I am on about layout....so the answer is yes, you want to isolate the switching noise from the rest of the circuitry as much as possible....
I agree. But no matter how good the layout it is it can't fix, say, a missing control loop in the design.
To use an example which just came up, the LT3439 in the quietpower project Jack linked on the previous page is unregulated, meaning the circuit's reliant on regulation of its supply rail to avoid dropout of the LT1761 and LT1964 regulating the output windings. So it's more a complex inverter implementation than an SMPS since it ducks the problem of generating the input rail from mains. And if one has one OK rail already other methods such as inverting with a charge pump might yield a more compact board, lower BOM, more LDO selection frequency, and ease noise management the layout.
To speak more generally to the thread, Jack's link is nifty. teho labs' quietpower writeup doesn't mention which Mean Well supply it is they measured but, looking through their catalog, from the photos it's the PD-2515. That's a pretty cost effective one. To go back to the OP's question, I can't DIY a regulated 25W +-15 linear supply for what Mouser charges for the PD-2515. I can come close. But it's more assembly work and it'd have more noise in the audio band. The downside is Mean Well's block diagram for the PD-2515 shows feedback from one, positive output rail. The linear implementation would have a 7915.
I find I'm often looking at these sorts of situations. One answer I frequently arrive at is point of load regulation to +-5, +-12, or whatever from a loosely controlled +-15 off an SMPS.
Last edited:
I set up both my linear & the cheapo SMPS side by sid & connected each of them one after the other to my small project preamp. Lo & behold there was hardly any difference! Both units were almost equally quiet/noisy I think & most of the noise emmiting was from my preamp? I then switched on my TV & the fluorescent workbench light to see if these introduced anymore noise. Yes, but only marginally a little more perhaps on the SMPS. However,there was no hum or any EMI/RF interference in either psus!
On the SMPS the BUF634 was dead accurate giving exactly +12v-0--12v! Mind you the SMPS is only a very cheap $7,00 affair yet performed quite well.
So what does this say in REAL LIFE? are SMPS not as noisy as many of us are lead to believe?
Now my quest is to find a well made laptop PSU. I gather most Macs run on 24v (1.85A-2A?) & seem better made than most generic or other models?
All UK versions come with the ground pin...don't know what significance this would have with/without this connected?
Here is a link which SMPS propagators may find interesting!
Resonant SMPS for Audio | H i F i D U I N O
On the SMPS the BUF634 was dead accurate giving exactly +12v-0--12v! Mind you the SMPS is only a very cheap $7,00 affair yet performed quite well.
So what does this say in REAL LIFE? are SMPS not as noisy as many of us are lead to believe?
Now my quest is to find a well made laptop PSU. I gather most Macs run on 24v (1.85A-2A?) & seem better made than most generic or other models?
All UK versions come with the ground pin...don't know what significance this would have with/without this connected?
Here is a link which SMPS propagators may find interesting!
Resonant SMPS for Audio | H i F i D U I N O
Attachments
Last edited:
I experimented with various layouts, until I got it right, and by right I mean working at about 75% of rated current before voltage output collapsed, choke started buzzing loudly etc.
I used at least 4 different types of diodes, I used at least 6 different types of chokes, I even wound my own chokes to make sure, at least 5 PCB layouts...
They *do* mention that there will be parasitics on the switching diode, that might need to be suppressed, well, no Sh*t Sherlock. Anyway the point is that I can build a 100MHz FM transmitter *much* easier than a "simple" LM2679, 5A switching circuit.
I built it on prototype board with ridiculous layout and got 1A-2A easy. The problems start when you exceed the 4A mark.
A three pin mains plug may mean that one side of the output is connected to Protective Earth (PE).
This limits what you can do if you require dual polarity or other series connections of two or more SMPS.
A 2pin (with or without a dummy third pin) plug should show on the data as a double insulated device with the concentric squares symbol.
