Popular Post Superdad Posted January 31, 2021 Popular Post Share Posted January 31, 2021 On 1/8/2021 at 1:36 PM, Miska said: This is pretty good one! And doesn't cost much either. https://www.meanwell-web.com/en-gb/ac-dc-medical-desktop-adaptor-output-5vdc-at-6a-ac-gsm60b05--p1j 100 milliVolts p-p DC output noise? No thanks. (Just to give the lay folk perspective, the output noise of the Linear Technology LT3045 is a few microVolts; 100 milliVolts = 100,000 microVolts.) Moreover, all the “medical” grade SMPS units do is reduce the low-source-impedance “touch current” leakage. They do not at all reduced high-source-impedance leakage. (I can post measurements of such if people wish when I’m back in the office.) Here is a long post on the subject that @JohnSwensonmade a bit over 3 years ago: Leakage current has been around since AC power went into houses. All AC power supplies have it in some form, including linear supplies. In the 60s a couple of,engineers actually measured and modeled leakage current in audio systems. Given the time frame it was all from linear supplies, SMPS were a long way in the future. Different LPS implementations turn out to have significant differences in the leakage they produce. In the audio realm the effects of leakage that were important concerned generating voltages across loads and sources, even with tube circuits these are usually significantly less than 1 Mega Ohm, thus in what I am calling the "low impedance" range. This analysis of leakage current became quite important in the emerging medical instrumentation business (heart monitors etc), since electrical equipment was being deliberately connected to human bodies it was very important to know if this leakage current could be dangerous to humans. Since they are worried about mA range of current the leakage that was important had to be fairly low impedance to generate significant current. Thus a LOT of leakage analysis, testing tools, testing standards etc were focused on low impedance leakage. It was not specifically decided to ignore high impedance, but the effects of interest could only be produced by low impedance leakage, so that is what was studied. The result of this was that all leakage testing was done with circuits and test equipment that was designed to work at 1 Mega Ohm or less. With linear supplies this was perfectly sufficient. Then along came SMPS. It turns out that SMPS are very different with regard to leakage then LPS. First is frequency, linear leakage is power line frequency related (60, 120, 180 etc), but SMPS have a huge range of frequencies due to the switching nature of their operation. They ALSO include the traditional 60, 120, 180 etc. SMPS have been extensively tested for leakage, but it has been done with all the existing test equipment and methodologies, thus focusing on low impedance leakage. Unfortunately it turns out that SMPS also include a high impedance component to their leakage, this is way above 1 Mega Ohms. The problem is that the existing test equipment and methodologies shunt this high impedance leakage to ground so they never see it. So nobody knew it was there. This high impedance leakage is significantly higher in intensity than the traditional low impedance leakage, so it can actually have a significantly larger affect on audio systems than traditional leakage, but nobody knew it was there. Do not confuse the high impedance with high frequency. The SMPS contains high and low impedance components at all frequencies. Thus even at 60 Hz, there are both high and low components. This MUST mean that there are at least two different mechanisms contributing to the leakage simultaneously. One with a high impedance and one with a low impedance. The same thing happens at the higher frequencies. That amplitude ratio between high and low impedance varies with frequency (this is varies radically from one model to another), but both components seem to exist across the frequency range. Currently I do NOT know what these mechanisms ARE, just that they must exist due to the behavior of the leakage. So please don't ask what is causing this, I don't know. If you have leakage from a source (PS), it can show up in several ways. One is direct flow to earth ground. If the PS that is the source of the leakage has an electrical path to something that is grounded (such as a DAC, preamp, poweramp etc), maybe an interconnect, USB cable, Ethernet cable etc, the leakage current will create a voltage across the impedance of the cable, frequently the "ground wire" or shield of the cable. This can add noise to the intended signal. This is how leakage current has traditionally shown up in audio systems, as low frequency "hum or buzz" at the preamp or poweramp, because they were grounded. Another way leakage can get into systems is through a DAC, the leakage current can go through the ground plane of the DAC PCB, that current creates a small voltage which modulates the oscillators(s) producing the clocks in the DAC, adding jitter to those clocks. Even if the leakage doesn't get to a preamp or power amp it can add jitter to the clock in the DAC, thus subtly distorting audio output. This leakage from a computer through a DAC has been particularly important in computer audio since most computers are powered by SMPS. In both the above cases the leakage here is composed of both the high impedance and low impedance components. The leakage current does not have to go directly to an earth ground, it can also go from one power supply to another power supply, even if both have two prong plugs. This is what I have called a leakage loop. I have already written extensively about leakage loops so I am not going to go into it here. So how do I know high impedance leakage exists and how do I measure it? A