Looking for Custom SMPS designer

[Potato Cakes]

Hi, Hucky Bridge,

This is for pro audio applications, specifically analog Class A preamplifiers. Space is a key factor as these will be need to be able to fit in a 1U chassis. Ideally it would be not take up more space than 3"x6".

I does need to be a universal power supply. Again this is for microphone preamplifiers so the output noise has to be low.

I don't know how to give a good response for the economics side. Obviously, the desire it have the cost be low and the length of service before needing to be replaced to be very long. I'm not sure how quantify that in terms of actual numbers that would be reasonable.

The timeline would be in three months time, depending on what the design costs would be.

Thanks!

Paul

 

[JohnRoberts]

have you looked at purchasing a commercial power supply module? That way you will get agency safety approval too.

JR

 

[Potato Cakes]

Maybe you should stick to linear power supplies. It's what the community likes and approves. SMPS and DC-DC converters are bad, they will "definitely" inject unwanted noise and have no place in audio stuff.

As for me, I have yet to encounter one that does that, so I will continue to use them.

I installed a linear power supply in this unit with the audio portion being at one side of the chassis and the transformer/PSU on the other side. MUCH noisier, especially in the lower frequencies, than the one SMPS configured described at the beginning of the this thread.

Thanks!

Paul

 

[Potato Cakes]

have you looked at purchasing a commercial power supply module? That way you will get agency safety approval too.

JR

I take it you mean an external PSU? For the particular project I am working on the PSU needs to be in the unit.

Thanks!

Paul

 

[Potato Cakes]

I've been working on this issue some more and found that these SMPS for whatever reason provide exceptionally low noise in the Class A preamps I'm trying to power:

https://czcl-powersupply.en.made-in...ini-Single-Output-Switching-Power-Supply.html
However, the manufacturer stopped making them three years ago. I happen to have two of them when I was buying all sorts of SMPS units when I began this insane process of trying to figure out how to make these things work in some of my custom projects. This is the replacement model per the manufacturer:

https://www.circuitspecialists.com/24volt-35W-enclosed-switching-power-supply-a-35fal-24.html
I have not tried the newer ones yet but have some on order. I did place two of the HTS-35B-24's in a client's single preamp that built for him and the noise floor was about -110dB, some places lower depending on the frequency. All I did was place a 2200uF cap between the +/-VDC connections (power supplies wired together for bipolar as described above). I did not have to add a bunch of filtering at the input or at the output like I had tried on other supplies per manufacturers' recommendations which didn't help with noise. These SMPS units are also cheaper than the medical grade and the other "better" models that I tried. I ordered what I thought was more of them from the same place but it was a different manufacturer (Hengfu) even though the power ratings and dimensions were the same. These produced high levels of noise centered around 500Hz. I even tried a similar MeanWell unit which had smaller dimensions and the noise was lower but still centering around 500Hz. I don't know what is different with the HTS-35B-24 units that makes them less noisy. There is some particular rating I'm not seeing which is the key to finding off the shelf, low noise SMPS's. Also of note, I could hear buzzing coming from both the Hengfu and MeanWell units. I have not heard an SMPS do this before.

Even with finding what is easily the best SMPS that I have found for the audio circuit I am using, the problem is that two of them in the same chassis as the audio circuit takes up too much space, so I am still looking for a custom solution.

I did get a price from another designer for a schematic and prototype. It's a reasonable amount but more than I can do at this point. In the meantime, I will have to keep trying to figure out a solution.

Thanks!

Paul

 

[JohnRoberts]

I haven't searched lately but recall seeing PS modules that could be mounted inside chassis from Digikey. Not as cheap as rolling your own, but comes with agency approvals. You may need to figure on adding additional filtering.

JR

 

[Potato Cakes]

I haven't searched lately but recall seeing PS modules that could be mounted inside chassis from Digikey. Not as cheap as rolling your own, but comes with agency approvals. You may need to figure on adding additional filtering.

JR

I'll search there during lunch. I built a capacitance multiplier to test but I screwed it up because I get 0V on the output.

Thanks!

