Wurth recently released their WE-LLCR transformer series for LLC resonant SMPSes. These are now in stock at Mouser at quite reasonable prices. Options are 12, 24, and 48V at 150, 200, and 250W. Dual secondaries, single 16 or 18V aux winding (usually 18), 600+100 to 400+65uH.
I've been thinking to build a small resonant dual supply suitable for chipamps (+-15V 150W, give or take) for some time as Connexelectronic's SMPS300RE's lowest available output is +-20V. The WE-LLCRs are the LLC trafos I've come across available in small quantities at cost effective pricing; beats buying a bobbin and winding it oneself. So I'd be curious to hear folks' thoughts on which resonant controller and feedback scheme they'd use with the 12V 150W 760895431 and why. It's not like there's exactly a shortage of options anymore (scroll to bottom; note: unfortunately itacoil doesn't do small orders or have much in the way of distributors).
I'm considering the L6599A Connexelectronic uses on the SMPS300RE (see here) as a default. However, I've not yet done the analysis to see how it interacts with the trafo and a combined Vcc+ground+Vee sense.
I've been thinking to build a small resonant dual supply suitable for chipamps (+-15V 150W, give or take) for some time as Connexelectronic's SMPS300RE's lowest available output is +-20V. The WE-LLCRs are the LLC trafos I've come across available in small quantities at cost effective pricing; beats buying a bobbin and winding it oneself. So I'd be curious to hear folks' thoughts on which resonant controller and feedback scheme they'd use with the 12V 150W 760895431 and why. It's not like there's exactly a shortage of options anymore (scroll to bottom; note: unfortunately itacoil doesn't do small orders or have much in the way of distributors).
I'm considering the L6599A Connexelectronic uses on the SMPS300RE (see here) as a default. However, I've not yet done the analysis to see how it interacts with the trafo and a combined Vcc+ground+Vee sense.
Last edited:
Hi twest820, thanks for the Mouser links.
couldn't you just modify the Connex yerself changing a few R & Cs without changing the 20V xformer, maybe take a hit in max power and eff.?
What chipamps are you using that can't take more than +/-15V?
edit > IMO, rather than starting from scratch and new design, yer probably better off taking that one apart and removing a turn on the secondary.
custom outputs? I bet they only use 2 or 3 different XFMR part # just changing feedback loop R.
couldn't you just modify the Connex yerself changing a few R & Cs without changing the 20V xformer, maybe take a hit in max power and eff.?
What chipamps are you using that can't take more than +/-15V?
edit > IMO, rather than starting from scratch and new design, yer probably better off taking that one apart and removing a turn on the secondary.
custom outputs? I bet they only use 2 or 3 different XFMR part # just changing feedback loop R.
Last edited:
So far as design motivation goes, it happens +-15V rails are adequate for about 75% of DIYers. It happens I fall into this category. There's substantial leeway in specifying thermals but, given reasonable conservatism, rails in this range are attractive for keeping junction temperatures below min thermal shutdown thresholds when multiple class AB chipamps share a heatsink. Mainly useful for triamp and other multiway applications, though there's some benefit for stereo and biamp. Linear implementations with trafos around 12V also work. But using an SMPS eases managing mains variations and power factor.
As an aside, modding is probably unnecessary as turnkey solutions should be available on request. The discussion here should be of interest for those looking for such.
As an aside, modding is probably unnecessary as turnkey solutions should be available on request. The discussion here should be of interest for those looking for such.
Last edited:
LOL
OR said more directly > most chip amps use rails in the range of what PCBstuff / conelectric already sells straight off the boat.
even class B stuff doesn't get hot on non bass channels,. so really no need to specify multiples of different power rails when bi/tri-amping.
edit> interesting grounding scheme on PCBstuff SMPS. I wonder if they will ever submit for safety agency approvals.
OR said more directly > most chip amps use rails in the range of what PCBstuff / conelectric already sells straight off the boat.
even class B stuff doesn't get hot on non bass channels,. so really no need to specify multiples of different power rails when bi/tri-amping.
edit> interesting grounding scheme on PCBstuff SMPS. I wonder if they will ever submit for safety agency approvals.
