sinusoidal amplifier 1 MHz

[Neviolograpy]

I have to build an amplifier to drive a power transformer for sensors applications.
On the secondary I need a pure sinusoidal wave at about 1 MHz, 50 Vpp, low impedance (some tenth of ohm). The frequency is fixed.
So I have to drive the transformer with a pure sinusoidal wave. Now as source of the sinusoidal wave I'm using a function generator but in future I'll build my own oscillator.
The transformer I'm using now has a 1:4 ratio, 10 mOhm of DC resistance on primary side, 3 uH inductance, 15 W.
I don't know how to do it... the frequency isn't too high so I thought I'd use the classic configuration of audio amplifier with a 50-100 MHz op-amp that drives a push-pull stage, with feedback connected directly on the output (I need low distortion).
Any suggestions?
Thanks

[TimFox]

If the operating frequency is fixed, can you use a resonant circuit or filter after the amplifier or before the load?

[langwadt]

what's the load?

[Neviolograpy]

Yes I can. Now I try to generate the sinusoidal wave with a cross-coupled oscillator, using the primary of the transformer as the main "L". At the sight works fine, perfect sine wave, but when I analyze the final signal from my sensor I see that it contains a strong modulation at around 4f, that destroy the signal that I have to measure. In practice the reason is that my excitation sinusoid is not pure, I think it will have harmonics at -30 -40 dB, and one of these will resonate with the downstream sensor creating the disturbance. Using the function generator instead of my oscillator, to drive the transformer, I see that my sensor works fine, and I don't have 4f modulation. The problem now is that I need more power compared to what the function generator gives me with its 50 ohm impedance. For that I need an amplifier.
 
The load is a resonant circuit, but as I have seen I have to stimulate it only at one frequency (of the main mode) because it also has other modes of oscillation (at -30 -40 dB) due to parasitic inductances and parasitic capacitances.

[PCB.Wiz]

I have to build an amplifier to drive a power transformer for sensors applications.
On the secondary I need a pure sinusoidal wave at about 1 MHz, 50 Vpp, low impedance (some tenth of ohm). The frequency is fixed.
The transformer I'm using now has a 1:4 ratio, 10 mOhm of DC resistance on primary side, 3 uH inductance, 15 W.

What output power ?
50V p-p takes you above the standard opamp ratings, unless you do a bridge drive.
It may be simpler to do a bridge drive with buffer stages, than to try to add gain stages to the output of a lower voltage opamp.
That keeps you in the realm of simpler and available parts.
TI have a lot of high slew rate, high drive opamps. 
Their THS3491 data shows how to parallel parts for more drive power, which is another option.

[David Hess]

There are some fast high power current feedback amplifiers intended for applications like digital subscriber line which might be powerful enough.  I would probably try designing a discrete amplifier but it may involve more work than you want.

https://www.analog.com/en/products/lt1210.html
https://www.analog.com/en/products/lt1795.html

You say that the load is resonate, but what output current level is involved?

[Vovk_Z]

Some audio amplifiers are capable to work with 1 MHz but they could be too complex.
I would look at some opamp application book to find a voltage boost circuit. I would start with the Art of Electronics possibly.  Here I found one more source: AN-106 (Analod Devices) 'A Collection of Amp Applications'. (Figure 14, p. 7).  OP37 is quite fast itself. If it is not fast enough then there are faster opamps.

[moffy]

Perhaps a modified (switched) sine wave output: https://blog.tripplite.com/pure-sine-wave-vs-modified-sine-wave-explained
 with filtering, the modified output helps a lot with harmonics and also provides a means of modulation to control the final output amplitude. The interesting issue is the low output impedance, how quickly does the output load change? or does the output need to change? If slowish changes are acceptable then feedback can lower the output impedance, if it needs to change very quickly then you become more dependent on a low open loop output impedance.

[Solder_Junkie]

If the load was 50 Ohms, between 6 and 7 Watts of RF would equate to 50 Volts P-P. However, at fraction of an Ohm, you would need a huge amount of power to produce 50V P-P across the load, maybe in the tens of KW.

