The AM Forum

I managed to get my hands on a nice heat sink, metal chassis and 120 volt/12 amp transformer.  I also already have a couple FQA11N90's and just need to variable caps.  I've borrowed some ideas out there and put together this schematic for a class E single FET transmitter with external drive.  I used Xl (inductive reactance) numbers of 260 ohms for the input inductor and 450 ohms for the output series inductor. 

How far off am I?

Jon
KA1TDQ

Jon

I never tried using my 1-FET class E transmitter on 40m. However, I suggest that you use a toroidal transformer with only 1 turn on each side. [There is another way to feed the DC to the drain of the FET; but a 1:1 transformer will work fine.]

When you have four or six FETs in parallel, you need a 1:2 transformer to match the optimal load impedance of the FETs to the impedance looking into the tuned output circuit. But, with a single FET, a 1:1 transformer will provide the proper match.

More details:

The optimal RF load impedance for the FET(s) is approximately 0.5 x the drain voltage at carrier / the total drain current at carrier.

With four FETs in parallel, the total drain current is 4x the drain current of a single FET.

Therefore, with four FETs in parallel the optimal RF load impedance is 4x smaller; and you need the 1:2 step up transformer. [A 1:2 step up transformer provides a 1:4 impedance transformation]

Even more details:

If the drain-to-source voltage, at carrier, is 45V, and if the total drain current, at carrier, is 1.5A (corresponding to a single FET), then: 0.5 x 45V/1.5A = 15 ohms. This is approximately the load impedance that the FET needs to look into. The parallel combination of the adjustable loading capacitor and the 50 ohm output load produces an effective resistive load of approximately 15 ohms... to match the load that the FET needs to look into.

With four FETs in parallel (45V drain-to-source voltage, and 6A of total drain current at carrier), the optimal load impedance, that the parallel FETs need to look into, is approximately 0.5 x 45V/6A = 3.75 ohms. The 1:2 step up transformer converts this to 15 ohms... which is easier to match with the adjustable loading capacitor in parallel with a 50 ohm output load. E.g. it reduces the current in the output network's tuning inductor... which translates into reduced heating losses in the tuning inductor.

Stu

Ok, 1:1 sounds good.  I'm trying not to get too complicated for my first rig.  I want to build something that actually works.  This approach uses minimal components with stuff I mostly have. 

I ordered the electrolytics yesterday for the power supply and I might put that together this weekend.  I found a massive commercially made isolation unit for 120 volts in a Goodwill store.  I figure I'll just rectify the output and get 120 vdc.  For testing the transmitter, I'll just use a tap from a lower electrolytic capacitor in the bank. 

Jon
KA1TDQ

Jon

The 125pF capacitor should be placed between the drain and the source. It's purpose is to supplement the transistor's intrinsic drain-to-source capacitance (around 300pF) to produce a total capacitance that keeps the source-to-drain voltage from getting too large on peaks of the drain voltage waveform. For 40m operation, and a single FET, 125pF is about right.

Note that this capacitor needs to handle a lot of RF current. A doorknob type capacitor is what is typically used.

The bypass capacitor between the top side of the transformer and ground (where you currently have 125pF) needs to have a low impedance to ground at 7MHz. It should have a value of around 0.05uF or more, and rated at 400V or more. There are "orange drop" capacitors that can be used for that purpose.

Okay, so the 125pf doorknob is essentially in parallel with the FET (one lead on the source and the other on the drain).  I understand that this is part of the resonant circuit supplementing the FET's internal capacitance.

 Then on top of the transformer where the HV DC comes in I need to put something around a .05uF.

Do I have that right, or am I having another GCE?  (GCE was on my Navy nuclear exams for "Gross Conceptual Error").

Jon
KA1TDQ

Jon

Yes, you have it right, except I wouldn't describe the 125pF capacitor as "part of the resonant circuit".

For roughly half of each cycle (0.5 x 140ns): the FET is turned off. During that time in each cycle, current flows into the total drain-to-source capacitance and then back out again. Therefore, as a rough approximation, the capacitance is being charged for 1/4 of each cycle, and discharged in the next quarter of each cycle. Then the FET is turned on for half a cycle; and the capacitor remains discharged during that time.

The average value of this (time varying) current, during the portion of each cycle that the drain-to-source capacitance is charging, is roughly equal to the overall average drain current: 1.5 ampere (at carrier).

The charge (Q) that accumulates on this capacitance during each charging (quarter) cycle is: 0.25 x 140ns x 1.5A = 52.5 nC. (i.e. 52.5 nanocoulombs)

But: Q = CV, and C= Q/V

If (conservatively) we want to keep the peak drain-to-source voltage (at carrier) below 12.5% of the 900V rating of the FET (to ensure that the peak drain-to-source voltage, on 100% modulation peaks, is less than 25% of the 900V rating of the FET), i.e. 112.5V, then we need to make sure that the capacitance is larger than 52.5nC /112.5V = 466pF.

Therefore we need a total drain-to-source capacitance (including the 300pF intrinsic drain-to-source capacitance of the FET + the supplementary drain-to-source capacitance) of 466 pF

A supplementary capacitor having value 125pF is close enough (-:

Stu

Jon:

I built a couple of these years ago:

http://www.its.caltech.edu/~mmic/reshpubindex/papers/QST.pdf

I now believe in push-pull as the derived output waveform is closer to a sinewave than a single ended one.

73,
Dan
W1DAN

Stu has a lot of very good advise.

Make sure the leads from the shunt capacitor (your 125pF cap) are SHORT and HEAVY.  Otherwise, there will be ringing in that part of the circuit.  The cap has to be a good one.  Not all doorknobs are created equal.  Some (probably the majority) are actually pretty poor.  Trust me, I've done a *LOT* of experimenting with shunt capacitors over the past 15 years or so with class E rigs.  Multilayer ceramic capacitors work much better in the shunt cap application (assuming the cap is rated at the current you are running).  Just make sure the capacitor is rated at the current you are going to run.  Since this is a fairly low power affair, you're probably ok with most doorknobs.  As the power goes up, MLC caps are a better choice.

