LLC Resonant Converter for HiFi use?

[dch53]

I’m trying one of these (https://www.aliexpress.com/item/32921995271.html) LLC resonant converters in a class A valve amp I’m building. It provides

•   HT: 200V to 520V 300W maximum.
•   DC1 and DC2: 2.5V-6.5V adjustable 15W peak 7.5A.
•   DC3: fixed output 2.5V-15V, customizable, maximum 10W (not sure how it can be both fixed and customisable!)
•   DC4: -10V to -120V 50mA maximum

My requirements are well within the supply’s specs even with a 50% derating and it bench tests fine with the loads I need.

I’ve reverse-engineered a schematic for everything but the controller which is on a daughter-board with an opaque coating (so I can’t identify the control IC or assess the control loop). The schematic is very similar to those found in application notes and evaluation boards. The power supply is well made on a nicely laid-out 2-layer PCB with lots of vias.

The results have been very disappointing with significant noise at 100Hz and harmonics and strange “humps” in the response at around 1.5kHz, 3kHz, 4.5kHz etc, decreasing in amplitude with frequency.

The attached image shows the high voltage output of the supply (1kHz input signal to the amp).

The supply rectifies the mains and filters the result with 3 150uF capacitors. I’m guessing that the residual ripple is going straight through the resonant tank to the output. I was expecting the regulator to filter all that out.

I’m planning on trying a linear regulator between the mains filtering and the converter itself.

Meanwhile, can anyone tell me whether this is par for the course with these converters?

Are they not suitable for HiFi use in general or is it just this particular implementation?

[Faringdon]

If you are getting these noises in your audio then it could possibly be that the layout of the amp, the  pre-amp, and/or the smps is poor.
Also, it is not PFC'd so you will get high 100Hz current pulses into the circuitry...but good layout and cap decoupling etc would normally take care of that.
Yes your LLC is likely  poorly regulated.....

If i was you i would do a  PFC'd LLC converter.....and do nice decoupling etc........but if your amp circuitry is poorly layed out, then even a really good psu wont work well....all switchers are noisy.
But yes.....it sounds like you are seeing the noise from the mains input current and its harmonics...which for you is bad, because you have no PFC.

I mean, you could build your own...or buy a say, 60v output , 400W SMPS...and add a boost converter to its output to give you your high voltage output....what high voltage output do you need...presumably fixed at , say 200V would be fine?

[jonpaul]

Bonjour DCH53: Bravo for the excellent questions. Veteran of  high end hifi / cinema sound, (Sequerra  Dolby, THX) 1971..1990s.

0/ Rather than a spectrum, it is wiserr to post the schematic of the amp and PSU, as well as scope shot of the PARD/AC ripple/noise on the PSU 300V bus.

Use AC coupling and take precautions with scope or analyzer against B+ transients.


1/ Class A and B linear audio amps have zero  PS rail noise/PARD  rejection.

Thus a highly regulated and filtered supply is used ib all class  A audio.

2/ Most high end Class A amps use a linear PSU with choke/multiple section filters or linear regs.



2A/ I can see no reason to use a cheap poorly documented noisy chinese switcher.
Concentional transformer/rect/linear reg  mains trsf, caps, and regs for 250...1000V are readliy available or easily made.

3/ Suggest to  repurpose the Chinese switcher and study some lineal Class a valve amps.

4/ If you are serious about good sound, Many high end /audiophine websites and DIY forms will be glad to assist. Also ask our old friends Nelson Pass or Mark Levinson (Daniel HERTZ.CH)

Bon Chance


Jon

[dch53]

Thanks for your response.

Performance of the amp improves significantly with a high-voltage regulator between the LLC and the amp. The amp design itself is well-regarded and has been widely built in the diy community.

If the LLC is poorly regulated would that suggest issues with the design of the control loop? Design of these converters is very complex and I'm over-simplifying, but if the control loop was overly-damped the regulation would not respond quickly enough to eliminate the 100Hz and harmonics.

