Simple inverter cct.: comments, please

[trh01uk]

This is a pretty standard circuit - well the improved version with the diodes is - to replace mechanical vibrators. It still suffers from problems - like the ones Radio_Wrangler listed in post #6. A better solution is to drive a flip-flop with the oscillator output, then use the flip-flop output to drive some MOSFETs (or bipolar power transistors). MOSFETs generally need some reasonable current drive to make sure they transition quickly. The flip-flop will give an accurate 50:50 mark:space ratio - without that you get "flux walking" where the transformer develops a DC bias, and its power output is limited by saturation.

If you are going to use it as an invertor with any radios nearby, you will then have to deal with the resulting RFI situation......

Richard

And good rise time of just shy of 40ns!

Ah, then add the gate capacitance that will slow it down on the up side, fairly fast down, could be a good thing stopping both FETs being on at the same time.

[MrBungle]

Indeed. Gate capacitance of the IRF630 is up to 700pF. That's going to take a LOT of current to drive into saturation quickly. To get a respectable 10ns time constant that's going to require 70mA impulse current up front. Ergo there's going to be a switching transistor based follower on the output of the oscillator I suspect

This sort of stuff is incidentally why Intel CPUs take up to 100A or so in impulse currents!

[MrBungle]

Actually I just tried it with a 680pF load and the rise time went up to 72ns. Not terrible.

[Bazz4CQJ]

This old thread on 'Solid State Vibrators" https://www.vintage-radio.net/forum/...ad.php?t=41944 may be worth looking at. Quite a few references to the use of 4047 chips as oscillator drivers.

B

[Radio Wrangler]

Don't forget the feedback capacitance from drain to gate, by the Miller effect, that is scaled up because it doesn't go gate-to-ground but to something which is swinging by twice VCC in the opposite direction to what you are trying to swing the gate.

The fast-off, slow-on effect stops cross-current which is important. If you go to fast drivers you need something else to do the job. A decent SMPS control chip with fast drivers, and a timed dead zone, has a lot going for it.

David

[MrBungle]

Haven't had time to write it up properly as the MOSFETs only arrived about 30 minutes ago but this works with two IRF630's generating a respectable 214 volts with a 12-0-12 / 115-0-115 transformer. Quiescent current is 148mA and the MOSFETs run cool to the touch.

Voltage is true RMS with a Fluke 8050 and I haven't put the scope on it yet so the DC potential might be higher. I don't have any capacitors I can safely stick across it without blowing myself up so I'll order some bits for Thursday delivery.

[MrBungle]

Found an average responding meter under the pile of junk. 244v at drive voltage of 13.8v.





I'll find something to load it with later as well. I have some LED mains bulbs floating around somewhere.

[MrBungle]

Just a heads up to anyone who builds this. I've blown the gate out of a MOSFET with it. Make sure you stick some protection diodes across the MOSFET S-D pins or it'll go phut when you turn it off.

Also with a 5.6k load on it, it smoked a resistor instantly at 200v.

With a 12k load it delivers 200v 16mA approx. 3.2W out for 5W in. You can see the transformer saturating. If I speed up the oscillator so it runs to just before saturation I should get power improvement.

[MrBungle]

Final spamming. At 200Hz it makes no difference to power delivery. On the primary side there are 224v spikes (!) which is a little bit above the 200v Vds rating of the IRF630. Voltage is consistent down to 30mA and after that it dies miserably. If you put snubber diodes across the MOSFETs it becomes unstable. No amount of decoupling fixes it. End result three dead MOSFETs.

In summary, this circuit is not very good. You can get 200v at 10mA out of it but it's going to be noisy and inefficient and unstable and gobble up FETs.

[G6Tanuki]

Why not forget struggling with modern recreations and instead dig out some historic Germanium and build a 'classic' inverter using only two transistors with minimum component-count ??

Look at the circuits for the military 62-set dynamotor-replacement, or the Pye AM/FM10 Cambridge or AM/FM25 Vanguard VHF mobile-radios.

http://www.vmarsmanuals.co.uk/newsle...es/gfn_vib.pdf

The military/commercial-radio world of the 1960s expected the transistorised versions of dynamotors and vibrators to be more-reliable than the mechanics they replaced.

[MrBungle]

That's basically the same but the tip3055's were MOSFETs.

[Argus25]

Skywave,

There are many important considerations with this sort of converter. The original type had a setup where the frequency of operation was not determined by timing capacitors in a separate multivibrator configuration, but by the time it took for the transformer primary to saturate, these were called a Royer Oscillator. The saturation never lasted long as it initiated the transistors to change states.

Still, even with a separate driver oscillator, if the operating frequency was set too low it would cause significant problems if the transformer was forced into saturation.

Also, for a self oscillating version, or an oscillator driven version, it is important that the leakage reactance of the primaries are low and they are bifiler wound. This makes mains transformers used in reverse a poor choice. As are the electrolytic timing capacitors a poor choice.

I have investigated a good number of designs for this application, including self oscillating versions, and oscillator driven versions, using both fets and bjt's to replace electro-mechanical vibrators in vintage radios. I have also examined the leakage reactance problem that results in the voltage spikes on the switching edges. It is all in this article if you want to examine the types of circuits and the issues involved. There are also efficiency figures for each design:

http://www.worldphaco.com/uploads/ZC...R_SUPPLIES.pdf

(I could add that one advantage of a self oscillating version is with the correct primary tuning is that it self commutates and switching transients become negligible compared to oscillator driven versions and, the whole converter is self short circuit protected as short circuit kills the feedback shutting it down).