Paul

 

[Potato Cakes]

The modules I found over there are indeed a bit expensive. I wouldn't mind paying $80 - $150 if I knew for certain that I could drop it in and everything would work with the noise floor being what I expect.

Thanks!

Paul

 

[JohnRoberts]

search here for discussions about noise in off the shelf switching PS. I would anticipate adding some filtering.

The price is OK for building 1x of something, for 100x design something but agency approval typically costs tens of $k.

JR

 

[gswan]

Hello, Everyone,

I'm reposting here for the same thing. I've hit a brick wall trying to get figure out how to get the SMPS's I want to use to be quiet for pro audio applications. I was in discussions with a designer that would do the circuit but I have not heard from him in some time. I keep running into the same solutions when searching the internet some of which I've tried and have not worked and others I do not quite understand.

To reiterate from previous posts on other threads, the one thing that has worked with these particular PSUs is using a capacitor that is well beyond the recommended value from the manufacturer. This causes the PSU to power supply several times and then it becomes stable and everything works and the circuit is dead quiet. This obvious is not how this is to be done this method doesn't work with other similar circuits with more current draw.

Right now, I do not have time or bandwidth solve this on my own. I am on tour, trying to get a studio up and running, and I have a number of tube amps and other custom studio electronics to build. I have the PSUs I wish to use, I just need a circuit that properly filters/smooths the DC output. Any additional regulation I can do myself.

With all of that being said, I am looking for someone who is experienced in working SMPS's and getting them to work noise free in analog pro audio circuits. I can adapt the circuit to suit the various layout needs. I just need a tested schematic designed that will provide repeatable success.

Thanks!

Paul

For audio applications you need to look at resonant mode SMPS, specifically ZVS type. These have far less noise output as they fire a pule when the voltage is zero (ie no voltage transients) into a resonant tank circuit, so the energy is transferred through the transformer as sine waves at a high frequency instead of square waves with lots of harmonics.

I use a lot of ConnexElectronic SMPS for class-D power amplifiers I build. Their power supplied are well made and have never had any failures after over a decade in continuous use in many boxes. They will also customise the voltage rails to some degree.

 

[thor.zmt]

Hello, everyone,

I'm looking for someone experienced in designing SMPS's to provide a schematic for a simple yet reliable/robust circuit to be used in audio applications. It is essentially two power supplies that will be on one PCB, one that is +48V/1A (approximate current) and one that is +36V/500mA (also approximate current). Each voltage has to have a separate V- reference, hence two separate supplies. What I'm after is a bipolar +/-24V (+24, 0V, -24V) that has a minimum of 620mA per rail.

Normally I would figure this out on my own and ask questions along the way, but a potential client is wanting to purchase a prototype of mine and the time frame for building one is about one month. I have been using separate off the self PSUs but eventually I was going to need a single board PSU before I could release a commercial product if there was enough interest.

Ideally this would be taking two existing layouts for the different voltages and combining them. But since I do not have any layouts nor do I know enough about SMPS design to understand the various schematics found on the interwebs combined with the time factor I requiring myself to seek assistance. And this has to be solid without the need for testing or tweaking, so that also eliminates me from being the guy that designs the circuit.

I ultimately just need a schematic and a BOM. I have a local PCB layout guy that I use, but if you do already reliable layouts that can be combined as mentioned above then that would be helpful.

I do not know what this service would cost, so that would need to sorted upfront. Please email me at paul(at)scodovasound.com if you are experienced with SMPS design and if this would be of interest.

Thanks!

Paul

Commercial SMPS tend to not be designed for Audio.

There are a number of challenges in Audio that do not apply to a PSU that is meant to drive LED lighting or charge phones.

Then there are material challenges meeting agency requirements.

The cost for a competent, easily manufacturable design that covers all requirements is likely not to be cost effective, unless you plan manufacturing run's in 10's of 1,000's.

An alternative would be to use a generic OTS module (say 48V/1A), test and characterise it and then use additional external circuitry to deal with any issues and rail-split the single supply to +/-24V.

Or use a 24V/2A OTS design and use a separate switcher to create an inverted -24V supply.

