class D amp

Your sketch circuit is completely open loop. Perhaps feedback would reduce your distortion and DC offset worries ?

piglet

--
This email has been checked for viruses by AVG. 
http://www.avg.com

There's feedback inside the TPA3251.

Allan

Sure, but JL was talking about the chip adding 60mV offset causing 1A primary current. External overall feedback could bring that to mere microvolts and better reduce distortion introduced by LC filter resistance.

piglet

You can get the bare board from you know who for $3.45 delivered:

Why?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Ferromagnetism is weird.

Once had a 10kW, "high frequency" (low kHz) stripwound toroid in an induction heater. The inverter (just simple square wave output) wasn't quite balanced, so it went into saturation (audible with the loud, harsh noise) pretty quickly..

So we put a coupling capacitor in, and ran it again. DC is gone, but it's still saturating. Huh. Even stranger, it's fine for a few seconds, then "walks" over to saturation.

Put in transformer #2. No problem, nice and quiet (well, relatively speaking).

Some ferromagnetic materials have saturation and history problems: NiZn ferrite often warns not to saturate the material (though I don't know why; I haven't tested that yet). I've seen the same behavior, the long time constant, "flux walking" saturation, with square permalloy cores.

Those transformers were interesting. We quoted a dozen transformer shops across the country. Most "no quote". A few came back with unrealistic designs, like, a 24" wide toroid. Guessing their "designer" was using winding and loss tables for 60Hz. ;-) The place we finally bought from, gave us something about $800 and 10" o.d., with a copper tube in the middle for cooling water, exactly what I was expecting. Couldn't find any thermal problems with them.

As far as I know, those units are still in service at LANL. Some sort of material processing system. Something about a rare, toxic metal -- probably nothing interesting. :^)

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

s

; I

le

al

bly

Staircase saturation can be difficult to manage with full confidence

Until recently I only looked at the primary side control, but you also need to consider unbalance on the secondary. Typically on the primary the stair case is prevented by a series capacitor, but that only works for voltage mo de control systems

For the secondary, one could have say a halfbridge converter with primary s eries capacitor, but secondary could be what is called a current doubler ty pe:

The problem with that type is that if the diodes are not equal (and they ne ver are), then different volt seconds are presented to positive and negativ e cycles leading to staircasing. In the halfbridge, the staircasing can be mitigated by series resistance which offers negative feedback. But staircas ing exists for sure

Cheers

Klaus

nt

h

t's

en

hy; I

ps

m,

ddle

rmal

of

bably

ed to consider unbalance on the secondary. Typically on the primary the sta ircase is prevented by a series capacitor, but that only works for voltage mode control systems

series capacitor, but secondary could be what is called a current doubler type:

never are), then different volt seconds are presented to positive and negat ive cycles leading to staircasing. In the halfbridge, the staircasing can b e mitigated by series resistance which offers negative feedback. But stairc asing exists for sure

Oh, forgot to link that to Larkins shcematics, is secondary unbalance possi ble?

Bytheway John, nice idea, good use of the IC if it works

Cheers

Klaus

Annoyingly, you can't feed low-impedance input to those digital amps, you gotta DC-block with capacitors. So, that pic has some issues (won't bias the inputs right). That also means that if you have a toroid doing voltage-boosting, and care about DC, you can't correct the DC level at those input terminals.

You might, however, run a second winding on the toroid with a modulated load, that would let you balance the induction with a resistor-to-small-voltage load that opposes the (DC-offset) current.

Have a couple of the older rack mount 3ph 115v 400Hz psus, one 500 and another 750Va. Used to build 400Hz inverters for mil surplus gyros etc for leisure flyers, transformers at the time, but Scott connected to get 3ph from 2 channels. 400Hz is useful to have around the lab anyway, if you do any avionics type work...

Chris

-----------------------

. ** No point discussing "why" when you do not know the facts.

There are several ways your scheme can get into trouble with core saturatio n, only one of which you have addressed.

I mentioned low frequencies, so good HPF filter at the input to the class D amp is advisable. Also large inrush surges are generated when signal is fi rst applied and this needs addressing.

The third one is if the class D amp has response down to DC *and* when the signal is asymmetrical. I don't mean that it has a DC offset, but the peak values are non identical. If allowed to clip the amp, a large DC offset is produced at the output that will saturate the toroidal core.

