High Speed, Unipolar Sine Inverter?

Thread Starter

johnyradio

Joined Oct 26, 2012
434
Hello,

What's a simple, low-parts-count way to do this?

I need to convert 120 V, 60 Hz AC USA mains to 5V, 5A, 100Khz varying DC voltage.

For convenience, assume we'll use an off-the-shelf mains-to-DC converter for the first stage. You decide your preferred specs of the off-the-shelf mains-to-DC converter for the first stage. It should include isolation for safety.

Therefor, this question only concerns the design of the second stage, the DC/AC inverter.

There are two acceptable outputs. We'd like either one, because we need to test both.

It could be the full sine wave, with peaks are at 0V and 5V.

1679409333423.png

Or it could be a truncated sine. I didn't say "half rectified" because that makes an assumption about how the pulse is produced.

1679409019097.png

Here are some corrections and more details on the requirement, based on my better understanding.

Voltage regulation should be +/- 1%.
Current limiting not needed, but yes short-circuit protection.
The "0V" mustn't go negative, not even a little.

We need sine to improve efficiency of all following circuitry, and reduce various drawbacks of a non-sine PS, eg noise, and because our load wants smooth power. So yes, wave-form-symmetry/distortion is important. I assume it can be a pretty decent but not perfect sine, and still get those benefits. So i assume a very low-res modified sine won't do, but a hi-res MSW may be fine.

1679409542152.png

I don't know if Ringing, Flyback, and/or, Inductive/Capacitive-Reactive-Load considerations are important.

I'll try to characterize the load. I believe the impedance of the load is: @1kHz AC, 8 mΩ

There won't be any sudden transients.

We want voltage regulation, not current regulation. So it's a constant voltage PS. The load will send a control signal back to the PS to regulate voltage as needed.

Frequency stability/accuracy isn't important.

PWM is fine, but not required. My thought is that we may be able to achieve this with PWM, a sine driver, and a single mosfet, rather than H-bridge.

I'd like to accomplish this without a uC, if possible. Unfortunately, the EGS002 is restricted to 60 Hz.

I'm fine doing this with all off-the-shelf modules. That's actually preferred in the short-term.

Thx
 
Last edited:

Papabravo

Joined Feb 24, 2006
21,225
Convert the AC to DC and then convert the DC back to high frequency AC. Short of building a mechanical generator with the associated losses, I don't know of another way to do it. Neither method will be especially simple or low cost.
 
Hi johnyradio,

What output waveform is needed at 100 kHz? Sine wave, square wave? Also: is isolation needed between the mains input and the 100 kHz output?
(I assume yes but curious).
 

strantor

Joined Oct 3, 2010
6,787
Short of building a mechanical generator with the associated losses, I don't know of another way to do it. Neither method will be especially simple or low cost.
100kHz mechanical generator would need to spin six million RPM (2-pole) or three million RPM (4-pole). Definitely not a cheap option!
 

Papabravo

Joined Feb 24, 2006
21,225
@strantor wrote:
100kHz mechanical generator would need to spin six million RPM (2-pole) or three million RPM (4-pole). Definitely not a cheap option!

Here's a fun exercise. What would be the maximum diameter of the rotor to keep the outer edge subsonic?
A sonic boom from a generator would be something else.
 

ronsimpson

Joined Oct 7, 2019
3,037
I need to convert 120 V, 60 Hz AC USA mains to 5V, 5A, 100Khz AC. No DC needed at output.
There are a number of ICs on the market that are built for lighting.
Power line voltage is rectified and filtered to make "Vbus".
The IC has a oscillator and two MOSFET drivers inside.
The MOSFETs place 1/2 the power line voltage across the primary of the transformer.
The turn ratio makes the 5V as required.
This version of the circuit does not really regulate. There are other ICs like this.
Years ago I used an IC that had the MOSFETs in the IC.
1679269860269.png
 

Papabravo

Joined Feb 24, 2006
21,225
The TS asked:

What's the simplest, lowest-parts-count way to do this?
In what way is this circuit simple and low parts count? I count at least 2 dozen parts and what will likely be a custom transformer.
 

strantor

Joined Oct 3, 2010
6,787
That would take some expensive machining on exotic materials — no doubt!
@strantor wrote:
100kHz mechanical generator would need to spin six million RPM (2-pole) or three million RPM (4-pole). Definitely not a cheap option!

