700W ZVS switchmode converter

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all

I've started this thread because interest has been shown in ZVS converters in threads about other stuff. The circuit I've posted is that for the the main converter for my PowerDAC 1 digital amplifier, and was designed about 5 years ago.

The converter is designed to run on the non-isolated +385V-400V dc bus from a PFC preconverter. Design minimum was 350V. A non-isolated +18V supervisory supply referred to COM, is required
The converter uses a full bridge with ZVS switching to deliver 75V 700W at a measured 91% efficiency, using an ETD39 coreset, gapped. Switching frequency is 176.4kHz and there is provision for locking this to an external sync signal through an isolated input. Another isolated input accepts a shutdown signal, which can be fed to a suitable PFC shutdown input on the same side of the isolation barrier.

There is a cycle by cycle current limit, and conventional voltage mode control is used. The three phat Transzorbs are to limit voltage buildup when a downstream adjustable synchronous regulator in digital amp has it's output voltage reduced rapidly.

Some suggestions for improvement: There is no independent OverVoltage Protection mechanism (OVP) and this should be added in the interests of safety. The ML4818 is rather long in the tooth now, and better, less power-hungry chips exist, such as the UCC3895. It would also be better to use current mode control and remove the transformer's primary coupling capacitor.
Another efficiency-boosting addition would be the use of MOSFETs as synchronous rectifiers in a current doubling topology as advocated by Unitrode's Lazlo Baloch in several appnotes from TI.

I've often wondered on the feasibility of using a (distributed gap) big MPP toroid for the main transformer. Losses may be too high, but it may be worth doing a few sums, as the gapped ETD39 is likely to generate considerable EMI.

I would hope that this circuit may be of interest to members and might stimulate discussion on improvements and possibly be evolved with group effort to a really neat SMPS PSU. The circuit posted has been in use for about 4-5 years with no failures.

NOTE!! This circuit, although isolated, relies on construction of an SMPS transformer to achieve this isolation. Anyone who hasn't constructed offline SMPS transformers before, or is not familiar with creepage and clearance distances and other safety aspects found in UL1950, should not attempt to build something like this. There are high voltagees involved and great care must be taken.


John H


  • mainconzvs.pdf
    43.7 KB · Views: 1,503
John, thanks a lot for sharing such an experience. After reviewing your design and ML4818 datasheet (I wasn't familiar to it before), I would ask you some questions, more as DIYer than engineer :)
From the first sight it seems not easy to get how zero-voltage switching is organized.
Is it about that we adjust time delay between upper and lower side FETs on-state, so called deadtime - 250ns specified on your schematic, and also transformer inductance value, which deals with pulse rise/fall time together with capacitances in full bridge? Is it that why magnetizing/stray inductances also specified on schematic? If I understand it right, then supposedly I need to adjust deadtime for any new similar design to get ZVS operation, due to different FET type/transformer, etc., right?
Can you please explain why we need here gapped ETD core for transformer. Anyway core operation seem to be like in PWM mode, from the transformer's "point of view", and there is decoupling capacitor C10. Or is it just to get such quite low inductance value to allow such high-frequency ZV switching. Please clarify this.
And regarding your idea to replace ETD39 for MPP toroid.
I've had not much experience about using toroid transformers for mains power supplies, just made a couple of forward type 300-400W converters with toroid transformer. There's been everything well about principal operation, but terrible problem with HF noise and switching spikes on output DC rails. I traced that problem to transformer primary-to-secondary capacitance. After replacing transformer with ETD type, having single turn of copper foil shield between primary and secondary, I had solved the problem. Later, I was not quite successful about making some shield winding on toroid transformer. Also efficiency was about the same with toroid and ETD transformer, so I gave up using toroids for mains converters.
And one more question. Is it good idea to move driver part to secondary side, together with all logic signals and its supply (as we use transformers to drive gates anyway), and thus omit transfer of feedback analog signal thru exotic double-photosensor optocoupler?

ZVS operation

Hi Alme:

Yes, I reckon you just about got it! The clearest reference I could find was MicoLinear Application Note 19, and even that
I had to read several times. Another good one is 'Designing a Phase Shifted Zero Voltage Transition (ZVT) Power Converter' by Bill Andreycak of TI/Unitrode. Then there's 'Design Review: 500W 40W/in3 Phase Shifted ZVT Power Converter', also by Bill Andreycak. I had the good fortune of attending a seminar presented by Bill today, as it happens. These appnotes have no identifying numbers on them, and I think they're from the Unitrode Power Supply Design Seminar series, which is on the TI site; like looking for a needle in a haystack,I know.