------>>>  Important:  You should have a transzorb from gate to ground.  A 1.5ke18ca will do it.  Otherwise you stand a very great chance of blowing the MOSFET by exceeding the gate voltage, even for a nanosecond.  That's all it takes to punch through the gate and it happens all the time with analog drive for a variety of reasons.

40 meters has been pretty good lately with good activity around much of the time.

Short leads on the MLC shunting capacitor and a transzorb on the gate... got it!  I'll also use a variac on my power supply to bring things down to testing level. 

I have two crystals now that I can use on 40 meters, 7.289 and 7.294 MHz. I'll be back on the air after my antenna switch comes in next week.  I found out that my transmit dipole receives much better than my receive antenna, despite all efforts to bring it up to snuff.  Now I'll have a TX/RX switch and an antenna switch to move during QSO's.  The sequence is very important not to get wrong. 

Jon
KA1TDQ

Making progress.  The chassis seen in the picture will house the rig along with a PWM.  The single FET is an FQN11N90 and the top rail will be ground for the modulator 2 FETs.

The electrolytic board is for the rectified 120 vac from a huge isolation transformer.  I will cut down the coil to about 9.6uH to form the series inductor on the output. 

I'm going to a hamfest tomorrow to look for a HV 300pf variable.

Jon
KA1TDQ

What would be the approximate value to use for an FQA11N90's gate capacitance at 40 meters?  I'm trying to calculate the series inductor value to use with an analog drive.  The datasheet says 2700pf @ 1.0 MHz, but this value seems a little high.  I'm going to use a T80-2 toroidal core for the series inductor.

Also, copying someone else's work, should I stick with a 10:2 transformer on the input using an FB43-1020 on 40 meters rather than on 75?  He used 5 turns to 1 (1 turn going on the gate). Should I cut that back to maybe 4 turn to 1 for 40 meters?

This weekend I'm going to start tap/die parts to the heat sink and begin soldering.  I want to eliminate as much guesswork and experimentation as possible.

Jon
KA1TDQ

Jon

If you are going to use a toroidal transformer to drive the gate of a single FQA11N90, then you should model the gate input impedance as approximately 3000pF and series with 2 ohms of resistance.

You will need to add a series inductor whose impedance at your chosen frequency of operation is approximately equal (but opposite in sign) to that of the 3000pF of gate-to-source capacitance. At 7.3MHz, the impedance of the gate-to-source capacitance is approximately -j7.26 ohms. Therefore, for operation at 7.3MHz, you will need approximately 158nH of series inductance. If the inductor is placed on the input side of a 5:1 input transformer, then you will need 25 x 158nH = 3.95uH of inductance

If the impedance of the inductor cancels the impedance of the gate-to-source capacitance, then you will be left with the 2 ohms of series resistance. The 5:1 (or 10:2) toroidal transformer will step this up to around 50 ohms.

The output voltage of the toroidal transformer (driving the gate of the FET) will be (1/5) x the input voltage to the toroidal transformer. You need a 10 volt (peak) sine wave from gate-to-source of the FQA11N90.

However, because the gate-to-source voltage is the voltage across the gate-to-source capacitance in series with 2 ohms of resistance; and because of the Q of the resonant input circuit (including the inductor): the voltage from gate-to-source of the FQA11N90 is V x (1/5) x Q; where V is the amplitude of the sine wave applied to the input side of the circuit (including the series inductor and the 5:1 transformer), 1/5 is the transformer's turns ratio, and Q is around 7.26 ohms/2 ohms. Therefore, you will require V= 10 volts x 5 /Q = 13.8V.

At 50 ohms impedance ratio, this corresponds to 1.9W of drive power.

As an alternative, you can drive the FQA11N90 with an IXYS IXDD414. This driver chip can be placed in close proximity to the FQA11N90.

http://www.classeradio.com/driver.htm

The advantages are:

The IXYS chip is designed to produce its output voltage into a high capacitance load (like the FQA11N90's gate-to-source capacitance)... and its output voltage will have a faster rise time (crossing the 4V off-to-on gate voltage threshold of the FQA11N90) than a 7.3MHz sine wave... which is highly desirable in this application.

The IXYS chip can be driven by a CMOS logic gate or by a sine wave + a 1.25VDC offset. The input to the IXYS chip need not deliver a lot of current, because the input capacitance of the IXYS chip is the same as that of a standard CMOS logic chip.

Good luck
Stu

Thanks for the detailed explanation.  3.95uH of inductance is easy to get from a T80-2 and I can put a few more turns on, then back off a few to get resonance.  And I think I will stick with the analog drive for simplicity.  My current exciter only puts out 4 watts, but I can at least tune for the peak using it until I get a nice commercial rig with adjustable drive.  I can then adjust excitation until I get the proper 10+ volt peak at the gate.

Besides, the circuit we're talking about is too elegant not to use.  To get the matching gate drive with a corresponding 50 ohm match with two components just seems right. 

...and, transzorb the gate so as not to punch a hole through it.

Jon
KA1TDQ

Jon,

   Be very careful with that analog drive. This is not something you can raise up from zero to the desired level. If you do, the FET will go through the linear region, and the result will be high dissipation. Prolong that for more than an instant, and the fet goes south to dead fet heaven.

   We repair commercial 3KW 13.56 Mhz class E amplifiers here. These use RF drive with a tuning adjustment to get ~ 13v + peak to each of two parallel connected FET gates. Had a new guy that mistook 13v peak as 13v peak to peak. Two $70 Fet's made a nice thunk-thunk as they hit the bottom of the trash can.

