The Inexpensive, simple, MOSFET A.M. linear project

[ka1tdq]

What's the procedure to tune this after building it? If someone is going to use this for AM, I guess you'd slowly increase the voltage on the gates until there was some resting current, and then back it back down to zero again?

And for getting the proper class E waveforms on the output, do the same rules apply as to a regular class E transmitter? For the first tuning, do you keep the voltage at normal operating level and just reduce the drive to say, 1 watt or do you reduce the drain voltage in half and put 5 watts carrier into the input? Or reduce both?

I use an Acom amplifier for my PW plate modulated transmitter. I could build one of these to free that up for my ricebox.

Jon

[M0VRF]

It's class B and broadbandish

Class E is switching and not linear!

I don't why anyone would build such a thing when class E is the way to go to generate lots of power efficiently.

I'm sure someone will let me know?

JB

[KD1SH]

Some of us neanderthals even build with toobs.

It's class B and broadbandish

Class E is switching and not linear!

I don't why anyone would build such a thing when class E is the way to go to generate lots of power efficiently.

I'm sure someone will let me know?

JB

[WBear2GCR]

It's class B and broadbandish

Class E is switching and not linear!

I don't why anyone would build such a thing when class E is the way to go to generate lots of power efficiently.

I'm sure someone will let me know?

JB

Easy.

People have QRP SDR tranceivers and QRP standard topology transmitters. Some want more power.

A present example is the Hermes Lite.

                              _-_-bear

[WBear2GCR]

What's the procedure to tune this after building it? If someone is going to use this for AM, I guess you'd slowly increase the voltage on the gates until there was some resting current, and then back it back down to zero again?

And for getting the proper class E waveforms on the output, do the same rules apply as to a regular class E transmitter? For the first tuning, do you keep the voltage at normal operating level and just reduce the drive to say, 1 watt or do you reduce the drain voltage in half and put 5 watts carrier into the input? Or reduce both?

I use an Acom amplifier for my PW plate modulated transmitter. I could build one of these to free that up for my ricebox.

Jon

Not being the designer, but we're fairly early in the design cycle.
I think the control and protection circuitry, if any, will come later...

[ka1tdq]

Funny.

All you need is a negative peak limit buzzer. Done.

Jon

[steve_qix]

I updated the initial post.

1) New schematic - I added 2 more MOSFETs, so now there are 12 MOSFETs.  With this, I can easily get a 300W carrier, and modulation around 125-130% positive.

2) New picture showing the top view of the amplifier

I do have a T/R system, which is very standard.  The use with any Flex, Anan or other transceiver will be very similar.

I am currently using this with a Flex 5000.  The Flex has several antenna connections, so I dedicated one of the antenna connections on the flex to the linear.  The Flex is then configured, when in transmit mode, to send the output to the antenna connection hooked to the linear.   There is a standard transmit/receive system, which will disconnect the antenna from the receiver input, and connect it to the amplifier output.  The T/R system also keys the Flex, and enables the power supply for the linear.

You can design your T/R system anyway you want.  DO use a sequencer.  There is a lot of information about sequencing, and a schematic for a simple but effective sequencer on the class E web site.

[KD6VXI]

Steve,

Is a lownpass filter required or does the tuned output give enough harmonic suppression?

Secondly, is the formula for these fets the same (or close enough) for the bipolar eq (Vcc x Vcc) / 2 x Pout to work on a broadband Ed output network?

I understand it won't give the best power transfer but I was looking at doing a set of 3 L networks on the output post broadband xformer. Use a Russian vac cap to switch them in band for band to get the tuning.  With the series C, it's not as easy to make a tribander.

--Shane
KD6VXI

[steve_qix]

Very interesting..

When one runs an AM linear, is it best to accentuate the positive peaks or raise the carrier level to achieve just 100% positive modulation?

My opinion is that anything more than 100% positive peaks is useless when it comes to using a linear amplifier on AM.