Operating at 1 MHz, you are right in the middle of the medium wave broadcast band... There is a big potential for widespread interference.

Something doesn't add up...

SJ

[Neviolograpy]

I measured the current absorbed by the resonant sensor exciting it with the function generator, and interpolating to 50 V it should draw an average of 2.5 W (sinusoidal current, with 5 W peak). So 100 mA at 50 V.
On the primary I expect to heve 11 Vpp 500 mA peak.

PCB.Wiz, you are telling me to drive the sensor directly with a bridge, but for balancing reasons I strictly want to put the transformer in the middle. So to reach 50Vpp voltage I can play on the primary-secondary ratio. I don't have +-30V available on my board and it would be very inconvenient for me to have to add a dual +-30V power supply.
What I would like to do is drive the transformer.

In the past I tried to use an LT3917 but I couldn't get it to work... without understanding why. I should pick it up again and try to make it work again because maybe using just an op-amp is the simplest way.
I also like the solution with the OP37!

Thanks to everyone for the precious advices!

[PCB.Wiz]

I measured the current absorbed by the resonant sensor exciting it with the function generator, and interpolating to 50 V it should draw an average of 2.5 W (sinusoidal current, with 5 W peak). So 100 mA at 50 V.
On the primary I expect to heve 11 Vpp 500 mA peak.

PCB.Wiz, you are telling me to drive the sensor directly with a bridge, but for balancing reasons I strictly want to put the transformer in the middle. So to reach 50Vpp voltage I can play on the primary-secondary ratio. I don't have +-30V available on my board and it would be very inconvenient for me to have to add a dual +-30V power supply.

Your original post simply said 50V p-p, so everyone assumed that was from the amplifier you are trying to build.

That's above 'standard' opamp voltages, so that was why I suggested bridge drive to get 50V p-p

If you actually only need 11Vp-p, 500mA pk things are simpler as many opamps can manage that on their own.
You could work backwards from the power supplies you do have, and vary the transformer turns ratio to give lowest amplifier dissipation, and lower drive current gives you more choices.

[boB]

There are many that are driving piezo transducers with similar frequencies but those are resonant and I don't think they are using a 0.1 Ohm drive resistance...

But they are somewhat higher voltages so it might be worth Googling for this at least.

I used the terms piezo and  MHZ to get some results.

boB

[Neviolograpy]

I did other tests: I took the old pcb with the LT1739 and made it work (I had a feedback problem). I drove the transformer with that but I dont get good results. Here too my external sensor is full of harmonics. I attach the schematic.

To recap:
- with function generator, if I drive the transformer with the parallel capacitor (resonating at f = 1 MHz), it behaves as if had an impedance of around 50 ohms. Sensor work fine, no harmonics at 4f;
- the same but without the capacitor that resonates with the primary, no way, you need a lot of power to drive it (and in my opinion it's not the right way)
- with LT1739 that drive the transformer with 24+24 ohm series resistors the waveform is not bad but contains harmonics, the function generator signal degrades a lot. Maybe because I was forced to amplify 10 times (standard configuration), I don't know.

I should test with a function generator that reaches higher in amplitude, because mine only reaches 2.5 Vpp (on 50 ohm).

Thank you boB but piezos have megaohm that I know of; I'll try googling, thanks

[jbb]

Before you go to extremes on the signal source, I have a couple of thoughts.

1. Any chance your test load is nonlinear? Maybe it generates its own 4th harmonics which only become apparent at higher magnitudes?

2. A coupling transformer may not be the best way to do the voltage ratio change; maybe you can use an impedance matching network to achieve adequate results? Maybe this network could be tuned to favour the target 1 MHz and reject other frequencies?

3. Are there ways you could make the testing coil (+resonant cap) less sensitive to harmonics? Maybe you could add some harmonic traps (eg an additional LCR series network tuned @ 4 MHz)

4. In principle, a differential output amplifier should provide better (ie lower) 2nd, 4th etc harmonics. But you’re already using that :-/

5. In principle, an H bridge inverter should provide nice low even (2nd, 4th etc.) harmonics. If operated at 120 degree conduction angle it should also have very low 3rd harmonic content. But it will generate 5th harmonics which could be trouble.