   I wonder if a Retro-40 (Small Wonder Labs) transceiver would be a good candidate as a driver? These use a class D Fet output with digital drive.....The Xmitter is OFF/ON, no in-between..

Regards,
Jim
Wd5JKO

How about this as a driver?  It's just a CMOS inverter/oscillator before a digital drive chip.

Jon

I would expect the digital drive to be simpler.   Very low drive impedance and of course very broadband.  I would imagine that the limiting factor as far a frequency goes would be when you run out of current to charge and discharge the gate capacitance of the FET.  With the gate capacitance a low impedance current source and sink is required which the digital driver provides.  In going from 75 to 40 meters on a multiple FET design, you will likely be able to drive less FET's with a single digital driver when compared to 75M.  Since your only driving a single FET a single digital driver should be extremely simple.  

I believe the driver chips are TTL levels, but have not researched all of the devices available.  

Joe, W3GMS  

I have never looked at the input spec's of the digital driver chip.  You may have to use a buffer amp between your cmos oscillator and the drive chip.  That certainly is no big deal though! 

Joe, GMS       

Must also insure that the FET receives a logic zero at the gate when the crystal is pulled.

I didn't see any specs on the IXDD614 datasheet, but I'm willing to bet that the input impedence is very high. If need be, I can add a voltage follower in between.

I was checking out threads on the Class E forum and I noticed that someone had a rippled waveform going into the gate and the problem turned out to be a voltage regulation issue on the 12 volts for the IXDD.  I will probably use a car battery to supply the digital drive for testing and simplicity. 

Jon
KA1TDQ

Also have a look at Jay's (W1VD) website...
Yes it's Class-D, but the driver section is the same as for Class-E

http://www.w1vd.com/375wattclassD.html

in one link he has a VFO design, that will directly drive the IXDD614's in the RF deck

"digital drive" makes things a bit easier... and much less worry about the more likely parasitics
that can happen with analog drive....

Yeah, I'm definitely going with the digital drive.  I've seen that VFO schematic but I think that I'm going to stick with the crystal oscillator design.  Also I'm going to switch gears and make this a 75 meter class E rig.  I do have a crystal for 3870 (the west coast AM frequency). 

Everything for the digital drive will fit on a Radio Shack circuit board.  I've attached a picture and it shows the 5 volt and 12 volt regulator, CMOS hex inverter, voltage follower and IXDD614.  It will mount right next to the single FET.

This rig will only be 35 watts-ish but would be great for SOTA activations.  Lots of mountains around here.

Jon
KA1TDQ

That circuit with the crystal oscillator didn't work too well so I'm back to analog drive for 40 meters.  I've got the layout for the heat sink done with the modulator FETs on top and the FQA11N90 on bottom.  A BNC jack takes the RF drive and sends it through the T80-2 and FB-43-1020 to the gate. 

I need to make a DigiKey order for the components that will be associated with the FETs. 

Jon
KA1TDQ

Here is just one example of a high speed comparator (about 10ns switching time) that accepts an analog input and provides TTL output (to drive the IXYS chip), and comes in a DIP package.

http://datasheets.maximintegrated.com/en/ds/MXL1016.pdf

I used a comparator like this (I don't remember the part number of the particular comparator I used), with a dual voltage power supply (+5V and -5V), in the last version of my single FET 75 meter class E transmitter.

The sine wave input to the comparator could be as little as 10mV in amplitude (i.e. 1 microwatt into the 50 ohm non-inductive swamping resistor that I used as a termination at the input to the comparator); although I used around 100mV (0.1 milliwatt into the 50 ohm swamping resistor) from the unmodulated, low power RF output of my Flex Radio SDR-1000.

Be careful not to use too large a signal at the input to the comparator. For example, if you use a 1W input signal (10V amplitude into 50 ohms)... you will need to employ a low power resistive voltage divider at the input to the comparator to keep the input voltage below around 1V peak.

The TTL output of the comparator was a decent square wave when I used either a 3.8MHz or a 7.3MHz input.

In summary: the RF signal source provides a 100mV (value not critical) sine wave to the comparator => the comparator provides a TTL square wave to the IXYS chip => the IXYS chip drives the gate of the FQA11N90.

Stu

Jon,

Its very important for your learning curve to find out why something does not work.  Just switching topologies when the other one did not work leaves you with nothing learned.  Look at the circuit and find out why it did not work.  When you have your answers, then you will be in a much better position to make a decision on what way you want to go. 

We all know that digital drive is a very proven technique and it works very well.   

Joe, GMS   

I like the idea of using a voltage comparator with a small sine wave. The reference voltage would be ground and I would need a + and - 5 volt source.  The bottom half of the sine wave would slam the output voltage to -5 vdc, but I guess that's ok for the IXDD.  It doesn't care if the input is -5 or 0 volts for "off."

I'll try that... thanks!  I can use the BNC that I've already soldered on the board to input the oscillator waveform to the comparator. 

I built the circuit I mentioned a few replies ago and the oscillator would change frequency a lot simply by touching it. 

Jon
KA1TDQ

Jon

You wrote: The bottom half of the sine wave would slam the output voltage to -5 vdc, but I guess that's ok for the IXDD.  It doesn't care if the input is -5 or 0 volts for "off."


No, that is not correct.

The output of the comparator will be one of the two standard TTL logic levels. Logical "low": between 0V and 0.4V. Logical "high": between 2.6V and 5V.

The reason for using a dual voltage supply (+5V and -5V) is so that the signal applied to the "+input" of the comparator can be a sine wave, whose values are both positive and negative. The "-input"  of the comparator should be connected to a small positive voltage relative to ground... for example: +0.1V obtained from the +5V supply by using a voltage divider between the +5V supply and ground. When the voltage applied to the "+input" (relative to ground) exceeds the voltage applied to the "-input" (relative to ground) ... for example: when the sine wave has a value that is greater than +0.1V... the output of the comparator will be TTL logical high. When the voltage applied to the "+input" is less than the voltage applied to the "-input"... for example, when the sine wave has a value that is less than +0.1V... the output of the comparator will be TTL logical low.