Last edited:
SMPS transformers are a bit of a black art.
I found this out when designing an LLC SMPS transformer.
I had trouble getting the leakage inductance right and in the end resorted to a separate leakage inductor.
I also fell foul of the skin effect. I just wound thick wire on the transformer and it failed miserably.
I found this out when designing an LLC SMPS transformer.
I had trouble getting the leakage inductance right and in the end resorted to a separate leakage inductor.
I also fell foul of the skin effect. I just wound thick wire on the transformer and it failed miserably.
yeah extra cored inductor to meet L/C resonance goal is huge fail in the low cost arena that is commercial SMPS. In my 1st SMPS 100KHz fixed freq., I used hand made Litz wire. I think I calculated 100% skin depth at 200KHz was 27AWG.
edit> I like that ST L6599A part
with the bootstrapped high side driver makes for one of the lowest parts count design.
Last edited:
Yah, I'd pretty much decided I didn't want to mess about with winding my own series LLC some time back. If I hadn't your recent threads probably would have done the job. Taking turns off a commercially wound trafo, less of an issue, as at least things should be set up right to start with. Prior to the WE-LLCRs showing up at Mouser the lowest cost option I could find was Wurth's prototype winding service, which was around 100 euro per trafo with a minimum order of five.
Care to be more specific? I can't tell if you're referring to the Y caps, 10 ohm ground impedance, ground lift option, or something else.
Share your design analysis, please.
As an example, consider a triamp with, nominally, a 4 ohm tweeter and 8 ohm mid and woofer. Within the passbands the effective impedances are something like 3, 6, and 6 ohms, give or take depending on driver to driver variation, variation in program material power spectral density, and so on. To make this a fairly minimal dissipation case, assume the chimpamps used current limit at 3.5A and the power demands of all three channels are equal. At the onset of clipping with a +-20V supply dissipation in the chipamp driving the tweeter is about 30W RMS, give or take a few depending on exactly which formula one's using. The corresponding dissipation in the mid and woofer is 16W, again give or take some, so the total power delivered to the heatsink is 60W. Take Tambient to be 40C due to, say, a warm day, inclement amp placement such as in the sun, theta_chassis_ambient > 0 C/W, or what have you. Then steady state operation of the amp at rated power puts the heatsink at 100+C if it's operating at a fairly decent 1 C/W. Chipamp theta_jc is typically 1 C/W in nonisolated packages, plus something like 0.1 to 0.2 C/W for the insulator and thermal grease. That puts the tweeter's chipamp junction temperature at 135C. So, if one assumes a 10% variation in the usual 150C thermal shutdown trigger---max and min junction temperatures are not speced but 10% is probably realistic---then the tweeter is operating right at the edge of thermal shutdown at chip's min limit. The chipamps handling mid and bass are a bit cooler at 115-120C though, as you allude to, the bass junction temperature will tend to track peak rather than RMS power, so it'll probably get up around 135C as well.
One could take Tambient to be lower. One could take Tambient to be higher. One could assume a tighter thermal shutdown tolerance. One could assume a looser one. Point is any margin which exists is minimal and the design is reliant on the amp not actually being operated at rated power distribution to avoid the chipamps going into thermal shutdown. Given a typical loudness war recording has about a 3dB crest factor it's probably fair to say the junctions would run closer to 100C in this particular example. So some margin probably exists in practice unless the program material is pathological.