As I recall, the dead time is adjusted equal to one quarter of the period of the resonant frequency of the FET's Cds and the L(leakage) of the transformer. I set this up on a scope.
The gapped core is to control the L(magnetising) of the transformer to achieve a desired rate of rise of current. (Gapping doesn't significantly affect L(leakage)). The ZVS requires a certain minimum current, either from the load or from a deliberately high magnetising current. If the I(mag)
is too low, then there's a significant minimum load requirement on the converter for proper ZVS operation.

I have found my transformer design notes for the ZVS. They're in pencil on paper (proper stuff, eh, no computers req'd). Over the next few days I will scan them in and try make a pdf file of them,
so I can post them on the forum.


John Hope
Thanks, John; thanks, Jaka. Things become clear step by step :)

Another question: how do I know transformer stray inductance value before actual transformer making? It doesn't seem apt to calculation :confused:
And if so then what is easy way to measure it and distinguish from magnetizing inductance (which is maybe easy to measure) ? Before closer look at ZVS, I never had the need to define it...

calculating stray or leakage inductance is difficult, but measurement is easy. Just measure with shorted secondary. When you measure with open secondary, you are actually measuring sum of leakage and magnetizing inductance which are series connected. By shortcircuiting secondary you are actually bypassing magnetizing inductance and all that is left is leakage inductance.

Best regards,

Jaka Racman
Resonant parameters

Jaka: Snap, you took the words right out of my mouth!
Alme: This is one of those 'chicken and the egg things'. You don't actually know the L(leakage) until you've wound the transformer. You know what ballpark it will be: 5-50uH is typical. But you don't need the L(leakage) to be known to design the transformer, so you can do this first and then calc the resonant parameters afterwards. If you have to mod the transformer the L(leakage) might change a bit, but you can accomodate this with the dead time.

I've posted my design calcs for the resonant parameters of this converter; they were done after the xfmr was finished. These formulae were taken directly from ML appnote no. 19. What's interesting is that even with a L(mag) as low as 250uH, you still need a minimum load to guarantee ZVS action. You can't decrease L(mag) too much (I wouldn't go below 250uH) because you will then have high primary currents flowing back and forth that will increase copper losses and make your trafo run hot. But a minimum load isn't such a big deal; in the given design the quiescent power drawn by the amplifier circuitry was such that the 'deliberate' load resistor I needed to make up the required minimum load was only a couple of watts. It may not be enough to guarantee full 100% ZVS action, but even 'near ZVS' switching is far preferable to hard switching.

Although I previously said that it would be an improvement to go to current mode control, I'm no longer so convinced of this. CM is at a disadvantage at 0 load and very light loads because the current feedback signal corresponding to the very low currents at low loads tends to get swamped by noise. Audio amps in a domestic environment spend a lot of time at light load and it might be wiser to stick with the voltage mode control if this is the target application.

I totally ruined the family kettle testing this powerr supply, but that's for another forum.

I will post the transformer design details soon.


John H


  • resparms.zip
    77 KB · Views: 690
ZVS transformer design

Here's file 2.

Some postscrips to the attachments:

1. During development I found it difficult to fit all the windings AND screens comfortably on the bobbin, and my final transfomer design was modified to use 81/0.1118mm Litz wire for the secondaries. (I somehow acquired a big roll of this. . .) Using this in practice resulted in a higher measured temperature rise, but still acceptable (35C as measured on the trafo outer winding tape, if I recall)

2. The design notes are based on 2mm spacing tape against each wall of the bobbin, giving a 4mm creepage in total. (For 400Vdc primary bus and with secondary earthed to mains earth. In my experience this has generally satisfied CE safety agencies, but UL can be more sticky and require 5mm which would mean 2.5mm tape on each side. (Squish!!). Even as a DIYer, you should adhere to proper safety guidelines when constructing offline supplies. :eek:

3. Gapping was done by inserting a small fibreglass shim of about 0.8mm between each of the outer leg faces of the ETD core. One has to push quite hard to get the core clamps to 'click' into position. (More Squish!)


John Hope


  • xfmrdsn2.zip
    87.2 KB · Views: 494
Gate driver transformers

Alme: The FETs used are quite PHAT (I like TO-247's the most) and could easily handle up to 1kW or more. I chose them largely for their Rds(on). The transformer would have to be redesigned for powers > 700W and it would then be wise to use the slightly bigger ETD49 core.
I'm still keen on considering a PHAT MPP toroid, despite your misgivings over this - I'll do the sums and see what falls out when I get a few quiet hours. I suspect the core losses will be too high at B=(+/-)100mT, so this is the first thing to check.

