Matt Garman, I'd also like to add that the separate cover/shipping expenses on ebay are higher cost than a single competent supplier.For a stereo MOD286, the Connex SMPS300RE (±30 V version) is perfect. For a 4-channel amp, I'd go with the SMPS500R (±30 V version).
Or to paraphrase: Can I get my 2018 Audi RS8 configured with $50 Walmart tires, please. I'm trying to save a buck.
The quality of Alibaba/eBay/etc. electronics is dubious to say the least. Sometimes you get what you pay for. Many times you don't. If the supply fails catastrophically, it might take both the MOD286 and the speakers with it.
The Connex supplies tend to fail to 0 V when they die due to overload (ask me how I know). They just stop putting out voltage. That's the best scenario for a supply failure.
The majority of the cost will be the chassis, board, and parts. A Connex supply is maybe 15-20 % of the total build cost.
Tom
Too much politeness is Tom's one failing. But, if he has hinted that you won't enjoy an unfortunate supply, then simply apply some gain to that comment:
SO, Ima translate:
That awful shite will instantly neuter your pets and yourself, and turn all your furniture to the modern dove gray color, same as unpainted drywall, which is not a concern while it is on all on fire, deliver 1900 pounds of manure directly atop your house; and, you won't even notice those problems due to the many other things that have gone wrong.
Matt, Perhaps I have errored the translation slightly; however, it is still likely that using an unpleasant supply is something that you should make a new goal to avoid.
toms amps are immune, but what's a smps doing to your source and preamp?
I'm still not a fan, I run too much old hifi with psus that predate smps and hence tend to poorer HF psrr.
Have you looked at the performance of my DIFF PRE 8x2 preamp or HP-1 headphone amp?
I'm getting noise levels a hair below 1.0 uV RMS and THD below -130 dBc on the HP-1. It uses two Mean Well switching bricks (with post-filter and linear post-regulators).
You're free to believe whatever you want. All I can do is to encourage folks to take a look at the measurements as they do document absolutely state-of-the-art performance.
Tom
I like my old school power supplies, and it is sure that the consideration prior to build was a week or two's more effort than the soldering. No matter how well they work (they're huge, they're heavy, but they're fantastic), what I don't like is the expense. I just don't like explaining that one need go through all those steps prior to doing any audio thing.
Instead, it is possible that using an SMPS might be easier, had it been tamed.
Ah! An smps makes higher pitched noise (noise, this time used to explain non-music signal polluting the DC reference).
It is sure that the higher pitched noise of an SMPS will defeat most regulators and capmulti's.
There are solutions:
Firstly, I'd like to mention the confusion between the voltage sense line and using the dc umbilical cable as a helpful resistant element as a portion of a filter. These needs cannot match up unless you have an additional conductor included in the SMPS's dc umbilical cable, with one of the conductors a sense line that runs the whole length of the cable. . . instead of the sense located at the edge of the circuit board inside the SMPS. Having the sense line terminate at the edge of the amplifier board will make it sag-proof and it will also retard the maximum amount of audio-relevant noise.
If that cannot be done, you'd need a series linear regulator capable of including HF, and possibly a relevant description follows.
You can't do solutions unless you know what the problem was. Oh, my apologies, but that's not the least bit true. A more accurate way to say it is that it will be more convenient and more directly applicable to make solutions to an identified problem.
So as not to disappoint, here's my crudeness: Put knots every 6" in the dc umbilical cable and utilize a ferrite at the edge of the amplifier board. It isn't much of a surprise. I would have mentioned my favorite mis-use of diodes and caps, except that these different approaches are entirely cross compatible without interference.
For more orthodox, there's a few ST datasheets that have "nested regulators" when double the parts count has been used to reach the higher frequencies more successfully, including SMPS noise. I have always loved the tone of regulators. And, although they can sound a little flat, my easy crude means can wipe off that problem.
For more measured, Keantoken has arranged a capmulti that can work for any signal imaginable, is adaptable to high current, and has already been proofed for audio. Had one added up relevant concerns, then Kean's parts count is quite low. One effort to solve, one board, and low parts count? Kudos!
There is curiosity:
What ever happened to SMPS? The older (at least a decade older) SMPS worked for audio directly, and so easily well. However, the new production SMPS can be a lot of trouble to apply for audio. So, what inconvenience has happened recently?
Last edited:
I'd like to see your math there. With a well-designed LC filter, I'm having no trouble knocking any switching noise down to levels approaching the noise floor.
Also: Modern switching supplies switch at frequencies well beyond the audio band. Many, including the Mean Well units, use spread spectrum clocking, thus create a low level noise in wide-ish band rather than a tall peak.
The switching frequency is low enough (usually 50-100 kHz, some higher) that modern opamps still have plenty of PSRR and loop gain left.
A ferrite bead will do absolutely nothing to the 50-100 kHz switching frequency of a modern switcher. Even at 500 kHz - 1 MHz used by some switching ICs, the ferrite beads remain ineffective.
An umbilical cord will also do nothing to the switching frequency. It will, however add supply impedance, which will hurt performance within the audio band.
Tom
These statements are fine with me.
I had done an incomplete job of the prior post. It had to do with assumptions of the average/basic audio amplifier, had it been run from SMPS.
My apologies that such a thing was not relevant to this thread, and not relevant to your amplifier.
Probably, concerns about SMPS power should have been on a different thread/page.
Compare the 1969 JLH amplifier, had it been run from circa 2017 SMPS or the circa 1969 supply intended for it. One of those options would work passably and the other would sound awfully sour.
This matter doesn't relate to your more elaborate amplifier. Well, it does still relate, except that the decimal points have been relocated so far out that their effectiveness is inaudible.