However, cases such as all drivers being 4 ohms, less even power distributions across channels, all channels running full range, chipamps with more typical 4 to 6A current limits, and/or supplies above +-20V, absorb much of this margin or exceed it at certain corners. One can use large heatsinks to get below 1 C/W in natural convection, one can make them more independent so they run cooler, one can use forced air, one can put the amp in the fridge to reduce Tambient, and so on. But I am most curious as to how one guarantees junction temperatures comfortably below 150C across a typical range of use cases with three or four chipamps sharing a heatsink of moderate size cooled by natural convection and operating at power levels which make use of the output swing enabled by rails at +-20V or higher. In particular, take moderate size to be a segment of extrusion four or five inches wide and six inches long. Optimal orientation for convection can be assumed.
yes program power density is a big factor, the next biggest is not using drivers at 4 ohms esp the tweeter. ( look at the chip amps data sheet, limited SOA it's far easier to deliver voltage not current ) It's a given bi / tri amp system design is best approached from a speakers system POV IE top down. A speakers bass driver is usually run 4 to 6 dB hotter than the tweeter even if they have the same sensitivity. The smaller the bass driver the more likely the sensitivity of the mid-bass driver is a few dB lower. The bass driver is taking a lot more power right off the bat, up to 8-10 dB more for the ubiquitous 6/7 inch mid bass dome tweeter combo. thro in the music's spectral density and the balance leans against the woofers amp, heat wise, much much more.
It's simple choice to use the same power supply rails for both chip amps otherwise you run into much greater cost and complexity for the power supplies. wouldn't it be better to choose the woofer at 4 ohm and the tweeter at 8 ohms, doing the reverse is kind of a system designers big fail> although tying his hands a logical guy would probably pad the tweet down to 8 ohms straight away rather than add another power supply. LOL
It's simple choice to use the same power supply rails for both chip amps otherwise you run into much greater cost and complexity for the power supplies. wouldn't it be better to choose the woofer at 4 ohm and the tweeter at 8 ohms, doing the reverse is kind of a system designers big fail> although tying his hands a logical guy would probably pad the tweet down to 8 ohms straight away rather than add another power supply. LOL
grounding PE, primary ground and secondary's connections in SMPS is complex, relates to just about everything.
I get very concerned when I see multiple connections to PE ground esp. in regards to EMI conducted noise. connections using capacitors without common mode inductors inject significant noise energy into chassis PE. The closer the connection to the switches (primary and secondary ) the more problematic. This is a system level design problem for the integrator ( IE final safety or FCC approvals ).
I get very concerned when I see multiple connections to PE ground esp. in regards to EMI conducted noise. connections using capacitors without common mode inductors inject significant noise energy into chassis PE. The closer the connection to the switches (primary and secondary ) the more problematic. This is a system level design problem for the integrator ( IE final safety or FCC approvals ).
Eh? The SMPS300R and 500R schematics both show mains inlet chokes. It's unclear to me from the schematics how much PFC may be built into the LLC drive but, given ST's reference designs for the L6599A and L6599AT all include a separate PFC controller, I would suspect the probable lack of PFC is a larger issue for certification.
Sure, special case thermal solutions may exist for driver selection in specific powered monitor or speaker kinds of cases. Most DIY is admissible to that since the amp to speaker ratio is usually close to 1:1. However, I don't see any reason to eschew the increased driver selection latitude reducing dissipation by removing unused headroom on the rails affords.
If you're interested in low parts count Fairchild's FSFR1700HS and similar integrate the switches. ON's NCL30051 integrates the PFC, albeit with some optimizations for lighting which limit its audio use to class A. It'd be handy if there were more such parts, though for this design iteration it's separate PFC and resonant control. I'm thinking not to employ synchronous rectification. Since typical home audio use is in the lower part of even a small supply like this's capability it's probably a wash or slightly more efficient not to use a synchronous primary. The secondaries are high enough voltage a Schottky bridge is fine and synchronous rectification pretty much overkill.
Sure, special case thermal solutions may exist for driver selection in specific powered monitor or speaker kinds of cases. Most DIY is admissible to that since the amp to speaker ratio is usually close to 1:1. However, I don't see any reason to eschew the increased driver selection latitude reducing dissipation by removing unused headroom on the rails affords.