I guess this scenario is similar to "staircasing".

The normal fix is to make sure the class D amp's NFB loop reduces the gain to unity or zero at low frequencies as is the case with the vast majority o f audio amps. Another way is to use an offset servo loop.

See this article on by my colleague Rod Elliot, whom I alerted to the pheno menon.

..... Phil

The loads will be low-Q inductors, like synchro or LVDT windings. I don't expect any diodes or anything.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

I didn't show the input caps in the sketch, but the chip needs them. Given the input caps, the half-bridge output DC offset is 60 mV max.

It's impressive how much important stuff they can leave out of a 42 page data sheet.

The caps in the transformer primary are supposed to kill any DC. The diodes make it more interesting.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

I've successfully kept Ethernet transformers happy by watching the current*time balance in the center-tap. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| STV, Queen Creek, AZ 85142    Skype: skypeanalog |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website. 

Thinking outside the box...producing elegant & economic solutions.

I like the sound of that.

You mean JL could compare the primary center-tap against supply mid-point and feedback to eliminate asymmetry problems?

piglet

--
This email has been checked for viruses by AVG. 
http://www.avg.com

Not quite, but close. Rather than observing volt-seconds, which is slow, watch current to/from center-tap for asymmetry, cycle-by-cycle. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| STV, Queen Creek, AZ 85142    Skype: skypeanalog |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website. 

Thinking outside the box...producing elegant & economic solutions.

FYI, nothing is staircasing. At least, whoever came up with it, chose a terrible name.

Staircase flux would only happen if you apply "pulsed DC" (i.e., a unipolar, constant voltage source) to a zero-resistance winding.

Note that's not /switched/ DC, which allows flux to return to zero (flyback). Because the source isn't constant voltage in that case: it switches between constant voltage and constant current (the current is zero).

Even resistance will "tilt" the staircase, so that it goes backwards during each "tread", and after a few L/R time constants, it will simply be ripple: or, if you like, it'll be a very "tilted" staircase that doesn't climb at all.

So it would be very difficult indeed to have a circuit generate a proper staircase.

Maybe "staircase" is meant to mean, if you sample the flux at a consistent point each cycle -- that way if there's an imbalance, it's clearly rising; but that's still a poor description, because it's no more a staircase (zero order hold) than a sampled ramp, or any other of infinite functions that happen to have that result when sampled at a particular time.

The phenomena I described seems to have little to do with the switching frequency or duty cycle (given that DC balance is good), and more to do with material properties. It took on the order of 10k cycles to drift off towards saturation. It's not obvious if the saturation was the same polarity each time (hmm, I recall measuring inverter current, but I don't remember which phase was saturating, if it was consistent or not). The timing was inconsistent as well; sometimes it would saturate in a few seconds, sometimes it would come and go, then go back to saturation.

Not that that phenomena is terribly common either, but it's a material thing more than a circuit thing.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

ar,

Staircasing, or a better name for it flux walking. The name staircasing com es AFAIK from push pull converters, when if you observe the current in one winding, it will show a semi staircase look cycle to cycle (envelope)

It is not an effect caused by the magnetics, but caused by unequal volt sec onds product by the converter control. For example in a push pull converter with non equal duty cycle for push and pull, or from other non-linear effe ct (RDSon, matching of push/pull paths etc)

It can be remedied by deadtime and negative feedback of the paths (increasi ng resistances, bifilar windings etc)

A current mode converter can also remedy the effect, but also cause it to b e worse, due to interaction of the coupling capacitor and the current mode control, which makes the problem a lot worse

See page 7 for explanation:

Cheers

Klaus

** JL's inductance figure is a nonsense, the tranny is toriodal with no effective air gap so the L value varies widely with magnetisation.

If JL just used a coupling cap, say two 1000uF 35V electros back to back, all DC offset issues disappear.

If he clipped the input signal so the amp never clips and followed that by even a 6dB/oct HPF set at 300Hz - another bogey is gone.

The final bogey, inrush surges, can be *eliminated* by using a tranny with double the primary voltage rating at 300Hz his class D amp can deliver.

Yep, AC supply transformers do not surge at switch of if run at half rated voltage or less. Fact.

.... Phil