Another alternative would be to use double-conversion where a commercial OTS SMPS provides (say) 12V/4A (can be external plastic Brick) provides a fully agency compliant safe low voltage and a secondary low voltage and a high frequency (1MHz & up) switcher to produce the required voltages (e.g. +/-24V for audio and +48V for Phantom power), which will have low audio band noise (generally comparable to middle of the road LDO's) and where any switching noise can be easily attenuated using small size inductors and capacitors.

I have previously implemented all of these options in audio products.

I have also designed "super low noise" Audio SMPS (commercially produced) with 1uV audio band noise and very low mains leakage by using a specialised transformer design, which are OEM'ed by a general SMPS manufacturer, however these have 10kU MOQ.

We can discuss options more if you are interested.

Who am I IRL?

https://www.linkedin.com/in/thorsten-loesch-studio-raumklangg/

Thor

 

[RogerAF]

I've played around with SMPS for the last decade. There are several chips that are available from TI, ADI, etc. TI now owns Burr-Brown and ADI owns Linear Devices and Maxim. All of these have chips for SMPS designs. The data sheets often have a recommended PCB layout. (I have heard remarks that some of these suck, but haven't researched that myself.) I'm still using DIPs on solderless breadboards.
One thing I've read, which you might try, is to have the SMPS output a higher voltage, then use a linear regulator to get your desired output. The advantage is that the SMPS can be designed to match the LDO's voltage drop, so it doesn't have to dissipate a lot of heat. If you use the LDO following the SMPS, you can get a low EMI and lower noise floor than the SMPS can deliver. You therefor gain the high efficiency of the SMPS and the low noise of the LDO. I found a presentation at TI that may be of interest: https://training.ti.com/ldo-or-switcherthat-question
You haven't specified the input voltage, so I'll assume you want to plug into 120VAC. Perhaps an external AC to DC converter would be worth looking into.
PM me with as many specs as you can including size restrictions and I'll look into what is available. If you then want to discuss actual design and PCB layout, we can negotiate suitable compensation. There are PCB and assembly companies who will do prototypes, so you don't have to do a run of 10,000.

 

[JohnRoberts]

One thing I've read, which you might try, is to have the SMPS output a higher voltage, then use a linear regulator to get your desired output. The advantage is that the SMPS can be designed to match the LDO's voltage drop, so it doesn't have to dissipate a lot of heat. If you use the LDO following the SMPS, you can get a low EMI and lower noise floor than the SMPS can deliver. You therefor gain the high efficiency of the SMPS and the low noise of the LDO. .

I did pretty much that in a DSP rack unit about 20 years ago... I don't recall if I used a LDO or just a normal 3 terminal regulator after the switcher. In that same unit I made a cap doubler/tripler to make phantom voltage for it's one mic input using the HF clock from the switcher, with small signals diodes and tiny SMD caps.

JR

 

[RogerAF]

I've mostly worked on linear PS, for consoles and power amps, etc. But when I came to Florida in 2007, I applied for a job at what turned out to be the repair department for Vari-Lights. I had no experience with SMPS, so didn't get the job, but the manager took me through the shop and explained how they took the 120/240VAC, rectified it, boosted it up to between 400 and 500VDC, then bucked it down to whatever was required for the motors. The reason they boosted so high was to trade current for voltage and thus be able to use smaller wires (but the EMI must have been horrible within the cases, although I guess it doesn't matter much up on the trusses).
So this fascinated me and I started playing around with SMPS on my breadboards. I have in mind a SMPS for Tubes that can deliver a clean 300VDC. But it's kinda dropped by the wayside as my apartment is already too full of guitars and recording gear. I have enough room on my table for my computer, scope, sig-gen, and breadboard, but not so much for fabrication anymore. I used to use CircuitMaker and TraxMaker, but they were bought out by Altium or the other big one--pardon my brain fog--now they are online, but I like to keep my files to myself and prefer my software on my own computer. Like the song says "...Get off my cloud!"
I found KiCad a couple of years ago and I really like it--being free, it's well within my budget. There is criticism that the learning curve is steep, but I've found that most EDA software is basically the same. I just prefer that the lines on the schematic are straight with 90° or 45° turns--not jogging up and down before and after the component which I'd carefully lined up with the pin of something else. Anyway, KiCad makes nice drawings and has a BIG database of components, produces BOMs and PCB layouts and gerber files. (I haven't explored those much as of yet, just a little bit of the PCB layouts--still have to help the autorouter sort out the rat's nest. But there may be things I don't know about.) There is a large online presence with lots of help and lots of models for those parts that aren't in the BIG database.