All linear amplifiers produce a certain PEP level.  Why wouldn't one opt for the biggest carrier (quieting) and just modulate up to 100% perhaps using an all pass filter to limit those useless positive peaks?

A good 6 to 8 pole all pass filter will eliminate that nasty assymetry in the human voice in your audio chain.  Plus that assymetry does NOTHING for intelligibility.


YUP....A bit of heresy but fact.

Chuck K1KW

Given that, the current amplifier design and implementation meets the goals of 300 watts carrier output.  However, at 300 watts carrier out, the positive modulation is limited to about 120% positive.

I suppose to have asymmetry (or not) is a personal choice, and there are always compromises.  Personally, I prefer the sound of an unmodified waveform.  For whatever reason, it sounds better (to me).  Some people may not notice it.

So, if you don't have a lot of asymmetry (or reduce the asymmetry by artificial means such as with a phase rotator), you can run more carrier output given the same amplifier because you don't need the positive peak headroom.  And, lots of people do this to get the most carrier.

I prefer the asymmetrical waveform but of course that comes with the compromise of lower carrier output given the same linear amplifier.

The final goal of this project is an amplifier that will deliver 300 watts carrier output and at least 160% positive modulation, and then the user can actually run it any way he (or she) wants !!

Give 'em choices 

Regards,  Steve

This is an interesting thread.
Steve, I just have a couple of questions:
1. You spec 80-85 vdc as the voltage used. Is there any reason why one could not go to 100vdc in this circuit and reduce current to keep dissipation down?
I have some nice 70vac @14amp transformers. Could put a variac in, if needed, but they will give about 100vdc output with full input AC voltage.

2. The 1.5 ohm resistors on the source leads- are they a special low inductance thick film or just 5 watt thick film resistors with spiral windings?
The implied question is whether inductance matters at this low resistance and if so- would 3-4.7 ohm 2 watt resistors be better?
Same question on the 4.7 ohm input resistors.
3. If one went to a 4x4 setup (16 FETs) how will that affect the shunt caps?
4. Can this be bandswitched by tapping a larger output inductor and using-say- a 1000pF at 10KV vacuum variable and using vacuum relays?

Answer 1:  You probably could increase the voltage, but I have not tested at anything higher than 85VDC.  I'm currently using a 70V transformer secondary, and I get about 82V.  The filter capacitor is probably too small for the current.

Answer 2:  The 1.5 ohm resistors are 3  4.7 ohm resistors in parallel.  The resistors are metal film resistors (low inductance).  You could probably get away with a 2 watt resistor, but a 3 watt resistor would be better. 

Answer 3:  I haven't tested.  But, the shunt cap is a non-critical value.  I had 1500pF around, so I used them :-)  I have not experimented with other values.

Answer 4:  You could probably tap down on the inductor, but I would not short it out.  This will create a LOT of heat ! 

Hope this helps !!

[steve_qix]

Steve,

Is a lownpass filter required or does the tuned output give enough harmonic suppression?

Secondly, is the formula for these fets the same (or close enough) for the bipolar eq (Vcc x Vcc) / 2 x Pout to work on a broadband Ed output network?

I understand it won't give the best power transfer but I was looking at doing a set of 3 L networks on the output post broadband xformer. Use a Russian vac cap to switch them in band for band to get the tuning.  With the series C, it's not as easy to make a tribander.

--Shane
KD6VXI

The tuned circuit gives a LOT of harmonic suppression.  I added an FFT of the output to the main post.  I will also put it here.  The 2nd and 3rd harmonics are WAY down !!!  In the picture, the yellow trace is the output waveform, looking into a broadband dummy load, and the red trace is the FFT of that waveform.  The vertical scale is 10db per division.  The 2nd harmonic is down around 55db (maybe a bit more, actually) from the fundamental. 

I can't comment about the 3 L networks.  You'd have to experiment, but let us know !! I haven't tried that.

[w9jsw]

Steve,

Missing a ground on the neg side of the PS bridge.