If I end up concluding the LLC is of no use as is I'll disassemble one (I have 2) to fill in the gaps of my reverse-engineered schematic and maybe improve it.

I've been considering building a PFC. There are plenty of design guides, spreadsheets and reference designs around for both PFCs and LLCs. The main problem is the magnetics.

[dch53]

Thanks for your reply.

The amp is a well regarded design built by many people in the diy community. I'm reluctant to post the schematic of the LLC resonant converter since it's a commercial product. I'll look at posting AC-coupled scope shots of the 400V rail. I already have some of the switching noise.

While I've built, repaired and modified amps and other HiFi equipment for myself and others I vowed I'd never build a Class A valve amp due the weight and cost.

Then I came across these power supplies which would eliminate quite a lot weight and some of the cost (no heavy mains transformer and banks of capacitors to get the HT good enough for an amp with no PSR).

I've built enough amps and the world of SMPSs and LLC resonant converters in particular has been a very interesting diversion.

[Faringdon]

yes you could do PFC followed by LLC for the 300W.
So what vout would you choose.?

The LLC is easy to design...basically, when you are at the resonant freq it just converts at the transformer turns ratio.
Attached is LLC in LTspice and the excel design file.


The PFC is also easy for you……BCM PFC is simple…..CCM Av mode PFC is also simple for you….you just start out by designing a boost for the mains pk = vin, and with your wanted vout…..but just remember in that case the actual power throughput is twice average when at mains peak

I'm reluctant to post the schematic of the LLC resonant converter since it's a commercial product.

OK But its likely just a bog standard LLC....There are schems all over the web, and its all knwon tech....most of the audio LLCs i saw werejust open loop.

[T3sl4co1l]

Was this measured under load, or what?

LLCs may run in burst mode at light load, which would explain the kHz peak.  With sidebands due to line frequency modulation, which clearly comes through.  Mains ripple may suggest low control loop gain.  Conversely, squealing could be bad compensation.

How much of that is due to EMI aliasing/rectifying into your ADC, I have no idea.  They did at least put one single CMC on there, and Y-cap between grounds.  I highly doubt it passes FCC/CE, let alone UL/CE safety.

Tim

[mtwieg]

From the pictures it looks like it's a single stage conversion (as opposed to a PFC stage followed by a DC-DC stage), so it's certainly going to have significant output ripple at twice the line frequency. There's a direct tradeoff between power factor and output ripple, dictated by conservation of energy, so not the power supply's fault.

LLC converters are somewhat limited in their transient response performance across wide line/load conditions. If you want a stiffly regulated output, you're better off with a fixed-frequency converter (forward, half bridge, etc).

[dch53]

Thanks for that. It would appear then that I'd be wasting my time trying to improve the regulation of the converter.

Pity, it includes all the voltages I need in such a small package.

I could precede it with a PFC. This is a Class A amplifier so I'm not so worried about load transients.

I'll also investigate fixed frequency converters.

[dch53]

Thanks Tim.

It was measured under a load which was around 20% of it's rating.

When I get a chance I'll get my oscilloscope out and check whether it's running in burst mode or not. I haven't been able to identify the controller so I don't know whether it's capable of it.

CMC? I suspect you're right about not passing FCC/CE or UL/CE safety.

Dave.

[T3sl4co1l]

(Common Mode Choke)

EMI depends on environment, and a lot of commercial power supplies fail when put in real equipment anyway*, but it's entirely possible they didn't test it at all.

*Typically, they're tested with a load resistor hanging off the output -- no common-ground capacitance, or direct grounding for that matter, only whatever antenna value the resistor (and its heatsink if applicable) contribute, which will only matter for radiated emissions.  So, you can get quite different conducted results depending on what you have connected -- how much (more stuff = bigger antenna = emissions at lower frequencies), or how it's grounded (maybe grounding the output "sausage effect"s more emissions up the mains).