If you're interested in low parts count Fairchild's FSFR1700HS and similar integrate the switches. ON's NCL30051 integrates the PFC, albeit with some optimizations for lighting which limit its audio use to class A. It'd be handy if there were more such parts, though for this design iteration it's separate PFC and resonant control. I'm thinking not to employ synchronous rectification. Since typical home audio use is in the lower part of even a small supply like this's capability it's probably a wash or slightly more efficient not to use a synchronous primary. The secondaries are high enough voltage a Schottky bridge is fine and synchronous rectification pretty much overkill.
an example of an active speaker crossover
Behringer B2031A best selling active 2 way monitor
AFAIK it has bridged chip amps for the 8" woofer and a single TI - National LM4701 for the 1" dome. The back panel shows LF room compensation for up to 6 dB boost to the woofer, this what is sometimes referred to as "baffle diffraction loss" in speaker design language. This is implemented as a cut in the active filter section.
image of power output VS load impedance ( most National devices will be similar showing optimum loads peaked for 8 ohm drivers. )
Behringer B2031A best selling active 2 way monitor
AFAIK it has bridged chip amps for the 8" woofer and a single TI - National LM4701 for the 1" dome. The back panel shows LF room compensation for up to 6 dB boost to the woofer, this what is sometimes referred to as "baffle diffraction loss" in speaker design language. This is implemented as a cut in the active filter section.
image of power output VS load impedance ( most National devices will be similar showing optimum loads peaked for 8 ohm drivers. )
Attachments
Disagree about the PFC at least for the USA. yes I see a choke but Y caps are on the noisy side shunting any noise on the primary side, (remember the common mode filter is directional here it's not working as an internal noise SMPS noise attenuator ) also other connections to PE exist without much series impedance. IMO they don't care about common mode noise! just like the PC industry. LOL
Last edited:
what does this mean in English. maybe you should review audio gear LOL
audio headroom is a good thing in DIY land. saving a few watts on a tweeter, invest in solar panels perhaps.
edit> I gave a concrete example of how to specify audio power for an active speaker and an actual product that comes close to that theory. look at it closer! your analysis seems inside out from where I sit.
Last edited:
nxp has some nice design tools for PFC LLC supplies up to 500W. these include the winding procedures and core selection of the transformers. quite handy and instructive.
besides the chips have a low part count ..
besides the chips have a low part count ..
thanks for the helpful links
Fairchild's part certainly is low parts count with both power FETs onboard.
not sure how the "independent" SMPS OEMs would take to it, I could see a lot of resistance without the ability to upgrade/ downgrade the switches and source alternate controllers. given they often end up in a price war or the race to the bottom with their competition. In my experience as a small competitor for TV tuners in set top boxes , Most of the chip products from the real big boys like NXP and ST often have whole turn key solutions for TVs and set top boxes that punish OEMs from straying too far away (mixing and matching parts) from their part portfolios. when it came to pricing, they ended up practically giving away the tuners cuz our customers were locked in on their video decoders and processors.
Last edited:
Yes, I came across those not long after NXP launched them. Haven't revisited since (still working through other controller suppliers as DIY time's been limited lately) but I'll get there.
Considering the design motivations are more or less opposite of the position you're advocating this seems unsurprising. 😉 Between prior amp and supply builds, oscilloscopes, audio interfaces, and DAWs I'm not exactly lacking in the peak voltage assessment department. If it makes you feel better some of the amps I've run for years do just fine on +-5V. And they're single ended.
Mayhap, mayhap not. My instinct is to think the same way you're doing but one doesn't have to look at many reference designs to find both bridge and mains side placement of inlet Y caps. There's insufficient data in the documentation to determine if the placements are optimization or oversight. What seems most useful would be comparative measurements, or at least detailed sims, but I'm not having any luck finding such in the literature. Though every paper I've come across which looks at Y caps to secondary ground shows reduced EMI provided resonances are controlled.
Connex does choose a placement opposite what ST uses in their L6599A(T) reference designs. That might imply it's less than optimal. Or perhaps it indicates the board influences the optimum more than controller and switch selection does.
Interesting. Thanks for sharing.
- Status
- Not open for further replies.