Since the input sine wave will be AC-coupled (i.e. via a capacitor) to the "+input"... and since there will be a 50 ohm non-inductive resistor between the "+input" and ground... removal of the sine wave (turning it off) will cause the "+input" voltage to go to 0V; and the output of the comparator to go to TTL logical low.

Stu

That I can do.  Thanks for the oscillator picture too.  I'll just copy the Retro's oscillator and apply it to a comparator (with the .1 volt reference and 50 ohm resistor). 

I'm making a parts order to Digikey next week, so I'll tack on what I need.  The other stuff is heat-sink mount components. 

Thanks, I'll let you know how I make out!

Jon
KA1TDQ

Update:  I built the oscillator board and I'm getting about a 1 volt pp which will be used to drive the comparator.  I will add a fixed and variable resistor to the comparator (-) input so that I can fine tune the reference for the duty cycle. 

Jon
KA1TDQ

Jon

Good!

As a clarification:

In this case, is not necessary to have a 50 ohm swamping resistor at the point where the oscillator connects to the comparator. The 50 ohm swamping resistor (and an associated RF voltage divider/ RF power attenuator) is needed in the case where the RF source is the output of a transceiver or a transmitter that is designed to feed into a 50 ohm load. In this case, you can just connect the oscillator to the comparator with a twisted pair... one wire of which is connected to ground at the location of the oscillator and at the location of the comparator. Make sure that the oscillator is "AC-coupled" to the comparator. I.e. somewhere between the output of the oscillator and the "+input" of the comparator, there should be a capacitor in series. The capacitor can be located directly at the output of the oscillator or located directly at the "+input" of the comparator. A 0.001uF capacitor will be fine. The value is not critical. Looking at the picture you posted, you may already have an AC-coupling capacitor, in series, at the output of the oscillator. In that case, you don't need to add another capacitor.  

On the comparator side of the AC-coupling capacitor, place a resistor between the "+input" and ground. The value of this resistor should be around 1000 ohms. The value is not critical. The purpose of this resistor is to force the "+input" of the comparator to be at 0V relative to ground when the oscillator is turned off (i.e. not producing any RF output). This resistor will be part of the RF load on the oscillator output. A resistor having value of 1000 ohms should be no problem... but you could use a 10,000 ohm resistor instead.

When the comparator is powered on (i.e. when both the +5V and -5V supplies are present on the comparator's power pins) you can verify that the voltage between the "+input" pin of the comparator, and the comparator's ground reference, is a sine wave (not larger than 1V peak) when the oscillator is turned on, and 0V when the oscillator is turned off.  

Stu

Progress update:  Oscillator board is done (left of the BNC connector) and the digital drive board connected to the gate is done too.  Blue connectors allow for unregulated power inputs of gnd, +5, -5, and +12 volts.

The PWM section is on top with another Radio Shack board that will sit on the right side of the FET.  That will be the PWM output board and I will get a PWM generator board from Steve to complete that area.

That leaves the power supply, PWM filter and output RF section.  Still a long way to go, but getting there.

Jon
KA1TDQ

.

You may already be planning it...

But you must have a heatsink for the IXYS driver IC.

it is going to get very hot at 40m operation.


Also, I see a lot of grease, but is the FQA11N90 insulated from the heatsink?
The the metal backing on that is connected to the drain, not source.

The big diode, and modulator mosfet would have to be insulated as well.

There are mica insulators on the back of everything, so the drains and diode are insulated from ground. 

And actually this'll be used on 75 meters... the crystal is for 3.870 MHz.

Yeah, a heat sink for the IXDD...  didn't think of that.  It's an 8 DIP socket.  Is there a heat sink product out there for that? 

Jon

Oh, thought it was the normal to-220 package.
I didn't even look close enough to see there were no devices with 5 leads.

Yes, there are heatsinks for DIP packages...

http://www.digikey.com/product-search/en?pv357=307&pv357=249&FV=fff40012%2Cfff80068&mnonly=0&newproducts=0&ColumnSort=0&page=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=25

I don't know how that one will act... I haven't tried one.
Though I suppose it should be the same.

When you DO build one for 40m, I would definitely use the to-220 package, so
you can attach it directly to the large heatsink.

I had problems working with the digital drive and decided to give the analog drive a try.  You can see the schematic on the computer screen in one of the pictures along with the waveform at the gate.

I'm driving it on 40 meters with 4 watts and I'm getting about 8 volt pp.  L1 is calculated for 200 ohms Xl at 4.37uH.  T1 is on a FB43-1020 and is 5 turns primary to 1 turn secondary at the gate. 

I didn't get the class E waveform at the gate like I expected, and it was just a regular sine wave with negative excursions for half the wave. 

I have Transorbed the FET everywhere too so I'm safe for experimentation.

I'm not sure where to go from here.

Jon
KA1TDQ

Here's a clearer picture...

no, you won't see a class-e waveform at the gate...
But the gate waveform will change when the fet is under load.


see picture... top right

http://classeradio.com/driver.htm

Oh, ok.  I guess then that I need to work on getting at least 24 volts pp out of the transformer.  I'll tune L1 to resonate with the gate for max voltage and then adjust drive as necessary.

I haven't added the shunt capacitor yet, so I can experiment with class C first.  All I need to do is take the RF out of the output transformer.  I have a 130 volt DC supply and I'll just variac the input to that for the lower DC.  I'll start really low, around 12 volts, and work up to 35-45-ish. 

