps hard you may want to stick with old iron, or, oversize the SMPS to something like a 2KW or even 3KW - then you'd be fine.
• The rail voltages don't sag as much under full load with the SMPS compared with the linear supply - could this be going against the design topology and affecting the sound somehow?
• I also admit, the amplifier will likely no longer meet the brochure specs with regards to the full FTC sections, as the power supply would overheat before getting there. I don't listen to full scale sine waves though so I'm happy to live with this.

Conclusion;

I can say for certain my testing method is flawed. I cant say for certain that SMPS installed into any class AB (or other class) will net an improvement as this is only one amplifier. There could be much more out of audio band noise that I didn't test for/look at. But the result here is a subjective and objective success, at least to me. First, it fixed the actual physical problems with the amp and second, appears to measure better at least in the areas the power supply affects the output. I didn't set out to say that old iron linear supplies are sh*t - not at all, there are amps with fantastic low noise iron supplies, but for sure the one in my amp here was either no good, or had deteriorated so much over time. Now I don't get a sore back lifting it.

Thoughts, was it worth it, should I stick the iron back in?

P.S. I got the matching SM2 preamp with it and am going to the E1DA to see if a good old fashion 'recap' does anything to the preamp.

The switch mode supply likely has voltage regulation, whereas the 60Hz standard supply does not have that. Maybe the regulation is what's improving the performance of the amplifier, not whether the supply is switch node or not.

 

[restorer-john]

if it wasn't for the measurement artifact I get at 16khz. (shows up in all my testing - think its inherent in my laptop)

Did you ever get to the bottom of the 16kHz artifact? The reason I ask, is I get the same artifact on a PCs with with a PCI slot soundcard (not USB).

Not present on loopback, but appears if you put any line level preamp in the loop.

 

[McFly]

WIFI and/or Bluetooth - I think it was. I did find the source but now I forget, but pretty sure it was WIFI. Switch off all radios - routers, laptops, desktops, phones, repeaters, taplets etc anything that transmits OR receives WIFI. If you are measuring on a desktop PC you may need to disable the WIFI/BT radio in the BIOS, not just switch it off in windows. Sometimes it was a spike at 16kHZ and sometimes it was a spike at 8khz, and a lot of times I would expand the FFT scale and see a bigger spike at 24khz (direct derivative of 2.4ghz i think?)

I think for these measurements you speak of I just didnt bother switching off all nearby radios.

Best to measure with a laptop, disconnected from mains and all wifi/bt radios disabled.

With the E1DA ADC, I'm seeing all kinds of cray sheet. I see a rolling HF noise floor if I have the cheap LED garage lights turned on, above the measurements. I have to turn my lights off!

 

[McFly]

The switch mode supply likely has voltage regulation, whereas the 60Hz standard supply does not have that. Maybe the regulation is what's improving the performance of the amplifier, not whether the supply is switch node or not.

The Hypex does not. The DC rails fluctuate up and down with the incoming mains AC, exactly like a linear supply does.

 

[restorer-john]

WIFI and/or Bluetooth - I think it was. I did find the source but now I forget, but pretty sure it was WIFI. Switch off all radios - routers, laptops, desktops, phones, repeaters, taplets etc anything that transmits OR receives WIFI. If you are measuring on a desktop PC you may need to disable the WIFI/BT radio in the BIOS, not just switch it off in windows. Sometimes it was a spike at 16kHZ and sometimes it was a spike at 8khz, and a lot of times I would expand the FFT scale and see a bigger spike at 24khz (direct derivative of 2.4ghz i think?)

Interesting. I killed the WiFi, bluetooth and all my SMPS/LED supplies etc and ran the PC on the LAN and it was still there back when I got the sh#ts and just left it. But... there was an internal PCI WiFi card in the PC, right near the M-Audio card. I'm gonna pull it out and see if that does the trick.

This is what the old M-Audio AP 24/192 will do on loopback:

 

[McFly]

I just remembered - There is also a lot of history online about 8khz noise and it having something to do with USB - 8khz is the close cousin to 16khz. Check all USB drivers, keep in mind different sockets on a desktop PC are connected to different parts of the motherboards chipset so moving sockets or using different computers can solve it.

I think I had more success with usb2.0 sockets (usually grey or black, placed highest on the mobo near where your PS2 ports are/used to be) than I did on any of the 3.0,3.1 and 3.2 sockets (blue, red and type C, respectively, usually)

Desktop computers are full of noise, unfortunately. Other than fan noise, some even squeal at 10-15khz with the damn inductors on the VRMs that power the CPUs.

 

[McFly]

I actually just found this Old ASR thread/post - food for thought

 

[Chrise36]

WIFI and/or Bluetooth - I think it was. I did find the source but now I forget, but pretty sure it was WIFI. Switch off all radios - routers, laptops, desktops, phones, repeaters, taplets etc anything that transmits OR receives WIFI. If you are measuring on a desktop PC you may need to disable the WIFI/BT radio in the BIOS, not just switch it off in windows. Sometimes it was a spike at 16kHZ and sometimes it was a spike at 8khz, and a lot of times I would expand the FFT scale and see a bigger spike at 24khz (direct derivative of 2.4ghz i think?)

I think for these measurements you speak of I just didnt bother switching off all nearby radios.

Best to measure with a laptop, disconnected from mains and all wifi/bt radios disabled.

With the E1DA ADC, I'm seeing all kinds of cray sheet. I see a rolling HF noise floor if I have the cheap LED garage lights turned on, above the measurements. I have to turn my lights off!

I just remembered - There is also a lot of history online about 8khz noise and it having something to do with USB - 8khz is the close cousin to 16khz. Check all USB drivers, keep in mind different sockets on a desktop PC are connected to different parts of the motherboards chipset so moving sockets or using different computers can solve it.

I think I had more success with usb2.0 sockets (usually grey or black, placed highest on the mobo near where your PS2 ports are/used to be) than I did on any of the 3.0,3.1 and 3.2 sockets (blue, red and type C, respectively, usually)

Desktop computers are full of noise, unfortunately. Other than fan noise, some even squeal at 10-15khz with the damn inductors on the VRMs that power the CPUs.

Archimago described it as 8/16 khz usb packet noise:

MEASUREMENTS: Computer USB port noise, USB hubs and the 8kHz PHY Microframe Packet Noise

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

 

[Piere]

Interesting. I killed the WiFi, bluetooth and all my SMPS/LED supplies etc and ran the PC on the LAN and it was still there back when I got the sh#ts and just left it. But... there was an internal PCI WiFi card in the PC, right near the M-Audio card. I'm gonna pull it out and see if that does the trick.

This is what the old M-Audio AP 24/192 will do on loopback:

That M-Audio Audiophile192 was one of the best PCI cards you could get at that time at almost a bargain price. 12 years ago I modified mine a bit by placing extra RC's at the + an - 12V lines and replaced the NE5532 by LM4562 op-amps. My Lab system PC was a dual core Athlon at that time running XP on an Asus mobo, Matrox entry video card next to the Audiophile card and a no-name 450W standard power brick. No special measures apart from careful audio cabling (balanced loop-back). The result:



I think AKM was way ahead with their AD/DA converters. Nearly no PC related harmonic spurs! Only a slight 3rd harmonic. You will have hard times to get a better result with external USB cards today. My current Topping D30Pro - E1DA measurement combo does not better! (But also not worse )