Typically output ripple is also tested at a 20MHz bandwidth, on wire leads into a bypass capacitor.  So you might get unexpectedly high VHF-band noise from commercial supplies that otherwise seem reasonable.

Such a power supply might be filtered by putting it in a box with inlet/passthru filtering all around, but that's a lot of effort, and still doesn't address the safety issues.  (Issues which could at least be minimized if a grounded chassis is used -- this allows basic insulation between primary and secondary; and perhaps the box is itself fireproof, containing any flame eruption that might occur.  It still might not pass UL, but maybe it's enough peace of mind for a one-off.  Or you might inspect the internals, take it apart a bit; but at that point, has your supposed cost savings actually paid off anymore? Not really.)

And yeah, I know, it's easy for me to say "do it right".  I do this for a living, I could slap together and test a suitable (non-isolated) converter in a few days.  That's a lot harder if you don't know SMPS, and a whole solution is very attractive.

The problem is not necessarily even anything I've laid out here -- I'm only speculating.  And that's exactly the problem: you don't know, you can't know.  Even if the manufacturer went through the entire process and made a respectable product, the simple fact that the seller hasn't provided such information means it's legally useless.  The main thing in legal terms is, if they place the UL mark (or other NRTL, or a similar process for CE), either they're liable for infringement by placing it without license, or insurance will pay out in event of destruction caused by a recognized device.  If you have insurance, consider the rare case where the unit fails, massive damage results, and they refuse to pay out because unrecognized equipment was involved -- or even just present!

And, I don't see much if any route to sue the overseas manufacturer, or even the seller (and, mind, Ali most likely has strict immunity clauses in their terms, so you're almost certainly not suing them, you have to find the seller directly), so they don't have much reason to get approvals.

(Or equally well, most of this stuff shows up on Amazon too, and they just don't give a shit.  Hm, I forget if I've heard any news about electronics like this, power supplies, or if that was some years ago much of anything popped in the news, and if anything's changed since then. I'm sure there are active court cases on the linked matter and others, and, they can stretch cases out for years before having to take action or pay remediation.)

Tim

[dch53]

Thanks again Tim.

Given my entire tri-amped HiFi system is diy-built tube gear that insurance scenario is probably moot.

Since the schematic of the LLC Converter is bog-standard I thought I'd post it. The actual controller board is missing because it has an impervious shiny black coating. There's a connector labelled "J1" that goes to the little controller daughter-board on the primary-side schematic.

I haven't included the schematics for the other voltages. 3 of them use MP2482 buck converters. The -150V is a simple transistor regulator.

Perhaps someone can explain to me why there are 2 control loops.

[jonpaul]

https://www.ti.com/seclit/ml/slup263/slup263.pdf

https://www.st.com/en/applications/power-supplies-and-converters/llc-resonant-converter.html

Use of 2 optos allows jumper selection of output V on SEC side

Circled RC/Diodes are detecting peak V Cres

Jon

[Faringdon]

..yes the red bit in your post 2 above here is current sensing....ayk, its by way of i = cdv/dt

[dch53]

Use of 2 optos allows jumper selection of output V on SEC side

Thanks Jon. The jumper positions vary the voltage across the optocouplers. Why do 2 voltages need to be fed back to the controller?

[Faringdon]

There are three levels, Def, Dn and up.
These three different jumper selction combos, give the three  different opto currents, which get detected....the optos have different series resistors.
Most likely the controller will read the optos, and then adjust the switching frequency, so as to change the vout.

You could knock this circuit up with a infineon ICE2H LLC controller...you can set it up to give three different frequencies as you require.

Your circuit has inherent leakage inductance in each transformer..(same for each)..you could of course, make it easier for yourself and just interleave wind them and have an external "leakage" inductor.

Shovel some overvoltage protection on.....and as you know, HV DC is most most dangerous.

Shove a opto  feedback loop on it if you want, and get whatever vout you want.....external error amp with variable reference...and you could get rid of that 100Hz ripple.