Jon
KA1TDQ

Ok. I increased the drive to 10 watts.  I then kept pealing turns off the toroid until I ended up with 24 volts pp.  I wound up with just 10 turns on the T80-2, but I got there. 

Next, I'm going to hook this up to some DC and see what I get for RF output.  For a more compact unit, I could leave it class C (no need for variable capacitors and a coil).  I could use this to replace my single tube, low-level transmitter.

Jon
KA1TDQ 

I found a junked 32 volt printer power supply to try the RF board out, and it puts out 22 watts.  Since this is configured for class C, I used a 1:2 output transformer at the drain of the FET.

I'm going to get an isolation transformer so that I'll have 130 volts-ish DC to play with and add a PWM.  I'm going to etch my own PWM output board for my customized setup, and I'll get the PWM generator board from Steve.

I'm going to need help designing the PWM filter though.  I haven't taken any measurements, but I did calculate once that this would consume .72 amps at carrier.  From there I haven't messed with any of the online filter calculators. 

Jon
KA1TDQ

I decided to go with a traditional mod transformer design for this transmitter too.  I found an article on the web using a 120/40 volt power transformer as the mod transformer for a class D RF deck.  I've attached a picture showing the transformer with a negative peak limiter found on the AMwindow.

The other picture shows the transformer along with a negative peak limiter I've put together.  This circuit uses a jumper that I can take in/out for testing.  Also I've put in an LED to show if/when negative peaks show up.  The anode is connected to ground with the cathode connected to the mod transformer side going to the FET. 

I'm going to use a 30 volt transformer for the DC supply.  I will measure the DC voltage drop under heavy modulation and set the zener for 2 volts lower than that.  That should account for the .7 volt drop across the diode and any variance in zener voltage.

Jon
KA1TDQ

Jon

The Antek transformer will saturate if you allow significant amounts of unbalanced DC to pass through it. I think the intent of the design you found on the web is to try to pass DC through both the primary and the secondary sides... in amounts that will balance out the DC magnetic field in the core. It's an interesting approach, if done right. It's not very energy efficient. The balancing current will dissipate power in a resistor, comparable to the average electrical power delivered to the RF stage. The resistor will also place a load on the audio amplifier that consumes as much power as will be consumed by the RF stage. If you don't include the balancing current on the primary side, the modulator will not work.

The negative peak limiter that you propose will make the audio amplifier very unhappy... because the amplifier is a voltage source that doesn't like to look into a short circuit.

Stu

According to the schematic I found, it looks like only audio is passed through the 40 volt windings.  To keep things balanced, where does the other DC come from to keep things balanced?

If the negative peak limiter won't work, a least the diode to give an indication of a negative peak should be fine.  I would probably need to come up with another way to limit negative peaks.

Jon
KA1TDQ

Jon

There are a variety of hypothetical modulator designs that endeavor to use DC flowing in an auxiliary winding to work around saturation of the mod transformer core resulting from the average current flowing into the modulated RF stage. Some of them won't work; and most of the rest will dump half of the DC power and half of the audio power into a resistor. Some designs will dump more DC power into a resistor; but will dump less audio power into that resistor. Some designs employ a reactor of sufficiently high inductance... to produce a sufficiently high impedance at audio frequencies... to avoid dumping DC power and audio power into a resistor. [These last designs can be compared to the modified Heising design... but they employ a separate, auxiliary, mod transformer winding]. All of those designs require the balancing current in the auxiliary winding to be adjusted to cancel out the magnetic field produced by the DC flowing into the modulated RF stage.

For the hypothetical design that I think you were looking at, B+ is connected to one side of the primary, and a resistor (roughly 16 ohms) is connected from the other side of the primary to ground. The direction of the DC in the primary has to be chosen to cancel the magnetic field produced by the DC flowing through the secondary. The value of the DC flowing in the primary has to be the value of the DC flowing in the secondary x the primary-to-secondary turns ratio. This also assumes the the output of the audio amplifier blocks DC.

If you wish to use a modulation transformer that cannot handle unbalanced DC (like the Antek transformer) you should use a "modified Heising" design. For your application you will need a 0.25H Heising reactor capable of handing 1.5A of current. You will also need a 100uF electrolytic capacitor rated at 450V.

Stu

Modified Heising... got it!  I can do that, and thanks for the component values. 

I have the new schematic in the picture below.  As you mentioned before, this is assuming that the audio amp can handle the transformer load without coupling caps.

I've added the negative peak limiter/indicator around the choke now.  How does it look?

Jon
KA1TDQ

Jon

On each side of the transformer (the primary side and the secondary side) the two available windings have to be wired in series or in parallel with the correct relative polarity. The placement of the dots in your schematic puts the two available windings on each side of the transformer in series opposition. I.e. current flowing through the series-connected input windings will produce magnetic fields that cancel. The output windings, as you have connected them, will produce equal and opposite voltages that cancel. The correct series interconnection would have the side of one coil that is marked with a dot connected to the side of the other coil that is not marked with a dot.

If the audio signal at the output of the transformer goes 2 volts more negative than: - (the DC supply voltage), the LED will instantly burn out (you will get one "negative peak" flash... and that will be it). The Zener diode will, as I stated previously, place a differential resistance (dv/di) of essentially zero ohms across the amplifier... and the amplifier will activate its automatic protection circuitry (or worse).

I suggest that you look up Steve's (WA1QIX's) approach to negative peak limiting, using 3 diodes.

http://www.amwindow.org/tech/htm/3diodeka.htm

In this application, the negative peak limiter's "keep alive" supply should have a value of 5V. The resistor, R1, ... which becomes the differential load on the modulated B+ line, when the diodes switch in the 5V keep-alive supply... should have a value (ohms) roughly equal to following:

R (modulation) = B+/I(carrier), where I(carrier) is the drain current that flows into the FET at carrier level.