[boB]

Is your load really a parallel resonant circuit ?  That would be high impedance at resonance.

These typical Google project piezo resonators don't take a lot of power but there are BIG ones like in ultrasound transducers that take quite a bit of power.
That was the reason I thought of it but piezo might not be the right word.

Another idea that crossed my mind is to use GaN.  It might be the perfect newer technology for a  1 MHz power stage.

boB

[PCB.Wiz]

- with LT1739 that drive the transformer with 24+24 ohm series resistors the waveform is not bad but contains harmonics, the function generator signal degrades a lot. Maybe because I was forced to amplify 10 times (standard configuration), I don't know.

'not bad' and 'degrades a lot' somewhat contradict.
Maybe the LT1739 is going into high frequency oscillation with that load, or is simply clipping ?

I should test with a function generator that reaches higher in amplitude, because mine only reaches 2.5 Vpp (on 50 ohm).

You can also make a simple emitter follower buffer, to increase the drive / reduce the loading.
The simplest is connected bases, or you can add diode connected transistors to bias the bases to reduce crossover distortion.

[Neviolograpy]

Maybe the LT1739 is going into high frequency oscillation with that load, or is simply clipping ?

Clipping didn't, I could still go higher in amplitude; I lowered it to 2.5 Vpp to compare the dirt with what I (didn't) have using the function generator.

1. Any chance your test load is nonlinear? Maybe it generates its own 4th harmonics which only become apparent at higher magnitudes?

I also thought the same thing, so today I looked for a signal generator capable of generating a sine wave with greater amplitude to clear my doubts.
I found a cheap (but better this way for now) Multicomp MP750064 (DDS 14 bit 125 Msps). This can generate a maximum of 10 Vpp sinusoidal (measured on 50 ohm) so perfect for me to drive the primary of my transformer (with resonating cap). I measure 9.4 Vpp on primary and 40 Vpp on secondary, with sensor connected. Harmonics are present, but low, negligible. Let's say that if I could build a generator with the same performance on my pcb board I would be happy.

I was looking at the AD9954 + AD9954 combination, but I saw that it's not exactly simple to build a function generator with low distortion...
In any case I have to do the whole board and make sure a dual supply voltage of +-15 V for the power, 1.8 for the digital part (uc + dds) and another 1.8 for AVDD dds.

2. A coupling transformer may not be the best way to do the voltage ratio change; maybe you can use an impedance matching network to achieve adequate results? Maybe this network could be tuned to favour the target 1 MHz and reject other frequencies?

I don't know, thanks for the tip, I'll try to look... But how do you raise the voltage without a transformer? If you're telling me that the transformer I'm using isn't the correct one, ok that's probably true.

3. Are there ways you could make the testing coil (+resonant cap) less sensitive to harmonics? Maybe you could add some harmonic traps (eg an additional LCR series network tuned @ 4 MHz)

yes, I could add a filter to the final signal that attenuates the harmonics... but we are talking about 5-10 millivolts so I don't know if it's more the noise I'm introducing than anything else. I'll try!

[PCB.Wiz]

.... so today I looked for a signal generator capable of generating a sine wave with greater amplitude to clear my doubts.
I found a cheap (but better this way for now) Multicomp MP750064 (DDS 14 bit 125 Msps). This can generate a maximum of 10 Vpp sinusoidal (measured on 50 ohm) so perfect for me to drive the primary of my transformer (with resonating cap). I measure 9.4 Vpp on primary and 40 Vpp on secondary, with sensor connected. Harmonics are present, but low, negligible. Let's say that if I could build a generator with the same performance on my pcb board I would be happy.

I was looking at the AD9954 + AD9954 combination, but I saw that it's not exactly simple to build a function generator with low distortion...
In any case I have to do the whole board and make sure a dual supply voltage of +-15 V for the power, 1.8 for the digital part (uc + dds) and another 1.8 for AVDD dds.

