Hey, Svart! Big switcher needed

[bcarso]

The class D amp will be something somewhat special, but not intractable. IR have high side drivers that will handle 600V for instance. They are not as fast and low jitter as their fancy full-bandwidth audio-oriented ones, but we don't need that for 60Hz sines.

Once you get the drive circuit together and select the HV FETs and output filter, the PWM need not be very complex: a triangle wave generator and a comparator, or a free-running feedback oscillator around the whole shooting match with if you like a sync injector. The high/low driver chips will help you steer clear of Fire Emitting replacing Field Effect more or less out of the gate (no pun intended, and why I am on a sporting event kick tonight I can only guess).

 

[Svart]

I've used the IR2110 on a few things, that would likely be a good start for the HS and LS drivers for the FETs even though it's a little slower. Besides, I have a few laying around the lab.

What do you think about semi-discrete quantization of the schematic? Would you rather try something like this, an osc/comparator or a ready made ic?

The next question is, would you like to drive the output from a full bridge or from a half bridge? It seems that a full bridge adds complexity but solves a few problems, i.e. rail pumping.

I use the sg/uc3525 for a lot of projects. It's super easy to use and less than 1$. it has two channel outputs that can each run 50% duty or they can be summed for 100% duty(which we don't need here.) We could more or less drive an IR2110 directly from the outputs which in turn would drive the FETs.

 

[Svart]

I've found that the IR2110 will be quite capable of what we are looking at doing. Something i have found is that the VCC of the IR2110 will need to be -120V+VDD or roughly -120v+15v=-105v. yet another rail to create. We could do this with some Rs and zeners fairly easy unless you have a trick up your sleeve...

Edit: some quick sims show this working

www.theopiumdenproductions.com/PWMinverter.pdf

 

[bcarso]

The rail-pumping issue is an important one for class D amps with LF response down to 20Hz and fairly small reservoir caps, true. I suspect for 60Hz transformer primaries it becomes less significant. I guess though, if someone puts a half-wave rectified load on this thing we don't want it to get upset.

Bridge mode does solve some problems, especially for high-quality audio, since the power supply noise tends to be common-mode and rejected to first order by the load. But learning how to correct for pumping in the PWM is probably a good thing to learn, too.

Where is the semi-discrete quantization stuff---in that IR slideshow? I must have breezed past it (I have been insanely preoccupied and it's probably out of the proverbial frying pan and into the tokomak now).

 

[Svart]

Take a look at page 50 of 54, the schematic of the IR demo/reference board. after the level shifter and before the FET gate drivers.

TC7WH04FU and TC7WH08FU
triple inverter and an AND gate array.

Interesting little circuit that i have never seen before.

 

[Svart]

www.theopiumdenproductions.com/PWMinverterbridge.pdf

? :guinness:

Bcarso, do you think they'll get their latest Tokamak to work?

 

[PRR]

> full bridge or from a half bridge?

One side of the load has to be tied to Aussie dirt.

My gut reaction is that half-bridge is more natural for that.

I suppose if the 50KHz isolation transformer really isolates, it makes little difference (except at the debugging bench).

 

[Svart]

Although full bridge would be more elegant, I'm not so sure the additional parts count and the potential problems for timing of the bridge halves are worth the extra effort. The parts count is going to be fairly high from rough estimates because the PWM system will need to source it's own power rails from the switcher's outputs. this means some form of regulation and additional caps/zeners/Rs or LDO reg. we can't let the PWM ic go into UVLO or let the VS of the PWM IC rise beyond the rated inputs for the FET drivers (+18v). We also need to think about some kind of feedback for the PWM duty and some kind of SD feedback for instant fault situations.

 

[PRR]

Low-tech illustration.

I spent an hour cutting and pasting a schematic, foolishly not saving a partial copy, and the backhoe met the city power line. Power was out just long enough to lose the image.

So back to ol reliable yello pad and pencil.....

 

[Svart]

revised schemo, half bridge.

www.theopiumdenproductions.com/PWMinverter.pdf

the 180-0-180 is fed from the first stage of switcher which would be almost a normal switcher with improved specs and duty for supplying 360v of swing at a lower duty cycle. we should not allow the switcher to supply more than half of it's rated load before it reaches 50% duty. in other words this thing should never have to supply more than 70% duty to give a full output or else we run into trouble and will be damned to PSU hell.