This MUST have complete isolation between Mains and Output.
These are more flexible in that all will work as series combinations to create dual polarity and other voltage combinations.
I have a 3pin with output -ve connected to the PE.
I also have a double insulated.
I am able to use the 3pin as the +ve supply with zero as the common.
The 2pin then provides the -ve polarity by connecting the 2pin +ve to common (PE) and the -ve pin becomes the -ve supply.
I now have a +, 0 - dual polarity supply with 0 as common AND connected to PE.
dual double insulated avoids the PE connection.
This limits what you can do if you require dual polarity or other series connections of two or more SMPS.
A 2pin (with or without a dummy third pin) plug should show on the data as a double insulated device with the concentric squares symbol.
This MUST have complete isolation between Mains and Output.
These are more flexible in that all will work as series combinations to create dual polarity and other voltage combinations.
I have a 3pin with output -ve connected to the PE.
I also have a double insulated.
I am able to use the 3pin as the +ve supply with zero as the common.
The 2pin then provides the -ve polarity by connecting the 2pin +ve to common (PE) and the -ve pin becomes the -ve supply.
I now have a +, 0 - dual polarity supply with 0 as common AND connected to PE.
dual double insulated avoids the PE connection.
A two stage 4pole LCLC filter between the SMPS and the analogue circuit will massively attenuate the VHF that can enter the analogue circuit.
The voltage loss through the two Ls will be quite low and although these will affect the voltage regulation, you may find values of r that are tolerable to your analogue circuit. The Ls will have some value of "r" as do the connecting cables/traces.
The HF will not be attenuated as much, but careful selection of the Ls and Cs should be sufficient to prevent the analogue circuit reacting badly to impulses on the supply rails.
The voltage loss through the two Ls will be quite low and although these will affect the voltage regulation, you may find values of r that are tolerable to your analogue circuit. The Ls will have some value of "r" as do the connecting cables/traces.
The HF will not be attenuated as much, but careful selection of the Ls and Cs should be sufficient to prevent the analogue circuit reacting badly to impulses on the supply rails.
Well, almost. There's almost always a Y2 ceramic (typically 1 or 2 kV rated) between the primary 0V (not Neutral) and secondary 0V.
I suppose there's always the small risk of breakdown on this Y2 ceramic.
Honestly, I've never seen anyone work it all the way through from proper measurements. So "SMPSes are noisy" goes round and round with hardly ever data to quantify what it means. Hence, it's the subject of myths and legends. And not, you know, actual engineering design. Both SMPSes and linear supplies rely on the same switching mechanisms---it's not uncommon to use the same diodes secondary side in both types of supplies and an off state FET is essentially a reverse biased PN junction just like an open diode. The main difference is the epic quantities of text DIYers have devoted to mains frequency snubbing rarely do a good job of recognizing the current paths and parasitics one has to get right to get good performance from an SMPS. It's actually rather a plug and chug exercise to look at this stuff in Spice. But few DIYers set up the sims, have measurement gear at the level needed to validate the sims, or the skills to understand and extract package and layout parasitics such that the sims are properly informative. The result is, understandably, a retreat into oversimplification.
If you're interested in moving past that I think the most fundamental thing grasp is both an LC post-filter and PSRR of linear amplifiers (op amps, power amps, whatever) tend to roll off at second order. This means that, above the post-filter's corner frequency, the amount of amplifier GBP and slew rate consumed by motion in the supply rails is more or less invariant of frequency. Within the limits of the filter stop band and amplifier noise floor. Since the diodes in linear supplies switch at usually twice the mains frequency (full bridge) the only circumstance where linear post filtering is effective is class A loads where one's willing to sacrifice line regulation---both a supply's load and line regulation fall with its post filter attenuation and the loss of regulation (especially line) tends to be overlooked by DIYers implementing CLC or CRC filtering in linear supplies. In the more typical case of loads with some amount of class C behaviour (the class B part of class AB or class D in general) a linear supply's switching harmonics can't be usefully filtered as the required cutoff frequency is orders of magnitude lower than the load bandwidth.
