The AM Forum

 In response to my Easy Class D posting, Steve, WA1QIX, suggested I use an analog modulator.
After days of experimenting, I came up with a solution based on a novel circuit called a SZIKLAI
circuit, which is similar to a Darlington pair. The SZIKLAI circuit does not introduce the
two-diode-drop penalty that a Darlington does.

The circuit shown in the attached pic file is a complimentary-symmetry amplifier based on
SZIKLAI circuits, and meets the following requirements:
1. output swing of 23 volts on a 24 volt supply, and
2. second and third harmonics down at least 50 dB, and
3. drive current requirement well within op-amp output range.

FYI, have fun, and see you on the air soon with my new modulator.
Rod

email me for the LTC spice file

I could not get the neg half cycles, so i changed around thge lower pair. Is that something wrong?

While it may suffice for modulator use, -50dB harmonic distortion is actually very poor for
audio amplifiers. A two stage solution with some loop feedback might knock that down another
20dB or more.

Trying to get a boatload of gain out of one stage in a practical amplifier, with zero loop
feedback usually ends up with a lot of non-linearity and possibly stability issues.

The usual application for these compound high gain connections is to boost the overall
loop gain so that more loop feedback can be applied.

                             _-_-bear

I amended the original post to include output waveforms,  and FFTs.

Patrick, I sent an email to you at your QRZ email address with a spice file. Yours looks good, I don't know what the problem is.

Bear, on the  air, I don't think I could hear harmonic distortion at 50 dB down.

Of course, stability in the real world is the proof of the pudding. I will let you know.

Rod

Bear, note the linearity in the original post.
Rod

Hello Rod,

You might want to change the FFT window settings as well the number of terms. The resolution of the harmonic values is a bit rough. Also, all devices in your simulation are exact identical. Nice, but not realistic. Get a feel for distortion sensitivity, go to the model of the BIP devices and perturb it a bit, save it with a new name and re run. See what happens. I did this with a mixer circuit and it was quite an eye  opener.

Alan

Yes, Alan, I agree.
I will have to hand-select the transistors to match Betas.
Fortunately  the transistors in use are dirt cheap, and I can afford
to buy a bunch of each.

Regards,

Rod

What is the gain in dB for this circuit?

Why not just move the pairs of transistors apart, use a few more parts
and end up with a more stable and predictable result. Hand selecting parts
is ok for yourself, but not really usable by others.

Anyhow, why not use more devices in the first place? They're small and
cheap!

The typical design for an audio amp/modulator gets the voltage gain in the front end
and the current gain at the output side... as you likely know.

                       _-_-

Bear,

Are you inclined to make up a circuit and model it?

Remember, 23V swing on a 24V power supply.

Rod

Alan

r.e. matching Betas

Without going through laborious modeling to confirm it, I suspect that if

UpperHalf(Beta1 * Beta2) = LowerHalf(Beta1 * Beta2)

that would be sufficient for Beta Matching. That greatly simplifies
hand-selecting parts. What do you think?

Rod

Hello Rod,

To Bear's point, this is a driver. So it has current gain, little voltage gain. Of course it provides the needed Z transfer. I assume the op amp prior will provide the required voltage gain and there is nothing preventing feedback added around the whole system.

On matching, you might consider this arrangement as the output devices are identical either NPN or PNP and improved  matching-balance of the output. So here is the Darlington and your circuit, so called also, the complimentary Darlington. Three diodes are needed to minimize crossover.  This arrangement may work out better with less effort involved in the device matching process.

Hi Alan,
Interesting variation on that circuit. Yes, indeed it would ease the task of beta matching.

What do you think about my Beta1*Beta2 conjecture above?

Rod

Hello Rod,

Beta matching is fine. Key is diode matching to the output devices. Preventing thermal runaway may require small R in series with the output device emitters and temperature tracking between the diodes and the output devices.

It may make sense to construct the diodes out of the output transistors; tie collector-to-base.

I was surprised that the crossover distortion appears so small. See what the breadboard produces.

You should do a sanity check on the simulation. I played with this configuration a bit and the crossover distortion was essentially non existent. I messed up the AB bias setting with intention to see the distortion and still not the case. Needs checking.

Hello Alan,

What a good idea " construct the diodes out of the output transistors; tie collector-to-base."!

And I will mount the "diodes" on the same heatsink as the output transistors. Thanks for the ideas.

I will play with the small-emitter-resistor concept.

Thanks for the time and energy you have put into this.

Regards,
Rod

Think the simulations need an "R" as the load to draw current.
Seem to recall that LTSpice "knows" about loads?

                          _-_-

Bear,
In the original circuit R7, 12.5 ohms, simulates the RF deck load.
Rod

Rod, in your first post, it seems to me that Q3 and Q4 have incorrect polarity, as the emitter on the pnp device is connected to negative (ground) and the emitter of the npn device is connected to the positive side of the circuit.  Is this the circuit you simulated? 