For example, if the B+ is 30V and the drain current, at carrier, is 1A... then R(modulation) = 30V/1A = 30 ohms.

Separately, to be more precise about the values of the Heising choke and the Heising capacitor:

The value of the Heising choke should be roughly 0.25H x [R(modulation) / 50 ohms].

For example, if R(modulation) is 30V/1A, then the Heising choke should have a value of roughly 0.25H x 30 ohms/50 ohms = 0.15H.

The Heising choke should be rated to be able to support a current equal to (or greater than) I(carrier).

The value of the Heising capacitor should be roughly 100uF x 50 ohms/R(modulation).

If R(modulation is 30 ohms), then the capacitor's value should be roughly 100uF x 50 ohms / 30 ohms = 167uF.

Stu

Very good stuff.  Thanks.  I'm sourcing all the parts now and will get back in the not-too-distant future.  

...and as for the placement of the dots and the phase relationship of the Antek transformer, I drew that schematic quickly and from memory just to represent the transformer.  I will place the winding in parallel and in phase like the original schematic shows.


Jon
KA1TDQ

One more quick question.  I built the keep-alive circuit last night using diodes with values recommended in the article write-up.  However, is there any benefit to using a Fast Recover Epitaxial Diode as the series diode?  It has faster switching time and less loss, from what I gather.  Instead of a .7 volt drop going to the FET, I could use more voltage.  Or, does that mess up the negative peak limiter portion since there's a voltage drop intended for both sides of the series diode. 

I have an IXYS diode that I could easily put in there if there's any benefit.

Jon
KA1TDQ

Jon:

I think that using a series diode with a lower forward voltage drop (v. a 1n4007) would be fine... provided it can handle the peak forward current.

Stu

I now have all the major components to finish the rig.  I'm going to use a cookie sheet as a mock-up to make sure the transmitter works before I put everything all together in a nice chassis.  From the cookie sheet picture you can see the layout:

Left:  160 uF capacitor/.15 H choke for the Heising circuit
Left near RF deck:  3-diode keep alive circuit/5 volt regulator
Center:  Antek modulation transformer
Right:  28 volt transformer for carrier and 6 volt transformer for the 3-diode keep alive circuit
Center-top:  40 meter RF deck

I've tested the output power using the 28 volt transformer and I'm getting 20 watts. 

Jon
KA1TDQ

Jon

Okay!

Please read before proceeding:

1. Start by re-testing the transmitter with the 0.15H choke in series with the B+... but with the 160uF capacitor and the Antek transformer not yet connected. Verify that the transmitter still works okay, except that the DC drain voltage will be lower, due to the series resistance of the choke.

2. Next, add the 160uF capacitor (double check the polarity... the + side should connect to the drain of the FET) between the drain of the FET and one end of the secondary of the Antek transformer. Verify that the two primary windings of the transformer are wired properly, and the two secondary windings of the transformer are wired properly (i.e. + of one winding to - of the other winding, if one winding, of a pair of windings, is in series with the other winding, or + to + and - to - if the windings, of a pair of windings, are in parallel). Connect the unconnected side of the secondary winding of the Antek transformer to ground. Leave the primary winding of the Antek transformer disconnected (no audio amplifier connected yet). Verify that the transmitter still works.

3. Next, turn off the transmitter. Hook up the audio amplifier to the primary winding of the Antek transformer. Keep the output level of the audio amplifier set to a low value. Turn on the transmitter (first), then turn on the audio amplifier. Then verify that the transmitter is putting out the expected carrier level RF power. Then talk into the microphone, and adjust the amplifier's audio output level to obtain the desired modulation index.

4. Additional things to do (okay to wait until after the initial testing):

a. Add a bleeder resistor (wire wound is okay) across the 160uF capacitor... with a value of around 10,000 ohms (i.e. 160uF x 10,000 ohms = 1.6 seconds, so the capacitor will be mostly discharged 5 seconds after you turn off the power supply). The dissipation rating of this resistor should be 1W or more. Without this bleeder resistor, the 160uF capacitor may stay charged to the B+ voltage long after you have turned off the power supply (depending upon whether the existing bleeder resistor, across the output of the power supply itself, is still connected between the 0.15H choke and ground, when the power supply is turned off).

Keep us informed of your progress.

Stu

Thanks for the testing procedure. I'll do just that.

The capacitor is non-polarized (and not cheap either... $67 from Digi-Key). 

I like this transmitter design.  It's one FET, two DC power supplies, and passive components. Not much to go wrong. It's also heavy metal for one FET.

The front will only have an on/off switch.

Jon
KA1TDQ

I finished it this morning and tested over-the-air.  I listened to myself on the headphones and it sounds good to me.  I'm getting 25 watts carrier and with the audio on half-way I get 70 watts peak by going "Ahhhhh" into the microphone.  

The next step is to rig this up so that I can switch easily between RX and TX for an actual QSO to get some real audio reports.  If that all goes well, I'll disassemble it and reassemble it into a nice chassis that I picked up at a hamfest.  This will be my main rig and I'll put the single tube transmitter on a shelf.

Jon
KA1TDQ

Jon

Congratulations!

I look forward to hearing the results of your on-air testing.

Note:

Most power meters... even those that claim to measure "peak power"... do not actually capture and display peak power. They are unintentionally or intentionally* designed to produce a reading that does not capture the peak envelope power of a voice-modulated AM (or SSB) signal. Therefore, while your power meter may indicate nearly a tripling of the peak RF output power when you say "Ahhhhh" into the microphone... your actual peak envelope power (corresponding to an audio waveform peak) is probably significantly higher.

I once modified a popular digital RF power meter to produce a reading that accurately corresponded to the peak envelope power. Doing this was not a trivial task. The time constants of the peak detector in the stock meter were just too long to capture peak envelope power.  