In comparison, a well designed SMPS in continuous operation switches at 100+kHz. So it's trivial to corner an LC post filter above the nominal 20 to 20kHz audio band and below the switching frequency. Switching is basically delta functions bandlimited to how fast the FETs and diodes turn on and off (typically a few tens of nanoseconds) so there's a baseband portion of the switch energy which passes through the post filter unattenuated. It's industry best practice to follow the switcher with an LDO to keep the baseband residual off analog rails. But DIYers seem unaware of this practice---this thread's one of two or three I can think of where it's been mentioned---despite its benefits for managing the interaction between linear amplifiers and SMPSes in light load/quiescent cases where the switching frequency collapses into the audio band. It's hard to provide a general analysis as there's a substantial number of variables in supply requirements, parts selection, and regulator control schemes. But, basically, LDO manufacturers targeting analog off SMPS are pushing up their parts' regulation bandwidths for good reasons.
The analysis gets more involved if you're operating a linear power amp off an SMPS and want to avoid intervening linear regulators. Conveniently, that scenario's out of scope to this thread.
But some awareness of it is useful here. Loosely speaking, for an SMPS with 100mV of ripple the baseband residual's under 100uV. It's easy to muster sufficient PSRR to put that well below the noise floor. The carrier harmonics are more pernicious as the post filter might provide, say, 10dB of attenuation at 100kHz, meaning one ends up with mV to handle in the frequency independent range. Even with really nice op amps you're probably not getting much over 50dB PSRR against the carrier residual. So the carrier may or may not be buried in the analog noise and may or may not present a SID/TID risk. This is where sims and maybe some help from a fast LDO become useful.A lot of this is really about noise shaping. In the sense an SMPS is a noise shaped version of a linear supply, in the sense post filtering, LDO selection, and in circuit PSRR all govern the in circuit noise residual, and in the sense the noise spectra varies with SMPS topology selection and implementation quality. The analogy with sigma-delta's move out of band and attenuate is inexact. But the basic ideas are the same. And not well understood in the community. So, if you put all that into the pipe, it's natural what comes out the other end is "SMPSes are noisy".
Consider gaining familiarity with how Y caps affect emissions---fair number of write ups on the topic if one cares to look.
Y2 capacitors run from Live to PE and from Neutral to PE.
If there is no PE then there should be no need for Y2 capacitors.
In a double insulated Class11 product, the Mains is isolated from the Output.
There should not be any capacitor between Mains side and Output side.
If there is no PE then there should be no need for Y2 capacitors.
In a double insulated Class11 product, the Mains is isolated from the Output.
There should not be any capacitor between Mains side and Output side.
What is this suggesting?
Thanks for the perspective, and lots of good stuff snipped.
I am indeed using a pair of flyback SMPSes to power a Class-AB power amp without intervening LDOs. However, that amp is an unusual topology with nested feedback:
http://www.diyaudio.com/forums/chip-amps/184165-miniref-schematic-pcb-layout.html
Basically, it nests a Howland Current-Pump 4-quadrant transconductance amp within a conventional voltage-series feedback loop of an outer opamp, a la Walt Jung.
The rails to the chipamp are unregulated (i.e. no intervening LDO) to avoid ohmic losses, and also to take advantage of reasonably high PSRRs in modern chipamps, at least in the audio band. However, I do use moderately large electrolytics on the unregulated rails - about 4700uF per rail, which feeds both channels.
However, the +/- 12V rails to the outer-loop opamp *are* regulated, with simple series resistances + zener shunt regulators and modest capacitances of 220uF/rail. I could improve the regulation of the opamp rails by using some combination of monolithic regulators or series-shunt regulators as needed, but I have not seen the need for it (from audible sonics) so far. I might also consider high-performance discrete regulators for the opamp in a future re-spin.