I'll try to cobble together a switching front end in the next few days. I think current mode and normal feedback will be sufficient for this unless you see something that I need to address.

EDIT: and some of the values in that schemo are just place holders, not actual values.

 

[Svart]

Bump. been too busy with work to work on this any today, but i don't want this to die.

:thumb:

 

[Larrchild]

Hey PRR.
Would a simple saturable reactor in series with the guitar amp primary work to lower the rms value sufficiently to unheat the core while providing full peak voltage?

 

[SonsOfThunder]

Sorry I'm late to the thread... a couple of comments that I hope will help/shed some light.

Not surprised that you cannot find a (good) single chip class-D controller solution. There is one: HIP4080 and I don't recommend it for anything HV or making more than about 500W. Even then, use a socket for easy replacement (20-pin DIP). :evil:

I beg to differ on the opinion that rail-pumping is only a real problem for 20Hz and "fairly small reservoir caps". Of course it gets worse down that low, but you would be surprised how MUCH capacitance it would take to *completely* overcome this problem even at 40Hz.

ALL class-D amps I am aware of (in car audio >200W) have an output filter, 2nd or 3rd order.

Hope that is helpful info.
Charlie

 

[bcarso]

The rail pumping is a serious concern if the load is pretty reactive, and I guess most lightly loading mains supplies are going to be. And you are right, 60Hz is not that far away from 20. There are other ways to deal with it besides going full bridge (for example making the power supplies efficiently sink current and return it to the source). But full bridge does solve a lot of problems, even if it does use a bunch more silicon. And it's essential for the open-loop "digital" modulator designs (and still not enough at that, IMO!).

Carver had a lot of trouble with rail pumping in his amps with the buck converters tracking the signal and a class AB stage in the middle. Some of the corrections for it are the subject of some later patents, not particularly good patents IMO.

A system with multiple amps for a A/V receiver I was just starting work on when I left Harman had one possible version where I was going to have the help of a very good power supply designer, and he did a preliminary design where energy pumped back into a rail ended up being returned to the main cap on the other side of the PFC regulator. Never did see the light of day, but it probably would have worked.

 

[mr coffee]

Not my field of expertise, but this link might be pertinent to what you guys are up to if you aren't already familiar with Don Lancaster's magic sinewave stuff.

http://www.tinaja.com/glib/mschips.pdf

As you already noted, you really don't need all that class D audio bandwidth for this app, and output filtering requirements can be substantial in traditional class D audio amps.

If you do your transformer isolation from the incoming mains in the high frequency AC-to-DC step-down conversion, keeping transformer cost down, and then run a magic sinewave dc-to-AC inverter at 60 hz for your output, you would need a LOT less filtering than usual class D if I understand it correctly, and you could then ground one side of the 117vac line output to earth. Meets PRRs specs if I understand them correctly.

Those guitar amp power trannies tended to be under-spec'ed and run hot to start with.

 

[bcarso]

Thanks mr coffee. I had forgotten about Lancaster's thing.

I'm not sure it's really the "billion dollar" biz he suggests but it is quite entertaining. It reminds me a little bit of another maverick, Henning Harmuth, and his championing of Walsh functions* for just about everything under the sun. I got his Transmission of Information by Orthogonal Functions years ago and marveled over it, although never found a specific application.

Sounds as if Lancaster's approach might be about to flourish though, for things like motor control. Can't see the audio use as yet but who knows. Perhaps if the pulse sequence is calculated on the fly and maybe combined with some amount of multilevel switching (extensions of the pseudosine three-level inverters) it might become a more general signal transform technique.




*see for example http://mathworld.wolfram.com/WalshFunction.html

 

[mr coffee]

I'm not sure it's really the "billion dollar" biz he suggests but it is quite entertaining.

Don's enthusiam for his idea and his entertainment value aside (ever read his case against patents piece? :wink: ), do I understand the app correctly and that

If you do your transformer isolation from the incoming mains in the high frequency AC-to-DC step-down conversion, keeping transformer cost down, and then run a magic sinewave dc-to-AC inverter at 60 hz for your output, you would need a LOT less filtering than usual class D if I understand it correctly, and you could then ground one side of the 117vac line output to earth.

would be an economical way to implement it?