couple months ago I was looking into leakage current and was trying out several different detector circuits and started seeing very strange results that didn't make any sense. I ran a whole bunch of tests on different SMPS models and had a hard time coming up with correlations, things just were not making any sense. I was trying to figure out what could be causing this. After many weeks of trying different things it started to look like the leakage might be very high impedance (over a hundred Mega Ohms). A few simple tests confirmed that this was in fact true. (I still didn't know it was BOTH high and low at the same time). But that presented a quandary, how in the world do you measure that. All my test equipment maxed out at 10 Mega Ohms which make it impossible to properly measure such high impedance signals. It turned out I couldn't even buy test equipment for this (at least not that I had any chance of affording) so I had to build my own. That took a little while to design and build, but I finally had a differential probe with around 10 Giga Ohms input impedance, AND very low noise. With this tool I could now properly measure this very high impedance leakage. Unfortunately it was STILL doing really weird things. Another round of tests revealed that the leakage was composed of both a high impedance and low impedance part at the SAME frequency. Wow that was something I had not anticipated. I devised a series of tests to check this and sure enough, the results clearly showed both a high impedance and low impedance component at the same time from the same supply. Unfortunately this makes dealing with leakage way more complicated than I had ever imagined. All the methods I had been using and discussing for getting rid of leakage were all focused on the low impedance component, which work for that, but frequently don't touch the high impedance components. So how do you deal with leakage now that we know about both the high and low impedance components? It turns out that there is no single method that works well for both, so you have to come up with different methods, one for high and one for low and figure out how to apply them together. There are two broad categories of how to stop leakage: 1) series block 2) shunt Series block sticks something in series with the leakage path which prevents the leakage from going through. But in order to be useful it has to let whatever the signal is go through. This manifests itself with various isolation schemes that have been tried over the years. These work by increasing the impedance to the leakage, but still letting the signal go through. These work fairly well for the low impedance components, but the rise in impedance for the leakage is not nearly high enough to block high impedance components, they sail right through these isolation mechanisms. This is where the shunt comes in. It turns out it is very to get the high impedance components to shunt around your sensitive components, instead of trying to block them, you just make them go somewhere else. The easiest way to do this is to shunt them to ground and the power supply itself. It CAN be done in other parts of the system, but shunting to ground at the source is the easiest way to deal with it. Unfortunately the shunt does not deal with the low impedance part. So you need to do BOTH the shunt to ground and the series block. THAT will get rid of it all. The series block is going to be different depending on what the "signal" is. For a power supply the "signal" is DC power. So just sticking in a resistor is not going to work, it will block the leakage but it also blocks DC. SO you need to get more creative. A magnetic circuit that passes DC but blocks 60Hz and up would work, but that is very large, heavy and expensive. This is where the LPS-1 comes in, it blocks all low frequency leakage, but does not block the very high impedance leakage. So use either an LPS to drive it or an SMPS whose output is grounded to shunt the high impedance component. For high frequency signals such as Ethernet the existing transformers are sufficient to block the low impedance components of leakage. Leakage even from SMPS is still significantly lower in frequency than Ethernet signalling so a properly designed transformer will have a high enough impedance at the lower frequencies to block the low impedance components, but NOT the high impedance components. SO you still need to shunt the high impedance components and the transformer will take care of the low. Theoretically you could do the same with USB, BUT USB is not just AC, it requires DC connectivity through the data pair, so a transformer will not work. This has made series blocking very difficult to deal with. There are a few solutions, but none of them block the high impedance components, so you still need to shunt the all the high impedance source before they get to the USB cable if you want to stop ALL the leakage from getting through to a DAC. Stopping the low impedance leakage from getting through an audio interconnect is a difficult task. The leakage and the audio are in exactly the same frequency range so you can't separate them that way. The only known way to do this is with a balanced system. In many cases the leakage will be the same on both signal wires, but the audio will be differential, a proper differential input can block the leakage. BUT most implementation will NOT stop the high impedance component, so you STILL need to short it out before it gets there. Unfortunately not all balanced system are created equal. There are several implementations that do the differential input in such a way that it still doesn't block low impedance leakage. So a differential input MAY block low impedance leakage, it may not. Its best to get rid of it before