Boost magnetics is also very easy for you.....just get a PQ or ETD core of the right size.....do the saturation calc with i(sat) = B.A.N/L
Remember to gap it.
L(nH) = N^2 * AL

LLC woudl also be easy for you...if you give me your spec, ill send you all the magnetics and winding instrucs and all.
...plus schem with ICE2H...and assistance with layout if you need it...
Ill do you a PFC with  a ti PFC controller (same mode as LT1248).....UCC3818.
..Plus LTspice sim to start it off.

Youll have it done in no time....just be mega aware that HVDC output....though i believe you know that...DC is not like AC...DC is a real killer.

..other way to do it is just get say a 60v >300w offline psu, and shove a cascaded boost at the output..to give you your x hundred volts at 300w

[dch53]

Thanks for the explanation. My question is more around why 2 optocoupler control loops are required rather than one. Maybe the controller is using the difference between them.

Thanks also for the magnetics offer.

I have read every LLC Resonant Controller tutorial, app note and data sheet I can find over the last few months. In this case I think the controller is a 6599 or 6599A. The current sensing components are right out of the data sheet https://au.mouser.com/datasheet/2/389/l6599a-1849566.pdf.

The units I'm working with have 3 problems
- The output voltage goes up and down with the mains. It doesn't regulate.
- The 100hz and harmonics. It doesn't regulate.
- Switching noise

The switching noise is so great I can't see the 100Hz and harmonics on my oscilloscope. Maybe the switching noise is due to the parasitic inductance around the length of the gate traces/TO220 MOSFET Leads.

I've attached some scope shots but I need to do the same for the 2nd unit in the amp in case I did something removing and replacing the controller PCB on the unit I used for the scope shots.

At this point I have some decisions to make:
- try and improve the 2 LLC resonant supplies I have
- abandon the whole SMPS idea and shoe-horn in a linear supply.
- design and build a custom SMPS (PFC + switching supply)

The third option is a significantly longer term one and I'd really like to have the amp in use.

[jonpaul]

DCH53:

Design and debug of any SMPS is best done with a lot of experience and the proeer test amd safety equip.

Reading your long thread its unlikely you can accomplish your objectv without additional study and experience.

The scope shots reveal a lot of deep issues.

In case you seek to make a class A or B hifi/dtudio amplifier with triamp X over (three PAs) best option:

 A single Linear PSU , feeds all 3 amps, most use toroidal mains trsf for low hum. Size and weight of the other PA parts, heatsinks, etc will be comprable to the toroid weight.

You can DIY this easily.

You will find DIY of resonant SMPS much more challanging.

Jon

[dch53]

Thanks Jon.

Based on everything I've read I have no doubt that I'd "find DIY of resonant SMPS much more challanging". And it would be an iterative process over quite some time.

I also take your point about test equipment. I'm limited to multimeters, a cheap oscilloscope and PC-based hardware and software for audio frequency testing.

I've built or repaired a lot of valve equipment for myself and other people going up to 550V. The SE amp I'm building is intended for the HF range in my system (>1kHz). I've already had the metalwork CNC machined.

What I'll probably do is shoehorn in a linear supply and come back to switching power supplies. It would be a very interesting challenge.

Dave.

[T3sl4co1l]

Yeah, sticking to Hammond or whatever for now is a safe bet; it's heavy, and lacks features (from efficiency to PFC to regulation), but damned if it isn't simple, and reliable.  Making boost/buck converters (common ground, not isolated) for specific loads is a lot easier, and a good project to potentially turn into solutions.

Tim

[dch53]

Indeed. Although I've been using toroidal mains and output transformers from Poland for a few years now.  Smaller external field, physically smaller and lighter for the same rating and have better regulation than EI transformers.

[jonpaul]

DCH53

if your goal is high end hifi and great sound the audiophiles and ausiophools seem to avoid any SMPS and use linear PS for amps, preamps.

The HF and   and noise and ripple need a lot of post reg, filtering of the SMPS.

Pliotron, Hammond may others make fine shieded mains toroids.