Here's a fun exercise. What would be the maximum diameter of the rotor to keep the outer edge subsonic?
A sonic boom from a generator would be something else.
This thing made me think twice too, but it makes me wonder if it is even possible, if one had means and determination. So I will approach it from a more (semi-)realistic direction.

3M RPM is already 10x higher than the fastest jet turbines. Some nonstandard configuration could probably achieve a higher number of poles, and a lower RPM (probably at the cost of a larger diameter rotor). The highest pole count generator I found in a brief google search was 16 pole pairs. There are probably higher, but none jumped out. A 16-pole, 100kHz generator would "only" need to spin 750k RPM.

speed of sound = 13,504 in./s
13,504 / pi = 4,298.45"

750k RPM = 12,500 revs/second
4298.45 / 12500 = 0.344"

So, your rotor would have to be about 11/32" (8.75mm) diameter.
I don't think 16 poles will fit there.
 

Papabravo

Joined Feb 24, 2006
21,225
This thing made me think twice too, but it makes me wonder if it is even possible, if one had means and determination. So I will approach it from a more (semi-)realistic direction.

3M RPM is already 10x higher than the fastest jet turbines. Some nonstandard configuration could probably achieve a higher number of poles, and a lower RPM (probably at the cost of a larger diameter rotor). The highest pole count generator I found in a brief google search was 16 pole pairs. There are probably higher, but none jumped out. A 16-pole, 100kHz generator would "only" need to spin 750k RPM.

speed of sound = 13,504 in./s
13,504 / pi = 4,298.45"

750k RPM = 12,500 revs/second
4298.45 / 12500 = 0.344"

So, your rotor would have to be about 11/32" (8.75mm) diameter.
I don't think 16 poles will fit there.
What a hoot!
I hereby bestow upon you the "Furlongs per Fortnight Memorial Award"
 

ronsimpson

Joined Oct 7, 2019
3,037
This circuit goes inside a light bulb. The feed back could be left out if you don't care about regulation.
The diodes and capacitors on the output you do not need.
1679322490459.png
 

LowQCab

Joined Nov 6, 2012
4,061
"" What's the simplest, lowest-parts-count way to do this? ""

It depends on what You mean by "do-this".

What are You trying to accomplish as an end-result ?

What are You driving ?

Sine, Triangle, Square-Wave ?
Is Frequency stability/accuracy important ?
PWM ?
Is Wave-form-symmetry/distortion important ?
Is Feedback from some other device important ?
Are Ringing, Flyback, and/or, Inductive/Capacitive-Reactive-Load considerations important ?
Current-Limiting ?
.
.
.
 

ronsimpson

Joined Oct 7, 2019
3,037
But it looks like the NCP1392 supports up to 1A peak. We need 5A rms.
That 1A is the gate current into the MOSFETs. It has nothing to do with output current. The size of the MOSFETs governess the current.

120V ac which gets rectified and filtered to make 170V dc.
The transformer will have 160Vp-p on the primary. Which is 80 volts peak square wave. (depends on what mode and values of caps etc)
I do not know what type of "5Vac" you want. ???
If the transformer had a 16:1 turn ratio the output will be +5V to -5V.
5A on the output of the transformer (16:1) will need 320mA on the primary. The MOSFETs can be very small.

Here is another IC. I cannot find the one I used years ago where the MOSFETs are inside the IC. (transformer not shown)
1679345829061.png
 
Last edited:

Papabravo

Joined Feb 24, 2006
21,225
No DC output needed.
You still don't get it. You may not need the DC output, but it is an essential step in changing the frequency of an AC signal. There is no tother PRACTICAL way to do this.

You also don't get the point that the diagram in post #11 has BOTH and AC output and a DC output and apparently NO INPUT!!! There is probably a typo somewhere, but a correction might be helpful.

Reread carefully and tell me again if you think I am wrong.
 

Thread Starter

johnyradio

Joined Oct 26, 2012
434
"" What's the simplest, lowest-parts-count way to do this? ""

It depends on what You mean by "do-this".

What are You trying to accomplish as an end-result ?

What are You driving ?

Sine, Triangle, Square-Wave ?
Is Frequency stability/accuracy important ?
PWM ?
Is Wave-form-symmetry/distortion important ?
Is Feedback from some other device important ?
Are Ringing, Flyback, and/or, Inductive/Capacitive-Reactive-Load considerations important ?
Current-Limiting ?
.
.
.
Reply moved to original post.
 
Last edited:
Top