What distortion and frequency control and stability do you actually need ?
A DDS gives you great fine control over frequency, and will give a degree of future proofing, but at some adder to price..
You could buy an AliExpress module and just add the Sine amplifier.

addit:
If you are fixed at 1MHz, another technique you can use, is what TI covers in this App note. It uses a twisted ring / Johnson counter to give Sine = CLK/16, with weighted resistor DAC.
A Si5351A/MS5351M clock synth could be used if you need to sweep close to 1MHz, at lower cost than DDS.
https://www.ti.com/lit/pdf/sboa550
Note their 0.1% ballpark harmonic numbers only go to the 8th harmonic, so fall short of the 16x clock sawtooth error component.
Still, that 16x element is more easily filtered.

[Terry Bites]

A DSL amp ic makes an excellent low cost driver

[boB]

"A DSL amp ic makes an excellent low cost driver"

Got a part number or company name ?

I would imagine these are less common these days but maybe not ?

[David Hess]

"A DSL amp ic makes an excellent low cost driver"

Got a part number or company name ?

I would imagine these are less common these days but maybe not ?

https://www.analog.com/en/products/lt1210.html
https://www.analog.com/en/products/lt1795.html

[boB]

"A DSL amp ic makes an excellent low cost driver"

Got a part number or company name ?

I would imagine these are less common these days but maybe not ?

https://www.analog.com/en/products/lt1210.html
https://www.analog.com/en/products/lt1795.html

Yes !   Perfect, probably.   Close as your going to find.

Might even be able to add a pair of current booster  transistors on the output if needed.

[David Hess]

Might even be able to add a pair of current booster  transistors on the output if needed.

If you are going to add a couple transistors for greater output, then a more common lower power operational amplifier can be used.

A composite amplifier with good performance at 1 MHz and higher would be an interested design project that I will have to try some day.  Unfortunately most of high performance transistors which would have been used in the past for this sort of design are no longer available.

[Zoli]

Might even be able to add a pair of current booster  transistors on the output if needed.

If you are going to add a couple transistors for greater output, then a more common lower power operational amplifier can be used.

A composite amplifier with good performance at 1 MHz and higher would be an interested design project that I will have to try some day.  Unfortunately most of high performance transistors which would have been used in the past for this sort of design are no longer available.

Rohm has some interesting transistors, like the 2SAR/SCR574/586.

[nfmax]

"A DSL amp ic makes an excellent low cost driver"

Got a part number or company name ?

I would imagine these are less common these days but maybe not ?

https://www.analog.com/en/products/lt1210.html
https://www.analog.com/en/products/lt1795.html

I have used a pair of LT1210's in bridge driving the primary of a coupling transformer, stepping up the output voltage to drive a piezoelectric ultrasonic transducer at about this frequency range. They work well.

You don't specifically say if you are driving an ultrasonic transducer. However, remember that the 'simple' equivalent circuit of such a transducer is a series R-L-C, where R represents the sound emission and internal losses, and L-C model its fundamental thickness resonance, all IN PARALLEL with a larger-valued Cp, representing the bulk capacitance of the transducer (i.e. as if it was just a ceramic capacitor). The impedance of Cp at the resonant frequency can often be much lower than R, so the bulk of your amplifier drive power is wasted pushing wattless current through Cp.

The classic fix for this is to place an inductor Lp in parallel with the transducer, so that Lp-Cp has a parallel resonance at the operating frequency. This greatly increases the impedance, reducing the load on the amplifier.

Without knowing the details of your application it's difficult to be more specific.

[David Hess]

A composite amplifier with good performance at 1 MHz and higher would be an interested design project that I will have to try some day.  Unfortunately most of high performance transistors which would have been used in the past for this sort of design are no longer available.

Rohm has some interesting transistors, like the 2SAR/SCR574/586.

Those look like ring/perforated emitter transistors because of their flat hfe versus frequency.

There are lots of options up to maybe 10 MHz, but 1 GHz NPN/PNP complementary power transistors used to be available.