[Radio Wrangler]

If you want a DC output, then it makes a lot of sense to use an SMPS chip and have feedback regulating it by pulse width modulation.

If you want a sinewave output, there is an interesting sine inverter by Neville Mapham of General Elsctric (USA) It uses SCRs. HP used this in the original model 70000A mainframe but running 40kHz. Then the fast SCRs went obsolete (They were the Mullard Syclops ones!) and yours truly got the job of designing a replacement inverter and upping the power a bit. It turns out you don't need SCR self-turnoff in this circuit. IGBTs driven with squarewave gate drive are perfect.

David

[MrBungle]

Agreed. Linear's LT1073 wins for me. Seems to work in every application I throw at it. Also as the switching frequency is much higher, the transformer/inductor doesn't weigh 500g!

[Argus25]

There is really no need to use IC's or driver oscillators unless you want more complexity and more fragile parts. All that is required is the two power devices.

If they are mosfets the circuit can be configured as the self oscillating mosfet version, as in the article I cited in the post above. If they are BJt's then all that is required is the addition of a small driver transformer, to allow for the fact that a standard power transformer is devoid of feedback windings.

Tek as an option in their 464 scopes made the main power transformer oscillate so the scope could run from low voltage DC. They used a very interesting small toroidal driver transformer to do that.

If BJT's are used, it is also possible to simply AC couple to the bases from the opposite transistor's collector circuit, but it is very wasteful of energy in the bias resistors compared to the driver transformer method.

The common DC-DC converter, using two power germaniums, with the feedback windings on the output transformer (Royer oscillator), was very popular in the 1960's and used for CDI units, typically used NKT404's or 2N174's. They nearly always used a toroidal transformer to keep the primary leakage reactances low. If 2N3055's are dropped into these circuits they are often unstable and oscillate at high frequencies destroying the transistors, but that can be cured with 0.1uF base-collector feedback capacitors added. In these, when the SCR fires to generate the spark the converter output gets shorted, and automatically shuts down while the spark is being generated and the converter automatically re-starts after that.

[julie_m]

I built a small inverter using a 4047, four BC548s, two BD438s and two BD437s in a H-bridge to drive a parts-drawer transformer without a centre tap, for powering a standard mains compact fluorescent lamp from a 12V battery. I upped the frequency to about 90Hz to squeeze a few more VA through the transformer than it was really good for. (This can't be done indefinitely. Too high a frequency risks running into iron losses -- the thickness of the laminations allows eddy currents to fl9w within each one -- and the recovery time of the diodes in the lamp's internal bridge rectifier.)

Some of the transistors probably could be optimised away; but this design was a follow-on from an earlier attempt to be clever which failed spectacularly.

With a 10W lamp, the H-bridge transistors ran cool with no heat sink and even the transformer (from a factory with the motto: Regulation? We've heard of it) barely got warm, even inside the ceramic base of the lamp.

[Argus25]

Originally Posted by julie_m

Too high a frequency risks running into iron losses -- the thickness of the laminations allows eddy currents to flow

Yes, that is a very good point. The hysteresis losses are linearly proportional to the frequency, but the eddy current losses are proportional to the square of the frequency, so these increase significantly when the frequency gets elevated. Most commercial early DC-DC converter transformers (typically made by the Triad company with part number like TY-81) for this reason had ferrite cores rather than iron where the eddy current losses, compared to a laminated iron core, are eliminated. They chose an operating frequency around 400Hz.

Prior to electronic drivers & controls for DC-DC converters becoming more sophisticated, there was a lot more attention to the magnetic materials for DC-DC converters. For the self oscillating types, the B-H curves were nearly perfect sharp cornered rectangles with sudden onset saturation, because it was the core that determined the operating frequency, not the driver electronics. This is the case for the unique Toroidal driver transformer core in the DC-DC converter option in the Tek 464 scope, it has an astonishingly sharp edge box like B-H curve.

[trh01uk]

How are all you guys designing and building these invertors coping with the RFI problem?

I am particularly interested in the situation where an invertor - say at the 10W level - is put inside one of the older transceivers which originally ran on primary batteries, and demands multiple supplies - using two HT rails and possibly two LT rails as well. The invertor has to go in the radio's battery compartment and is thus very close to the RF circuitry - the RFI has to be extremely well controlled for this to work.

Richard

[Argus25]

Richard,

I can help answer this question.

It is actually a difficult task to put a switching converter inside am MW radios/ transceivers. A radio makes an excellent sniffer for RFI.

If there is any type of reasonable sharp edge square wave (meaning the upper frequency harmonics are at least 23 or more times the fundamental frequency for it to look reasonably "square") then likely RFI is generated. There can also be direct radiation from the wiring and the transformer core. But even in transformer-less voltage multiplier designs, there is often RFI generated.

In the past I have experimented with step up supplies with switching designs that do not produce RFI. (of course a pure sine wave supply won't but they are less efficient).

I was able to create a step up switching 90V supply with zero detectable RFI. The definition being its running pcb could be in direct proximity/contact to a radio tuned to 100kHz and nothing would be picked up. This was achieved by:
1) keeping the fundamental switching rate low 2) controlling the rise and fall times of the switching waveform and 3) avoiding pushing the transistors too heavily into saturation (as they generate RFI coming out of saturation)

The design of such a switching 90V supply (suitable for a battery valve radio for example) that has no detectable RFI is at the end of this article, so scroll to the end, it is called the Omega Device:

http://www.worldphaco.com/uploads/WORLDFETRON.pdf

As you can see it looks a lot like a low frequency DC-DC converter, but with heavy primary tuning.