It would appear to me that the edited circuit by opcom shows Q2 and Q4 with the proper orientation.  I cannot understand how your original circuit would operate properly. 

If I am missing something obvious, could you please explain the circuit for all of us to understand?  Thanks!

Spot on Rick.

The Darlington-Complimentary pair upstairs should also be duped down stairs.

Not sure how spice played with this. Probably not... Maybe a cut and past error.

Thank you Rick and Alan. I see what you are talking about.

I turns out I was seduced into a senior moment by looking at circuits
such as the one shown in the file example.gif.

Having been dazzled by the stellar results of my original post,
I published it without having critically analyzed the circuit.

See the circuit in the file pnp-only.png. Therein I disconnected the lower
transistor and the results were the same.

So, it turns out the circuit is just a very good class A amplifier!

Many thanks to Alan and Rick who DID do more critical analysis!

I will go ahead and use just the upper half in my modulator!

Regards,
Rod

Good article about a simple solid state class A power amplifier....

http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1996/Class-A%20power.pdf

I wonder if this circuit has ever been modeled?

Jim
Wd5JKO

Alan and Rick,

Yes, I know the 2SAR533P will not handle the dissipation. I just used
that transistor for convenience of modeling.

In real life I will use a TIP42, rated 65W at Tc=25C.

Rod

Rod,

I do not see how you are going to sustain a desired peak to peak sinusoidal swing.
You have source capability but no sink. Make sure you do a run command in spice as your alter circuits.

Alan,
I am not sure I understand your comment.
Take a look at the files associated with this post,
showing circuit, output swing, and fft.
Regards,
Rod

It appears what we have now is a single-ended class-A series modulator. 

If we were driving a load with AC coupling, referenced to ground, (or with equal positive and negative rails instead of a single positive supply) then the complimentary sink circuitry would obviously be beneficial.

In the case of the latest circuit, the output of the transistor pair not only provides the AC modulation swing, but it also is the source of DC supply for carrier, whether or not it is modulated.

With the single-ended, capacitively-coupled input, the diodes are no longer needed to set the bias for minimum crossover distortion.  Just a resistive divider to set the operating point, providing the maximum linear positive and negative modulation excursions, while maintaining the desired DC level for setting the carrier level.

With this output circuit load configuration, I do not see any need for the sink.  Modulation percentage is limited to the linear portion of the transfer curves of the devices used.

Rod,

It is fine. I miss interpreted the disconnected schematic. Actually the single complimentary Darlington pair provide a source-sink follower action in their own configuration. The output signal swing looks ok.

As a class a series mod, you're going to get 50 pct voltage swing on the device being modulated....  Using a standard darlington that's the best I could do.

Maybe this configuration will do better, but..   

--Shane
KD6VXI

Shane,
Not sure I am following what you are saying. Take a look at the files associated with post #22..
getting full swing on the load.

Rod

What I am saying is:

With an non / non darlington pair, I can only get 1/2 Vcc swing under real world conditions.  As has been discussed elsewhere (recently) in the qso section. 

If the non / pnp darlington pair gets actual near full Vcc swing then this is amazing!

I don't always trust modeling.  Not saying it won't work, I really hope you get near full Vcc swing.

In a series mod with 2N2222 and MJE3055 I see half Vcc swing.

Because of this inefficiency I went PWM.

--Shane
KD6VXI

Excellent point Shane.

Rod,

One item I believe is missing from the simulation is associated with the AC source.
There is no source R attached. Or I do not see one in the screen shot. Rs is zero. It is set up as an ideal source. Although the input R for the device pair is high, there will be an input signal voltage divider. The consequence is reduced output swing. Re check that aspect.

Attached with sweeps on Rs from 1 to 100 ohms with a step through 50 ohms. You can see the consequence.

Alan

Yes, Alan, I agree. One needs to adjust the driving voltage to compensate for driving source resistance.

I believe Shane is talking about something else.

Shane, could you post your 2n2222/MJE circuit, showing the RF deck equivalent resistance (Vdd/Idd)?

Thanks,
Rod

Cool....

I built the experiment. All total, 5 minutes. See attached.

No curve tracer matching but required careful setting of bias control.
The Vcc is set at 24 V, the swing is 24 V. Now I confess, I do not have a fancy 6 digit readout scope... So counting scales on my Tek scope. But indeed, careful set the bias and supply is swung, 0-to 24 V. I'll try to post an ugly board pix and scope shot.

Initially setup was no bias tweak. In that case, I lost 4 V in swing. So yes, careful device selection or bias adjustment would be required to get complete swing. The voltage gain for this arrangement is 0.89.

The scope at 5 V/major div. The channel 2 set to 24 V top line. Bottom gnd.
The ugly card is it at 1 kHz. Yep... ugly.

Alan

Alan, good job! Ugly is the way to fly!
Rod

Here is the final version, showing circuit, Vswing, Vfft.  See attached files.