As things stand, it is dead silent between tracks, which can be uncanny for those of us who are used to hearing a small amount of 50/60 Hz hum from any amplifier with a conventional 50/60 Hz power transformer.
Last edited:
I came across this rather interesting explanation w.r.t reducing noise in SMPS psu, which I found
quite useful for noobs like me!
power supply - Can the noise of any switching PSU be attenuated if I put linear regulator before the output? - Electrical Engineering Stack Exchange
Q:Would adding a Toroidal, Ferrite or an air core inductor in the linear section of a p.s.u help reduce noise?
I've read that toroidals are the best, but a little bulky! How many uH/mH/H should one use in a 12-15VA 2x 15v supply? I tried to calculate but the math(s) was way over my head!
quite useful for noobs like me!
power supply - Can the noise of any switching PSU be attenuated if I put linear regulator before the output? - Electrical Engineering Stack Exchange
Q:Would adding a Toroidal, Ferrite or an air core inductor in the linear section of a p.s.u help reduce noise?
I've read that toroidals are the best, but a little bulky! How many uH/mH/H should one use in a 12-15VA 2x 15v supply? I tried to calculate but the math(s) was way over my head!
Last edited:
That was the first question. I like how we all have not done a dismissive thread. I have asked myself could we benefit from doing this? Try as I might I can not think we might.
On the other hand I have a Hypex SMPS for 2 x UCD180 that might run my Hitachi amps rather well. I have a hunch Hypex cared enough to do a good job and will be very interested to try it. I expect it to be no worse than that amount of money might buy but using less space. There is a chance to regulate the PSU at high voltage and loose the ripple. Doubtful they do but not impossible.
One project we did we did the other thing. We used a mains rectified supply and power amp at 340V DC in MOS FET. The output via transformer. No big deal if a tube amp except never direct from mains. The input opto isolated. This was an industrial amp.
One form of SMPS is charge/use commutator. That is where the equipment never sees the mains or SMPS. In galvanic protection that is used. The regulator after the commutator to stop spikes or brownouts.
Sorry if any of this said already.
Careful with that. What Hypex calls an SMPS lacks line regulation and hence isn't really an SMPS in the usual sense. If you search back through Bruno's early posts here there are some hints as to the though processes behind this but it's not at all clear why Hypex uses a voltage divider (or equivalent) in the feedback loop to set the output voltage rather than a voltage reference.
The supplies on the UcD3xMP backplane are probably of the greatest interest to this thread but one most likely ends up following them with an LM7x12 or 3x7 pair to keep output voltages within acceptable limits in the high mains case. It's not hard to do as good or better with a toroid or some of the more proper SMPSes discussed here. So the arrangement's not of much interest unless one has some additional purpose for the backplane in mind.
The supplies on the UcD3xMP backplane are probably of the greatest interest to this thread but one most likely ends up following them with an LM7x12 or 3x7 pair to keep output voltages within acceptable limits in the high mains case. It's not hard to do as good or better with a toroid or some of the more proper SMPSes discussed here. So the arrangement's not of much interest unless one has some additional purpose for the backplane in mind.
Please bear with my lack of experience....I'm trying connect a 24 v SMPS psu to a linear smoothing stage. The unit itself is completely isolated, but without a separate earth connection. Can I then connect it as shown in the diagram below or what other corrections/components should add to make it work?
I'm thinking especially to the -left- pre LM7812/7912 common & grounding sections of the diagram!
Thanks
I'm thinking especially to the -left- pre LM7812/7912 common & grounding sections of the diagram!
Thanks
Attachments
Last edited:
Drop out voltage for the 78XX series regulators is 2 volts. For 7812/7912 you need a min 14 volts of input voltage for a stable regulation.
Your circuit would work if the smps output voltage were higher than 28 V (14+14) with the condition that the currents of both postive and negative outputs are the same.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.