To lear about SMPS, LLC, resonant, see the many paprs, books, app notes.

Then get an IC mfg eval board to try out a typical reference design.

Jon

[dch53]

Thanks again Jon.

The unit in the amp does measure much better that the one on the bench although still not good enough obviously.

I'm not a purist. Cheap boost/buck converters are used regularly in valve amps for things like heater supplies. Switch mode power supplies are used in some top-end solid-state gear too.

I've actually been looking at evaluation boards. Infineon ICE2HS01G and ST 6599A are freely available. Infineon have a Design example with 400V from 30V too. onsemi and TI have them too.

As per earlier advice I'll look at fixed-frequency converters (forward, half-bridge etc) too.

Regards, Dave.

[mtwieg]

I'm not a purist. Cheap boost/buck converters are used regularly in valve amps for things like heater supplies. Switch mode power supplies are used in some top-end solid-state gear too.

I've actually been looking at evaluation boards. Infineon ICE2HS01G and ST 6599A are freely available. Infineon have a Design example with 400V from 30V too. onsemi and TI have them too.

As per earlier advice I'll look at fixed-frequency converters (forward, half-bridge etc) too.

Absolutely agree that there's no reason a SMPS can't meet the needs of hi-fi audio. But I do want to caution you that a SMPS implementing PFC is going to have 100/120Hz ripple on the output (of the PFC stage, that is). This results from conservation of energy, and has nothing to do with the quality of the design or components. And as others above have mentioned, single-ended open-loop amplifiers topologies are going to pass that ripple straight out.

So if you use a SMPS, make sure either:
1. It doesn't implement PFC, and just has a big bridge rectifier and filter cap on the input. If the controller is good it will provide a smooth output.
2. If it does implement PFC, you use a second DC-DC regulator (either SM or linear) to reject the ripple from the PFC output. Most isolated PSUs actually will have have separate PFC and DC-DC stages built-in. Sometimes the DC-DC stage is an LLC too, which is fine in most cases. I think the issue with the PSU in your first post is that it's only an LLC stage, possibly doing PFC as well.

[Faringdon]

sorry dch53, i didnt have time to look for long time at your scopes.....i didnt see what they were of.....but i saw the dreaded common mode noise...so you can reduce this by using non-dangly-ground clip type scopeing......do you have isolation transformer by the way?.....basically, get rid of loops in the flying probe wires etc......good pcb layotu will make the smps easier to scope aswell.....did you buy  that unit, or you layed it out?
...i always make up my own probe for eg scoping the gates of lower fets.......just like a coax type 1x probe..........are you having like a long ground wire to which you connect for all readings, and this gives a wide loop?..and hance noise causing.....try to take meaurements with point and gnd of scope close together if poss.

Sorry i cant address your points properly here.........i am still workign on a different project at the mo...but if you do send your spec...and can do you  pretty well all-the-way schem for REGULATED LLC following a boost pfc...and sim too...and it will only take me an hour or so to do the whole lot and the mags...and its always good to keep hand in so i enjoy to do that for you...but maybe you have other plans i dont know..its your call...i am not charging anything
What diff probe do you use.?

When you are showing those T=40ns vibes then you are swamped in the luvverly common  mode noise........you have floated yoru scope?...that sometimes makes common mode worse.

coax x1 probe can be made by buying a RF Coax cable assembly of say 1metre with BNC plug end...then just cut it up/
........or...just cut the head off a  cheap scope probe.....and pull out the inner wire and  unwrap the gauze shiled and make a little conductor out of it....then with perhaps a twisted pair flying probe...solder in to your nodes..and probe away.

Common mode noise and scope readings often comes up here on the forum...so if you search you will find.
Your vout is 450V?...if so, avoid forwards and bridges....stick to llc....the forward will need very high voltage diode...and suffer ringing...LLC suffers no ringing of the output diode if done right....so LLC you are nice with say a 500V diode for a 450V rail.

But  you may also see interest in the cascaded boost, as you say...to give you 450v from 48V say.