it ever gets to the audio section in the first place. Wow that was a lot longer than I thought. I hope this makes sense and is useful to people. John S. ================== Now besides all that, I will add these points: a) Some of the better quality SMPS units can have lower output impedance—especially at low frequencies—than many of the cheap linear supplies we have seen. And low output broadband output impedance is VERY important for digital audio devices, given the very “bursty” nature of how they draw current. b) Modern Class VI SMPS AC>DC adapters are required to comply with strict standards regarding what they put back into the wall mains. The harmonic switching noise they put back into the wall is extremely high frequency, extremely broadband (at those high frequencies), and extremely low in level. So I think the aversion to SMPS that some audiophiles have based on belief that they pollute the mains is unfounded. c) Most conventional linear power supplies kick quite a lot of harmonics back to the wall—owing to their diode bridges—because they are not drawing current during more than 50% of the AC wave cycle time. This lack of “power factor correction” and the cumulative distortion of the mains is, aside from consumption inefficiency, part of the reason that governments have been trying to regulate out of existence the use of linear AC>DC power supplies. [As an aside, using a large filter choke—besides making much easier the job of the output regulators—results in a “power factor corrected linear supply. Our own JS-2 is designed this way. Plug a conventional component/LPS into a Kill-A-Watt type device and you’ll see a power factor reading of about .5 (draw over just 50% of waveform); plug our choke-filtered JS-2 in and it will read about 0.97.] Thus the main “evil” of an SMPS for audio is not their mediocre DC output noise nor the spread-spectrum ultrasonics they put into the wall. Rather it is the copious amounts of both high and low frequency leakage currents (common-mode AC traveling over DC connections throughout our audio systems). —Alex C. R1200CL and guiltyboxswapper 2 UpTone Audio LLC Link to comment
Superdad Posted January 31, 2021 Share Posted January 31, 2021 49 minutes ago, plissken said: The problem is this is 100% bullshit. There are crap SMPS just as there are crap LPS's. Ah, the ever profane and incorrect Mark Brown. Yes there are a lot of LPS’s whose performance is poor for various reasons—typically high output impedance. But @airguitaris correct to state that SMPS typically have much higher output noise than a decent LPS. Proof? Let’s take this to the extreme with the example of the absolute lowest output noise SMPS available as far as I know: The Daitron RFS50A: https://www.daitron.com/documents/RFS50A_Catalog_En.pdf Most modest-current SMPS units will specify about 80mV (80,000 microVolts) of noise. The wonderful Daitron offers 1mV (1,000uV) output noise. That is indeed outstanding for an SMPS, but in 250 unit OEM purchase quantity the thing is still, last I checked, $302. Building an LPS down into the single-digit milliVolt range is trivial (though as discussed, DC output noise is not the most important criteria for an “audiophile” LPS—or at least not the only one. However, advanced LDO regs such as the LT3045 make producing supplies with output noise in the single or teens digit microVolt range pretty straightforward. Of course there are numerous other PS design considerations—and the LT3045 maxes out at 0.7A (most designers pile a bunch in parallel for more current; or choose from the wide world of higher current LDOs still in the microVolt noise range). UpTone Audio LLC Link to comment
Superdad Posted February 2, 2021 Share Posted February 2, 2021 2 hours ago, Miska said: Well, it doesn't have ground leakage because it doesn't have ground! We never said ground leakage. Leakage is common-mode! I have a whole series of leakage measurement graphs taken from a Mean Well GSM40B "medical" SMPS and for the form of high-source-impedance leakage John has been explaining, they are no better than any of Mean Well's non-medical models. UpTone Audio LLC Link to comment
Popular Post Superdad Posted February 6, 2021 Popular Post Share Posted February 6, 2021 4 hours ago, jabbr said: Substitute "common mode noise" for "leakage currents" and this discussion will become comprehensible again. There is no disagreement here. John and I have always stated that leakage currents are common-mode AC (traveling over various nominally “DC” connections. We talk about them as “leakage” primarily because of their source, which is often overlooked. 3 hours ago, jabbr said: One of the reasons that SMPS cause more common mode noise problems is that the higher switching frequencies increase the role of parasitic capacitances, also the small transformers used in such supplies tend to have higher capacitative coupling which allows common mode noise to sail through... I think a large part of the capacitive coupling letting that common-mode leakage noise through comes from the safety-mandated ‘Y’ capacitors positioned right across the (primary/secondary) of the transformers of nearly every SMPS in the world. 2 hours ago, jabbr said: The little itty bitty transformers in the copper Ethernet PHY are more effective at blocking lower frequency common mode noise but given their parasitic capacitance do let higher frequency common mode noise "sail through" Hence the reason our EtherREGEN uses (difficult to source) 12-core-per-port magnetics, whose center-taps are grounded—followed by our A>B side isolation “moat” consisting of differential isolators. No leakage gets past that moat. jabbr and One and a half 2 UpTone Audio LLC Link to comment
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now