 

[Potato Cakes]

I have made a number of discoveries regarding this topic as it relates to a couple of other topics I've posted here. I will make a new thread with all my findings later this week as it pertains to this and the related audio circuit.

Thanks!

Paul

 

[RogerAF]

Up all night, with Google and TI, ADI, LT, etc. etc. There are several possibilities.
1. TI has a very inexpensive chip (that you can get in a PDIP for $0.755 each! Yes, that is 76 cents!): https://www.ti.com/product/MC33063A
2. ADI has an LT chip: https://www.analog.com/en/products/lt8582.html#product-overview which is rather pricey and only available in surface mount (but teeny tiny). It is a dual output device that can do + & - tracking. ($11.300 each from ADI Buy Now. $12.+ from distributor.) But it still requires external components.
3. Then for LDOs I found these, but TI probably makes some too. https://www.analog.com/media/en/technical-documentation/data-sheets/lt3080.pdf
https://www.analog.com/media/en/technical-documentation/data-sheets/3015fb.pdfThere are also going to be tradeoffs. More current, lower switching freq, bigger inductors. But, better noise, ripple, and EMI. But, more SPACE.
Higher switching freq = smaller inductors and caps, so less space, but also more ripple, etc. etc.
Anyway I've shot my wad at this for 8 hours or so. I hope it helps.
BTW this doesn't include the switcher from the mains AC to whatever DC is required by the inputs of any of these. Just a cheery thought to look at your wallet.
Happy New Year!

 

[thor.zmt]

1. TI has a very inexpensive chip (that you can get in a PDIP for $0.755 each! Yes, that is 76 cents!): https://www.ti.com/product/MC33063A

The MC33063A/MC34063A is about as old as it gets. It also is limited to 100kHz maximum switching frequency.

That said, it is really interesting for DIY.

First, it's a chunky Chip, not a tiny VQFN with thermal flag that needs a serious SMD rework setup to solder.

Second, it's a "universal" Chip, with 40V maximum in/out. It can be configured as Buck (Step-Down), Boost (Step-Up) and Inverting Buck-Boost (Inverting) and additionally, by adding a cascode transistor (Mosfet preferred) to the output voltage can be made considerably higher than the official 40V limit.

Third, many internal connections for modern IC's are brought out and accessible, allowing many interesting configurations, including switching frequency synchronisation, including even with sample rates. Imagine a Tube Microphone Preamp with analogue out but with wordclock in... Yup.

Fourth, 100kHz is a bit close to the audio range, but LC filtering at 100kHz can be very competent. Panasonic (previously Sanyo) Os-Con Capacitors were specifically designed for this kind of Switcher and are readily available from Mouser. Again, these parts are more DIY friendly than the kind of SMD components I tend to use for my 1MHz+ switchers.

Extensive LC filtering is a must for switchers, or switching frequencies break through.

Fifth, the requirements on instrumentation to work successfully with this is much less. Even a 192kHz/24Bit Audio interface may be used as virtual oscilloscope and FFT at this speed.

Also, a post regulator, I like LM317/337 extended with NJM5534 (or 5532) as this gets sub 1uV noise out (10Hz-100kHz) and are short circuit/over temperature protected.

A switching pre-regulator can be made tracking to minimise thermals. This can mean interesting non-datasheet approaches to the feedback loop, separating AC and DC components.

This is rather "old-fashioned", but as said, much more amenable to real DIY.

Using modern parts is for PCB Design and BYO from a full service PCB/PCBA prototyping house.

