Thoughts for a low-power 60Hz 120Vrms signal generator please

[eq1]

With the AD9837 you'll get a very precise frequency and with the output peak-hold controlling the input peak into the transformer you'll get a precise output level too.

I was thinking about this while drinking my morning coffee and realized the peak-hold solution is way too hard. One could make the hold cap leak so the peak doesn't need to be refreshed, hence no need for a reset pulse, but the AD9837 is giving a pure sine and the THD of the VGA and power amplifier will be very low at 10s of Hz (with any part made after 2010) so the measurement of the output can be to just integrate this sine and compare that to a reference. This would provide a higher bandwidth on the control and is easier to implement. The easiest, and most accurate way, is to just buy an amplifier that measures Vrms directly. Or, rectify the output and integrate that with a time constant that's slow compared to 60Hz, not Vrms but good enough for a known pure sine. Diode drop is small compared to 120V so even that should still be a quite accurate measurement w/o much effort.

 

[berkeman]

I prototyped the tip from @tech99 today, and I think it will work with some changes to my initial try. The 5kOhm load that I'm driving takes around 3W to drive at 120Vrms, so my initial setup used a 15W audio amp to drive a 5Vrms:120Vrms (15VA) transformer. I tuned the output capacitance (using X2 caps) to resonate at 60Hz to increase the impedance that the audio amplifier was driving.

It worked pretty well, but my initial guess of a 15VA transformer sizing turned out to be a bit too small and it saturated at higher voltage drive levels (probably due to the low Power Factor using the output tuning). I ordered a larger VA transformer today and hopefully will be able to test it Monday or Tuesday.

As an update, even with a 40VA 5V:120Vrms transformer, at the higher drive levels I was getting a 3rd harmonic instability. I'm thinking it's an issue with parasitics in this line of transformers, but thankfully I came up with a way around it for now. I've added a simple isolation transformer at the output, and used a 110V:220Vrms version to get voltage doubling. That let's me run the 5V:120Vrms transformer at lower levels to avoid the 3rd harmonic problems. It's running on the bench now, and has low jitter like I wanted. I'll use it for a test production run this weekend to see how it holds up.

AD9837

That's an interesting part. It looks like it may have low enough noise that I could use it in this fixture. I'll read over the datasheet and maybe order one of their Eval Boards (Digikey has them in stock). Thanks for the idea.

 

[berkeman]

Have the amplifier reference be a sine from an AD9837. You can put a peak hold on the secondary side of the transformer and use it to control the voltage gain on the low voltage amplifier. With the AD9837 you'll get a very precise frequency and with the output peak-hold controlling the input peak into the transformer you'll get a precise output level too.

Could you say more about how you would level the output voltage from this sine generator IC using feedback? Are you thinking of adding a variable gain stage between this IC and the audio power amp circuit? The datasheet for this IC lists 200ppm/C as the temperature coefficient on the output voltage (0.02%/C), so I'd definitely need to do something to level the output amplitude. Thanks.

 

[Tom.G]

A slightly orthogonal approach:
If the current sensing does not need to reproduce the current waveform. (or maybe even if it does.)

Rather than using a high purity and stable 120VAC, can you use the ratio between the current reading output and the 120VAC supply? Stated another way, the transducing device is reporting the percentage of the 120VAC excitation as the output.

(bridge measurement or ratiometer concept)

Cheers,
Tom

p.s. Those X2 caps you are using for a tank, it is the next character in the part No. that specifies the temperature stability. See: 'Class II' at
https://ec.kemet.com/blog/mlcc-dielectric-differences/

The site also has a simulator (K-SIM) showing temperature curves.

 

[cmb]

@berkeman, IMHO I think any solution to achieve 0.01% is really not a simple project.

The proposal to use an AF amp into a audio transformer is exactly how we do it (I speak in the past tense having been a principal engineer in ISO17025 testing). We'd then quote a measurement error of a % or two.

The act of measuring this to provide a feedback loop is a main complication.

TBH, having thought a bit more about it, you might find value in the principle of the saturation amplifier. Given you can just run mains and 'modify it a bit', this is, on reflection, probably your optimum.

https://en.wikipedia.org/wiki/Saturable_reactor

The variable resistor here can be a Darlington (high current capability) fed by a precision op amp. You have to peak-detect your last peak, and smooth the transitions of the transistor, however that could be quite simple as the appropriate means would be to shift the AC to >0, smooth it with RC via a peak-detect diode, and selection of those components will then average out the last few peaks.

In theory you could get the monitoring circuit to feed the transistor and use all passives (and one transistor) only.

Now ... where to buy transformers for saturation reactors? You might find an AC isolation transformer with an additional low voltage secondary winding, this might work.

IIRC, it was popularly used as a 'magnetic amplifier' in the 20's early 30's I believe, particularly in Germany for some reason, maybe they had difficulty making or importing valves?

(Personally, I'd still look at a MHz rate FET-switching voltage follower with a precision comparator to an input signal.)

Good luck with whatever you do. If you want any specific circuits from me, let me know.

 

[Tom.G]

Another possibility is a ferroresonant transformer, i.e. a constant voltage transformer such as supplied by Sola (now part of Emerson Electric).

They are frequency sensitive though, ±1% frequency yields ±0.25% voltage change, so drive it with a stable oscillator - not the power line.