If I could spot a flaw, I'd hazard an attempt at the maths. There's not room for that.
I could try to calculate when there's something to do, but there isn't.
As for calculating the remaining bit by hand, I wouldn't live long enough to do it.
This why I mentioned that the SMPS sense line needs to be extended to go the whole length of the dc umbilical cable. The negative feedback of the SMPS would then subtract the cable. I though they were supposed to do that. Seems that the newest SMPS have come up short on features.
P.S.
Might have missed it; but, what I was saying (no matter how fine the amplifier) is that one probably shouldn't make the audio amplifier's job more difficult. Last time I checked, power noise rejection is a wonderful feature that one hopes not to use.
I had done an incomplete job of the prior post. It had to do with assumptions of the average/basic audio amplifier, had it been run from SMPS.
My apologies that such a thing was not relevant to this thread, and not relevant to your amplifier.
Probably, concerns about SMPS power should have been on a different thread/page.
Compare the 1969 JLH amplifier, had it been run from circa 2017 SMPS or the circa 1969 supply intended for it. One of those options would work passably and the other would sound awfully sour.
This matter doesn't relate to your more elaborate amplifier. Well, it does still relate, except that the decimal points have been relocated so far out that their effectiveness is inaudible.
If I could spot a flaw, I'd hazard an attempt at the maths. There's not room for that.
I could try to calculate when there's something to do, but there isn't.
As for calculating the remaining bit by hand, I wouldn't live long enough to do it.
Right on!
This why I mentioned that the SMPS sense line needs to be extended to go the whole length of the dc umbilical cable. The negative feedback of the SMPS would then subtract the cable. I though they were supposed to do that. Seems that the newest SMPS have come up short on features.
P.S.
Might have missed it; but, what I was saying (no matter how fine the amplifier) is that one probably shouldn't make the audio amplifier's job more difficult. Last time I checked, power noise rejection is a wonderful feature that one hopes not to use.
Oh no! I messed up again!
That theory is outdated. It had an inherent flaw of being power supply board centrist, which is not the most effective locale for audio relevant filtering. Oops! Sorry. That (quoted) comment is just bad.
Seems that it takes me so long to learn things, that by the time I have managed, the news has changed.
Not long ago (months? one or two years?) most amplifiers could not nullify power supply differences.
Recently, my favorites are those that do.
I can confirm the ferrite bead part because I had actually tried some, no effect. What is the umbilical cord for?
Sent from my iPhone using Tapatalk
Power supply in a separate chassis from the amplifier. Bad idea on most counts.
I have pondered with the idea, not a clean design, but when you have different constraints, moving higher radiating EMI generating parts would be a good solution. But definitely a trade off. Recently looking at some converter options using PCB transformers seemed like a very small and clean solution, seemed so small that I wondered how they get the power rating they specify when using LT chips.
Sent from my iPhone using Tapatalk
Sent from my iPhone using Tapatalk
I suspect that we probably have a DC power cable located in-between the power board and amplifier board. If a filter had been added to that cable, the filter would be more effective on the amplifier end of the dc power cable. One example is FB43 ferrite beads (specifically, on the DC power cable, at the edge of the amplifier board). Another example is a (one) small (1u~5u) 250v cap across the rails, same locale.
Also, these are really very small, so it doesn't require an extra enclosure.
Such small filters do nothing to the audio directly. That is good news!
What they could do is block some higher frequency power noise so that it doesn't interfere with amplifier stability. The difference is still pretty small, but so is the cost.
The power noise rejection of Tom's amplifier is so good, that if you add a power filter, then it is likely that the added filter will seem to do nothing. That's the usual case when the amplifier didn't need the help.
The only split rail SMPS that I can find, are made for audio power amp usage. Those are usually lower cost and easier to use than a linear supply. I usually derate SMPS to 40% (up to 40% of current capacity used, and at least 60% in reserve), so that they'll last much longer. After that, it is not especially compact.
Also, these are really very small, so it doesn't require an extra enclosure.
Such small filters do nothing to the audio directly. That is good news!
What they could do is block some higher frequency power noise so that it doesn't interfere with amplifier stability. The difference is still pretty small, but so is the cost.
The power noise rejection of Tom's amplifier is so good, that if you add a power filter, then it is likely that the added filter will seem to do nothing. That's the usual case when the amplifier didn't need the help.
The only split rail SMPS that I can find, are made for audio power amp usage. Those are usually lower cost and easier to use than a linear supply. I usually derate SMPS to 40% (up to 40% of current capacity used, and at least 60% in reserve), so that they'll last much longer. After that, it is not especially compact.
Last edited:
Actually no they couldn't. The FB43 bead is good for the 40-200 MHz frequency range. That's beyond the self-resonant frequency of common 1-5 uF, 250 V caps, thus the filter won't provide any meaningful attenuation.
The filter also does nothing at the frequencies relevant for loop stability on the LM3886/Modulus-86/286.
Amplifier instability at RF is often caused by supply inductance and lack of proper decoupling. That goes for the LM3886 as well. Adding inductance in series with the supply (i.e. adding a ferrite bead) works directly against you as it basically turns your output stage into a Colpitts oscillator.
Should you wish to try this out in practice, build an LM3886-based amp and remove the local decoupling. Run this amp with >30 V peak swing into an 8 Ω load and watch the output voltage on an oscilloscope. I'll guarantee you that you'll get spurious oscillations.
The Modulus-86/286 are stable with more than 80 º phase margin if built according to the instructions. All the necessary supply filtering, decoupling, and stability measures are handled on the board. I'm really not sure what perceived problem you're trying to solve here.
Tom