Thor

 

[RogerAF]

First, it's a chunky Chip, not a tiny VQFN with thermal flag that needs a serious SMD rework setup to solder.

I like that it also comes in an 8-pin PDIP, so I can play with it on my breadboards. Of course everything is more noisy on a breadboard, but component values can be worked out. I have a decent LCR meter that tests up to at 100KHz, so I wind my own coils whenever possible. There's also a surplus electronics store near me that has LOTs of parts. I can usually find something close to what I need. I looked at Coilcraft's products, and for anything in the 100 to 200 uH range, that can handle at least 1A at a 20°C rise, the size approaches a cubic ½" or more.

post regulator, I like LM317/337 extended with NJM5534 (or 5532) as this gets sub 1uV noise out (10Hz-100kHz) and are short circuit/over temperature protected.

I thought that using LDOs following the switcher would be advantageous over the LM3x7s, as they need 3V of headroom. However, looking over the data sheet for ADI's LT3081, the noise is about 27µV (10 Hz to 100KHz). If this can be reduced to 1µV, how great is that!!??? But please explain how you'd do this with a +/-24V supply using a chip with a +/-22V maximum? And if you can do this with the LM3x7s, can it also be done with the LT3081 and LT3091 (18µV (10 Hz to 100KHz))?

The MC33063A/MC34063A is about as old as it gets. It also is limited to 100kHz maximum switching frequency.

I didn't even think of the age or switching freq when I found the MC33063A. First off, I was looking for things that were in stock. A whole lot of the newer chips were listed as 0 in inventory. So that was first criteria. Second was NO BGAs! And third was can it output up to at least +/-28V. That search took a while. Then I was also looking at what ADI had available. (They have a whole lot, but again, something that puts out higher voltage at sufficient current, and is in stock, has become challenging to find.)
Potato Cakes say he needs 0.6A x 2, so I figure at least 2A for so me thermal headroom. The MC33063A can do up to 1.5A on its internal transistors, so I figured to do that. The difference between Buck and Inverting configurations looks like just swapping the inductor and diode on the same layout pattern. He also says there is about 3"x 6" for the entire supply; and I haven't even looked into how to knock the 120VAC mains down to 30VDC. It seems like a passive ½ wave rectifier followed by a LC filter would probably be sufficient. I don't know that there really needs to be another SMPS on the board. But hey, I've been wrong before (that's mostly how I learn.;-)

many internal connections for modern IC's are brought out and accessible, allowing many interesting configurations, including switching frequency synchronisation,

I didn't see anything in the data or app sheet on this, but I may have overlooked it in my skimming. Can you explain please? That's probably going to be important when using two of these guys.
A note on my search of ADI's LDOs: There don't seem to be any in stock! And they cost $8+ for singles. Maybe the LM3x7s and 5534s are a better option. Wow!

 

[thor.zmt]

I did get a price from another designer for a schematic and prototype. It's a reasonable amount but more than I can do at this point. In the meantime, I will have to keep trying to figure out a solution.

My suggestion would be to use a OTS 12V DC "brick" supply to provide isolation and agency compliance. This could be placed internally but makes more sense off-board. Make sure you get a rating of 400% of continuous power consumption to ensure the SMPS survives a reasonable time span.

Using a "special" mains cable (with build in additional X/Y Mains filtering and a "safe" earth lifter - see "Hum X") can get rid of the pesky primary side problems.

Use a small internal module using MC33063A / MC34063A to produce the required local voltages. They are chunky, PCB's easily laid out.

It is easy to produce almost any voltage and current if using external transistors (or Mosfet's).

Here a commercial solution from one of my products:



The small "Mezzanine" PCB on the right with the two DC Jack's and the many small inductors produces 48V * 2 (Tube HT plus amplifier frontend) and dual +/-18V with > 1A per Channel.

There is actually "triple conversion" here. A 24V max input 8A step-down converter produced a 6V DC Bus (including tube heaters etc.) and the +/-18V & +48V are produced from this 6V bus. The "input rating" is 9V - 18V @ > 60VA, with design center target 12V DC.