Capacitive loads confuse them and I don't know about output waveform, better check if important. Also, they are designed to run hot in normal operation due to designed core saturation

Here is an data sheet and synopsis:
https://www.emerson.com/documents/a...nusoidal-transformers-solahd-en-us-163820.pdf

Cheers,
Tom

 

[eq1]

Could you say more about how you would level the output voltage from this sine generator IC using feedback? Are you thinking of adding a variable gain stage between this IC and the audio power amp circuit?

I can only use text right now so I'll do my best. I was thinking the signal chain would look like this:

AD9837 -> VGA(e) -> PowerAmp -> Non Isolating Transformer -> 5K load -> Rectifier -> Integrator -> subtract reference -> e

e is the error signal which gets fed back into the variable gain amplifier (VGA). Design the VGA so positive e lowers gain. Set the integrator with the time constant of maybe 10 periods of the AD9837 signal. That should be sufficient to track out error introduced from the environment and be stable. The VGA should probably be able to respond to e faster than the integrator can change its output. So for 60Hz AD9837 maybe integrate at 6Hz and a VGA that can respond to e w/o attenuation at 600Hz. That should be safe without doing any math or designing actual circuits. Then you can support 50Hz/60Hz too. :)

Not totally sure the best way to do circuit startup. Maybe slowly ramp the VGA in an open loop way to give the integrator it's warmup time. Then after a timeout switch to e. That's probably the simplest.

The reference is the expected average value of the absolute value of the sine with 120Vrms or 240Vrms depending on the mode the user selects. If you want to really dial it in, design the reference to be tweaked with a pot so you can hand calibrate out the sensor offset (static error from things like the rectifier, etc.)

The output sensor (rectifier+integrator) is not measuring rms directly, but since the AD8937 is only giving a single tone, and distortion can be minimized by design in the signal chain, I expect it to be sufficient.

 

[PaulWDent]

It's a matter of accuracy and variation versus time. The BESTEK is very steady, except for the bistable problem. My expensive AC mains generator is still jittering more than I like. I tested our lab AC Mains feed near the end of today, and it was varying around 0.1% continuously, which is way too much.

I've tried to use a signal generator plus step-up transformer, but the 50 Ohm output signal generator can't drive the typical 5Vrms-120Vrms step-up transformer.

Then drive it with a unity gain op-amp! e.g
This seems a very trivial problem

Thanks, this is the most promising avenue so far. Based on my measured 120V winding inductance, I'm able to resonate that winding with about 4μF of X2 capacitance. That makes the input impedance at the 5V winding about 4 Ohms for the 15VA AC transformer that I'm using right now. That's in the right place for an audio power amp. More in a bit...

Your 4 ohms can't be right. With no load on the secondary it would be much higher.
I think you said the load at 120v was 5k ohms. So let's do a calculation: The power you need is 120^2
5000. That's 2.88 watts. You don't usually get that much power out of a signal generator so you do need an amplifier.
You need to build a unity gain amplifier using a rail -to-rail op amp followed by a complementary pnp/npn emitter follower using beefy transistors. Then use your capacitor-tuned step up transformer.

 

[PaulWDent]

As an update, even with a 40VA 5V:120Vrms transformer, at the higher drive levels I was getting a 3rd harmonic instability. I'm thinking it's an issue with parasitics in this line of transformers, but thankfully I came up with a way around it for now. I've added a simple isolation transformer at the output, and used a 110V:220Vrms version to get voltage doubling. That let's me run the 5V:120Vrms transformer at lower levels to avoid the 3rd harmonic problems. It's running on the bench now, and has low jitter like I wanted. I'll use it for a test production run this weekend to see how it holds up.
That's an interesting part. It looks like it may have low enough noise that I could use it in this fixture. I'll read over the datasheet and maybe order one of their Eval Boards (Digikey has them in stock). Thanks for the idea.

It's not the VA that needs to ne higher - it's voltage. Use a 240v winding for 120v if you want to operate far from saturation. It's volt-seconds per turn that determine the core magnetic field, not output current. Mains teansgormers sre designed with only 10% margin to saturation and it sounds like you need to be far from that. You might also like to know that transformers take an inrush current when first switching on the power because if you switch power on at a voltage zero crossing the first half cycle integrates to give double the steady-state volt-seconds. So operating with a 2:1 voltage derating is definitely what you need.

 

[berkeman]

It's not the VA that needs to ne higher - it's voltage. Use a 240v winding for 120v if you want to operate far from saturation. It's volt-seconds per turn that determine the core magnetic field, not output current. Mains teansgormers sre designed with only 10% margin to saturation and it sounds like you need to be far from that. You might also like to know that transformers take an inrush current when first switching on the power because if you switch power on at a voltage zero crossing the first half cycle integrates to give double the steady-state volt-seconds. So operating with a 2:1 voltage derating is definitely what you need.

Please repost your reply without the typos. Maybe wait until morning?

 

[cmb]

Then drive it with a unity gain op-amp! e.g
This seems a very trivial problem

Would you please name the 120V [3-nines-accurate] calibrated signal generator you have in mind, and then the 120V op amp, by serial number/manufacturer?

Thanks. Should be very trivial.

 

[Windadct]

Would KWHM test equipment work - since you indicated it was a power meter.

Ref : https://www.tescometering.com/

 

[cmb]

Would you please name the 120V [3-nines-accurate] calibrated signal generator you have in mind, and then the 120V op amp, by serial number/manufacturer?

Thanks. Should be very trivial.

Sorry, I just realized my error but timed out on editing ... One needs a 350V signal generator and op-amp (we're talking peak-peak here).

Question again to @PaulWDent . Thanks.