The best way (in my opinion) to measure the actual modulation levels is with a two channel digital oscilloscope, and an off-air RF sniffer (like the RF sniffer that feeds the AM detector in your off air monitor). You display both the audio input to the modulator, and the sampled RF output signal... using the two scope input channels. You trigger the scope off of the audio input. You can superimpose the audio signal on the top of the RF signal's envelope using the vertical scaling and vertical offset controls of the scope. If all is well, the top of the displayed RF envelope will faithfully track the audio input signal. [Note: this works best with high-level modulation, or with a high-level-modulated transmitter followed by a linear RF amplifier.  It doesn't work well with something like an SDR transmitter... because of the large time delay between the audio input and the modulated RF output].  

To do this on 7MHz, you need a digital oscilloscope with at least 7MHz of vertical channel bandwidth.

You could do something similar with a 2-channel digital oscilloscope having only a 96 kHz sampling rate... employing your computer's stereo sound card, and with oscilloscope application software running on the computer... by using the output of your off air monitor (AM diode detector) instead of the sampled RF signal... to provide a decent (but less accurate) representation of the modulated envelope of your RF output signal. Computer sound cards don't pass DC, so, with this method, you can compare the modulated RF envelope's shape to the modulating audio waveform's shape... but you can't determine the modulation index. I.e. with a DC-coupled A/D converter, you can observe both the modulation of the RF envelope and the carrier level... but, not with a typical computer sound card serving as the A/D converter.

Or, you could do something similar using an REA modulation monitor in conjunction with a computer.

*I once spoke with a fellow who sells digital power meters that use modern sampling techniques to measure the "peak" RF power level of a modulated RF signal. I asked him why he set the time constants (i.e. the averaging time) in his design so that the peak power readings produced by his product would not capture the actual peaks that occur with voice modulation. His (very logical and business-like) reply was that "other brands of power meters (e.g. a Bird 43 meter with a peak power adapter installed, and competing digital display power meters) do not capture and display the actual peaks... and if he set the time constants in his product to capture and display the actual peaks... then no one would buy his product". I.e., most hams who purchase his product want the peak power reading to be significantly less than the true peak envelope power... so that they can operate at higher peak envelope power levels, while still alleging that they are not exceeding the FCC limit.

Stu

I called CQ and W7CK came back to me.  He lives in the same metro area as me, but my signal was a little PW since I was just running 25 watts carrier.  At first, he said my audio was a little down.  I had the audio drive at 50%, so I raised it to 90%.  The copy got better and my Daiwa PEP meter was reading over 100 watts peak.  He said that he didn't hear any distortion (my main concern), but I will try again after I get a decent carrier using my linear.

...on the linear...

I connected it to the new rig and I'm getting 700 watts carrier.  I don't see how a 25 watt rig could drive it that high, but that's was the wattmeter said on the AVG setting. 

There is no way for me to reduce power on the transmitter (It just has an on/off switch).  I'm going to do the brute force method and put in a 2db attenuator on the output going to the linear.  I'm just guessing at 2db, but I might have to try 3 or more to get the carrier around 250 watts. 

I'll build the attenuator using resistors.

Also, at the end of the conversation, W7ISJ chimed in near Tucson to say that he was copying us both at 100%.

Jon
KA1TDQ

Jon,

You should be able to reduce the drain voltage to run lower output.  Of course when you do that, you will need less audio.  That seems like a better and more efficient approach that a resistive attenuator network. 

Joe, GMS

Joe

Reducing the drain voltage significantly may result in the need to change the negative peak limiter's threshold voltage... and may also result in undesirable artifacts on negative modulation peaks (if you reduce the negative peak limiter's threshold voltage too much).

In my own experience with a 35W (at carrier) single FET transmitter... with 40V on the plate (at carrier)... and a 5V negative peak limiting threshold... using an attenuator was a better way to go than reducing the drain voltage.

If you have a spare ferrite core (of the type used by WA1QIX, and others, to make transformers for class E transmitters), and a 50 ohm dummy load, its easy to make a 4:1 (or other ratio) attenuator. See the information at the bottom of this web page (power divider):

http://mysite.verizon.net/sdp2/id11.html

I do agree, however, that cutting the the HV B+ supply voltage in half works great for reducing the output power (at carrier) of my Ranger from 40W to 10W. With my Ranger, I actually use an external, continuously variable, HV B+ supply for the 6146, when I want to adjust the output power

Stu

I built a 2.9 db attenuator (the values worked out pretty good for common resistors) and the carrier power out of the linear is now 300 watts.  My only problem now is that when I turn the Radio Shack microphone preamp Master Volume control up above half scale, I get RF hash in the audio.  To properly modulate the rig the volume control needs to go 80-90%-ish.  FYI, I'm using the Radio Shack 4-channel USB mixer (catalog #3200026).

Running the rig barefoot I don't have the problem, but with the linear amp I do.  Of course, the rig isn't in a metal enclosure and neither is the linear.  I'm using plexiglass on the linear for an art-deco look.  But, I didn't have this problem before using my tube rig with the linear.

I've done some troubleshooting and the RF hash goes away when I unplug the XLR balanced cable from the Radio Shack preamp (this is the cable from the microphone to the preamp).  I've ordered some snap-on RF ferrite chokes that I can install over the cable to see if that helps. 

Also, I don't have a ground in the shack that I can use.  I don't know if it would help anyway if I were to install one since I'd be driving the ground rod into Arizona sand. 

Jon
KA1TDQ

Jon,

    With your grounding setup, or lack of, you have a formidable task to stabilize the station. Perhaps a 33' wire will act as a 40m ground; realize that the end will be RF hot though.