Of course, these are "MHz Class" switchers, using MC33063A et all would be much larger. Extra LC filtering is on the main PACB together with big electrolytic capacitors.

To prevent switching noise from escaping grounds go via high current, low DCR Ferrite beads and all supply voltages have inductors in the lines (the row of small indictors to the right) and the PCB is designed with "islands" for "dirty" (switching) sections and clean (DC out) sections with all "bridges" formed by inductors or ferrites.

Thor

 

[thor.zmt]

I thought that using LDOs following the switcher would be advantageous over the LM3x7s, as they need 3V of headroom.

Yes, but are we REALLY concerned about this? Most modern LDO's are in tiny SMD packages.

However, looking over the data sheet for ADI's LT3081, the noise is about 27µV (10 Hz to 100KHz). If this can be reduced to 1µV, how great is that!!??? But please explain how you'd do this with a +/-24V supply using a chip with a +/-22V maximum? And if you can do this with the LM3x7s, can it also be done with the LT3081 and LT3091 (18µV (10 Hz to 100KHz))?

The way to do this is to use the 5534 single rail, 0V/+44V DC or 0V/-44V DC.

The trick is to use the LM317/LM337 in effect as "intelligent transistor" in a classic "Super-regulator", that is a simple Op-Amp + Pass Transistor (replaced by 317/337) design.

Reference Voltage produced from LM334 CCS & LM329 Voltage reference.

Multiple 317/337 can be paralleled to provide more current/power. Or the external current boost transistor trick can be applied.

The 317/337 can also simplify the circuit as they can eliminate startup circuits normally required if the Super Regulator is supplied from the output of the regulator and no protection circuitry is needed.

Frequency Compensation is of course mandatory as the 317 is a lot "slower" than a typical NPN or PNP TO-220 Transistor.

I didn't even think of the age or switching freq when I found the MC33063A. First off, I was looking for things that were in stock. A whole lot of the newer chips were listed as 0 in inventory. So that was first criteria. Second was NO BGAs! And third was can it output up to at least +/-28V. That search took a while. Then I was also looking at what ADI had available. (They have a whole lot, but again, something that puts out higher voltage at sufficient current, and is in stock, has become challenging to find.)
Potato Cakes say he needs 0.6A x 2, so I figure at least 2A for so me thermal headroom. The MC33063A can do up to 1.5A on its internal transistors, so I figured to do that.

Easy to add external Mosfets.

Use P-Channel for stepdowns and inverting configurations and N-Channel for step-up.

With a high current step-up using an external FET you can easily make a supply with +/-24V & 48V using a single MC33063A and external switch and one big inductor, using charge pumps for the rest, with just capacitors and diodes.

I didn't see anything in the data or app sheet on this, but I may have overlooked it in my skimming. Can you explain please?

For example, the oscillator is brought out. This way many options are possible with an external synchronisation circuit. Just weak coupling of multiple 33063/34063 (place a single larger capacitor in the "ground" side of all timing capacitors combined can make sure beat frequencies between multiple converters are avoided.

A lot of this can be simulated in Tina-TI (which I find more intuitive to use than LT-Spice etc.) which is another DIY advantage.

That's probably going to be important when using two of these guys.
A note on my search of ADI's LDOs: There don't seem to be any in stock! And they cost $8+ for singles. Maybe the LM3x7s and 5534s are a better option. Wow!

There is a lot of life in the old dog's. I routinely use Op-Amp's (no external pass transistors) as "regulators" for low power and noise sensitive consumers, like Clock's, DAC's and ADC's.

The 4nV|/Hz from a XX5534 or XX5532 is hard to beat by any LDO and adding external pass devices gives us near unlimited current with up to 44V low noise output voltage (more if we power the 5534 from the regulator out using a Zenner Diode in the PSU line to drop off a few volt.

So a 5534 with series zenner in the supply and a 317 with series transistor to increase voltage handling totally can make super low noise 48V Phantom power at many 100mA to cover for example a big mixer or a 16 in Channel recording interface. Who needs 16 phantom powered mic's? I do... Hehehe.

Thor