I'm thinking bridge output driven with 180 degree out-of-phase magic sinewave signals, and earthing one output. Since it's floating relative to the mains from the initial conversion, it wouldn't be a problem. A simple uP could generate the signals with a lookup table and control both output stages consisting of relatively slow high-current (and therfore cheaper) switching transistors with only minimal filtering and no feedback to drive the guitar amp power transformer. Wouldn't any supply rail pumping from the inductive load then be addressed by the DC rail regulation in the AC-to-DC stepdown swithing regulator that buck converts the 240 VAC to DC?

I would think that could be done relatively cheaply and you wouldn't have to be a switching supply guru to design a HF transformer to do the down-converting and line isolation. I don't know if it could be made for under $100 like PRR was hoping, but it shouldn't be much more, even in small volume, with a 350-400 watt ouput. The soft-start circuitry in the AC-to-DC converter would be easier on the guitar amp power supply caps and filaments too (add that to the feature list for marketing :wink: )

And it could be frequency "tuned" for a sweeter hard-pushed hum-injection sound in the uP driver (add that to the feature list for marketing, too :wink: ) for the deluxe model. :green: Shoot, the way those guys love to fool around with variacs and all sorts of stuff to tweak tone, PRR may have a hot product on his hands

FWIW. :guinness:

 

[bcarso]

The issue with pumping is not solved by invoking somebody's AC-DC stepdown regulator, whether a switcher or even a linear---they typically do not have two-quadrant operation, i.e., they don't sink (much) current, only source it. Ones that do are rara avis. But it's not an issue anyway with full bridge operation as you are suggesting.

I have a feeling that, despite the galvanic isolation in the regulator ahead, EMC issues around common-mode noise will be a problem with the asymmetrical arrangement you suggest, that is, grounding one side of the output. Possibly a better approach would be to use a center-tapped d.c. supply and have 60VAC either side, a la the balanced power currently in vogue in studio apps.

But all in all, clearly an appealing way to get efficient monchromatic AC. IGBTs would be happy as clams as the switching elements, even doing 230VAC.

I've still got the nagging suspicion that there is a clever more direct way to do this with a kind of direct frequency changer approach, maybe even involving some of the Lancaster ideas. But the margin of my brain is too small to contain it at the moment... :grin:

 

[PRR]

> 60VAC either side, a la the balanced power currently in vogue in studio apps.

No. Many vintage guitar amps have a 2-pin power plug and a Ground Switch, which ties a 0.047uFd cap from chassis to either side of the power line. Ideally this has been replaced with a proper 3-pin plug, ground switch cap removed; but many vintage amps have not been updated to modern safety standards. So ideally the player sets the ground switch to tie chassis to the groundED side of the wall output. If the wall outlet is Balanced Power, there may be danger either way. 60V beats 115V, but still a bad idea.

The outlet should conform to best US practice of the 1960s. It should have a true groundING U-hole, but should also have one side of the power-holes tied to supply ground. Anything else violates unstated and maybe unsuspected ASSumptions.

 

[bcarso]

[quote author="PRR"]> 60VAC either side, a la the balanced power currently in vogue in studio apps.

No. Many vintage guitar amps have a 2-pin power plug and a Ground Switch, which ties a 0.047uFd cap from chassis to either side of the power line. Ideally this has been replaced with a proper 3-pin plug, ground switch cap removed; but many vintage amps have not been updated to modern safety standards. So ideally the player sets the ground switch to tie chassis to the groundED side of the wall output. If the wall outlet is Balanced Power, there may be danger either way. 60V beats 115V, but still a bad idea.

The outlet should conform to best US practice of the 1960s. It should have a true groundING U-hole, but should also have one side of the power-holes tied to supply ground. Anything else violates unstated and maybe unsuspected ASSumptions.[/quote]

That would rule it out fer sher then. Better a U-hole than an A-hole...

I recounted another case of the Law of Unintended Consequences a while back, in regards Balanced Power, when Keith O. Johnson got hold of a Harman powered speaker which had the misfortune of an anomalous fault condition internally---perhaps the only one in the universe, like the Cabrera magnetic monopole---where the AC neutral was tied to signal ground. It promptly blew up the output stage of his special Pacific Microsonics DAC, in his balanced power system, before the breakers went, and the incident had him hollering for a product recall.

I was able to talk him out of it, but it was a near thing. He had taken the device apart and destroyed the evidence, so we never knew how in the wide world of sports the unit had the errant connection.