A note on the gate drive transformers: I used 3F3 material toroids 14mm OD x 9mm high because I found myself with a big bag of these. But more easily availably toroids 12mm x 5/6 mm would do fine in 3F3 or 3C90 material.

The gate driver transformer windings consisted of a trifilar triplet of 600V teflon insulated 30AWG stranded wire, twisted 2 turns in 25mm. The teflon wire (from RS/Franell) is an absolute slippery pig to work with, but I was apprehensive (Great Fear of BANG :eek: :bawling: :mad: ) to use commonly available and more easily workable Kynar insulated wire of the sort used for wire-wrapping. I'd read somewhere that this stuff can't handle much more than 60V, but on the other hand, the 500W ZVS converter in Bill Andreycak's Unitrode appnotes uses this. (Perhaps Bill is braver than I am or there are better types of Kynar wire.) Maybe you have some useful experience of wire for GD trafo's you could share? For this sort of thing it's very useful to have different colours for each winding.


John H
Hi, John,
I also use teflon insulation wire for base/gate drive transformers, or sometimes plain PCV insulation wire (for it's easily available in different colors as you mention).
They say that usual enamel insulation wire withstands 500V between turns; I can confirm that enamel wire 1mm in diameter surely withstands 300V between neighbor windings for I used it a lot in transformers (including automotive converters, where one must varnish transformer before inserting, for significant vibrations in there can deteriorate insulation after some time) - and this is without any additional film insulation. But I haven't tried to use Kynar type wire for drive transformers.
I got interested in soft switching techniques last times, because feel bored to struggle EMI in every hard-switching power supply. Now that I went thru your paper notes and thru schematic again, would you please answer my next questions.
Is it of principle that I use phase-shift modulation driver for ZVS converter? I would like to try it first with simple half-bridge and cheap driver like SG3524/3525, and because of this driver, some lower frequency operation, about 100kHz, having available deadtime 500-600ns. Looks like nothing prevents from such simplification.
From your schematic it seems that you use 1:200 ratio for current transformer (CT). Is it so and what core type for CT you suggest for 500-700W converter?

Greetings Alex, and thanks for the feedback on GD trafos. My understanding is that a phase shifted full bridge is needed for this particular kind of ZVS. You'll see from the page of current flow diagrams in the Microlinear appnote that you need a special sequence of overlapped switching ops to support the ZVS, and you can't get these from a standard half bridge.

But ZVS is simply Zero Voltage Switching. If you can devise another topology which only switches the FETs when the voltage across them is zero, then you have ZVS.

The current transformer is 100T on a 3F3 toroid 10mm. (These things are such a pig to wind, I would know if it had been 200 turns!). For 8.3A current limit, the CT secondary current is 8.3/100, and this develops almost exactly 1V across the 12 ohm resistor.

I started design of an improved version some time back in which I dispensed with the CT and used instead a resistive sensor in the return between the lower FETs sources (tied together) and COM. I haven't tried this, but it is an option. Personally I don't like the idea that the FETs sourcves aren't tied directly to COM.

Another really neat option - if you're not too fussed about costs - is to use a GMR hall effect sensor IC spanning a pcb trace. These sensors are made by NVE (www.nve.com) They have a bandwidth in the low MHz range and can be arranged for bipolar or monopolar current.


John H
John, now i see my mistake about calculating current transformer.
Sometimes spending much time to get simple things...
And, during my searching for your mentioned Microlinear appnote #19, I also found another soft-switching converter, utilizing PFC in single stage.
I like that idea too :)
After reviewing and studying schematics as yours and the like what can be found,
I want to try something similar soft-switching by myself :cool:
Isolated full boost converter

Alex: Thanks for that, much appreciated. It looks like a very exciting concept. I will study it in detail once I've energised my brain with enough coffee.

Unitrode brought out an isolated PFC/converter chip a few years ago that they were very excited about, but it was really cr**. They used push-pull with overlapping drive, necessitating 1200V IGBTs switching at a pathetic 75kHz into what appeared to be quite a complex transformer. Then after all that, it had efficiency roughly the same as any conventional PFC feeding an isolated converter, but was bulkier and heavier due to a massive secondary capacitance.


John H
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.