   With my Flex 3000 station I am stable 80-6m with my 80m OCF dipole. if I run my loop fed with OWL, I can run 40-15m with the loop. The EFJ Matchbox is 15" from my Audio chain Mic input! Dumb!! When I turn on the Linear amplifier everything goes to hell with the loop, but not the inverted V. My computer USB keyboard and mouse like to lock up. A wireless mouse didn't help. Common mode chokes on the USB lines help on some bands but not others.

    If you have a high power dummy load, try that to see if the problem goes away. It might NOT with the amplifier wide open like that. If your good on the dummy load, then see about eliminating common mode feeder current, and moving the antenna further away (if possible). As I write this, I think you are running an end fed random wire out the window. True? Then were is your return current supposed to go?

  With the amplifier, tune the loading beyond maximum power (less capacitance) until the RF carrier power drops about 20%. Then modulate looking at the carrier versus PEP power. Tune for the highest ratio on your meter. I bet that takes your carrier down to 250 watts or even less.

Jim
Wd5JKO

Stu,
I do fully realize there are limitations on how much the drain voltage can be reduced, but with the 3 db or so required here, it should be doable.  Of coarse one needs to examine the drain switched waveform to assure it looks as it should.  As far as changes to the negative peak limiter, that should also be easy. 

One of the advantages of Class D or E is the efficiency and I realize we are dealing with very low power here, but it seems contradictory to do it with pads when other solutions are available.

Joe, GMS

Jon

To reduce the RF being coupled into the microphone via the balanced microphone output cable:

1. Slip a ferrite core (the same type and size as used to wind transformers) over the microphone cable. Run the cable around the ferrite core two or three times. This will require the removal of the cable/connector at one end; but it is more effective than a snap-on ferrite.

2. Connect the shield of the cable to ground at the end that plugs into the preamplifier, but leave the shield floating (not connected to the microphone shell) at the microphone end (unless the microphone requires phantom power). This will create a Faraday shield around the microphone cable, and will eliminate the common mode path (on the outside of the shield) between the microphone and the preamplifier.

Stu

Joe

I think you are correct. I did not fully understand the physics of a MOSFET.

Jon is using 28V on the drain of the single FQA 11N90. To reduce the power by 3dB, he would need to reduce the drain voltage to 28V x 0.707 = 20V.

For -90% negative modulation peaks, the drain voltage would be down to 2V, which seemed kind of "iffy", to me, considering the gate-to-source threshold voltage is specified as 3-5 volts.

However, upon review of the information on the web, and also reminding myself that the D-S voltage drops to zero on each RF cycle... there is no problem with a low D-S voltage, provided it is positive or zero. A negative D-S voltage would be a problem, because of the diode that exists between the drain and the source in a MOSFET. The specification sheet for the FQA11N90C indicates that the device will work okay, in the "on" state, with drain-to-source voltages below 0.1V.

So, it appears that a low D-S voltage is not a problem, provided the negative peak limiter's threshold is adjusted accordingly, to prevent negative D-S voltages on negative modulation peaks.


Stu

Yeah, my negative peak limiter has a 7805 regulating the voltage to 5 volts.  I cranked up the audio really high and the one QSO I had said that he didn't detect any distortion on my signal.

I spoke too soon on the RF hash source being the microphone cable.  I unplugged the XLR cable from the preamp and ran the test again.  The hash showed up again when the volume level went above 50%.  I did take a class E ferrite core (for the lack of a better name) and put it over the RCA cable which feeds my car audio amplifer from the preamp.  Bam!  Hash gone.

However, when I plug the XLR microphone cable in and put the volume above 50%, I again get hash.  So I'm back to putting cores on the XLR cable. 

For simplicity, I think I will just try the snap-on ferrite cores for now when they arrive.  My XLR cable has commercial, pre-molded ends and I don't want to cut them if I don't have to.  I'll try to get at least two wraps of cable through the strap-on.  If the hash is still there, I'll make up my own new cable and ends to your specifications.

Jon
KA1TDQ

Jon - It sounds like you're on your way.  I have found it interesting to follow this thread and your progress.  Keep us updated.

73
Brad

I just got my 3 snap-on chokes and added them one by one onto the microphone cable.  I added the first one next to the preamp and it didn't seem to help too much.  I put the next one next to the microphone and things quieted down alot!  I added the third in the middle of the cable and my carrier got totally silent.  I listened to myself in the headphones over the air and it sounds super clean.

I also changed my attenuator from a 2.9 db reduction to a 4.15 db (again, convenient values for resistors).  Out of the linear, my carrier is 210 watts.  This seems like a nice level with enough headroom for voice peaks.  

It feels good.  This transmitter is only class C, but I get warm fuzzies talking through a mosfet rather than a tube.  

I just got off the air with AC0RL in Kansas, but it was hard to get good audio reports due to band conditions being noisy.

Jon
KA1TDQ

Jon

Congratulations again!

What's your next project?

Stu

I'm going to rest for a bit after I get this rig permanently mounted in its chassis.  I still need to powder coat the outside and get the steel for the front and bottom.

But, I could do a medium power class E rig.  I know how to use analog drive for the gate and I can use this same modulator design.  With the components I used, they could easily do 100 watts carrier.  I'd just need to get a larger car audio amplifier to drive it.  The only thing new would be the output circuit for the drain. 

It'll have to wait a while though.  My wife is already asking me where I plan to put new things that I build.  I've run out of space in the shack, and I need to find a shelf for my tube transmitter. 

-or-

Build a solid state 40 meter linear.  This will totally get me off of tubes. 

I'm enclosing a last look at my final.  Look, no filament wires!

Jon
KA1TDQ

" My wife is already asking me where I plan to put new things that I build.  I've run out of space in the shack. "

Well, there is always the space inside partition walls. A good 3 inches....

And as Tuco might say, "I feel a man like you can manage it."


klc

I only get the warm fuzzies when my tube filaments consume at least 150W per tube.  ;D

Congrats on your project!

Phil