What is the thought behind the dual relays controlling both input and output of the PS? I would think that for T/R you would leave the PS on but cycle the output relay.

Does the input side of the PS need a step start?

Also, will need a bleeder to drain those big caps on power down.

Is 95V too much or is that ok? I have a 46-48V supply that I am considering series/parallel on the secondary to give me both voltages. The Trans is a 1000VA Antek that can do 15A on each secondary.  

John

[steve_qix]

Steve,

Missing a ground on the neg side of the PS bridge.

What is the thought behind the dual relays controlling both input and output of the PS? I would think that for T/R you would leave the PS on but cycle the output relay.

Does the input side of the PS need a step start?

Also, will need a bleeder to drain those big caps on power down.

Is 95V too much or is that ok? I have a 46-48V supply that I am considering series/parallel on the secondary to give me both voltages. The Trans is a 1000VA Antek that can do 15A on each secondary.  

John

Thanks for the schematic check!  Much appreciated.   I fixed the missing ground 

Double relays (AC input side and DC output side of the power supply):  I am a real big fan of removing HV from things when they are not transmitting.  Prevents all sorts of problems !!    I agree that a step-start would be in order for the power supply.  I don't have one, but I use the Variac to bring up the voltage.  A step start is necessary for non-variac use for sure.  I will update the schematic once I add a step start to my setup.

A high value bleeder would be a good idea once the step start is included.  I have this on all of the class E rigs.  Never good to leave capacitors charged up when the equipment is not in use.

I haven't done any experiments at voltages exceeding 85 VDC.  You could put a "bucking" transformer in the primary to buck the line a bit and bring the voltage down, if necessary.

[steve_qix]

My interest in this amp would be using it with my Flex 6500. I have two related questions. Has anyone though of some sort of TR switching? More important, how do you safely mute the 6500 during transmit?

Yes, a proper T/R system is VERY important, and there are tried and true methods.   You have several possibilities with the Flex.

I believe the 6500 has more than one RF output port.  You COULD dedicate one output port to the amplifier, and use the other for receive.  I do this in my own setup.  Or you could configure a relay in the Flex antenna terminal, and switch between the antenna and the amplifier input.  This relay would work in concert with the main station T/R relay.  There are several people using this type of arrangement with SDRs and class E rigs.  The concept is identical.

A sequencer is ABSOLUTELY necessary !   This will facilitate an orderly transition between receive and transmit, and visa-versa.

This is the sequencer I use in my own station:

[w9jsw]

On the tops down pic, I see you use the usual class E manner of attaching all of the drains together on a large piece of copper. When we go to a PCB do you still think it makes sense to preserve that feature or can we rather just use an expansive fill area on the board to combine the drains?

[steve_qix]

On the tops down pic, I see you use the usual class E manner of attaching all of the drains together on a large piece of copper. When we go to a PCB do you still think it makes sense to preserve that feature or can we rather just use an expansive fill area on the board to combine the drains?

A fill area of sufficient width should be fine.  There are some boards created by N9SHU that implement some of a class E transmitter, and the drains are connected with a fill area that's about 1/2 inch wide.

[K9MB]

Are silicone pad insulaors sufficient for voltages present?
Do shunt caps need to be porcelain or is a mica cap ok?

[K1JJ]

Steve,

Free online tone gen for two-tone test:

https://onlinetonegenerator.com/binauralbeats.html


I'm excited to build one. But we need to test a little further, past harmonic and voice yallo tests. Let's verify that these $3 FQA 11N90 devices are suitable for clean linear service using an  IMD test. Most common linear devices cost $100+ each.  The MRF150 is now over $100, so I am cautious.    A 700/1900Hz two-tone test at 1500 watts pep will tell the story. An AM and/or LSB test would suffice.   This will be the first thing the ARRL wants to know when they consider this concept for an article.

A closed loop SDR driver with Pure Signal (pre-distortion) will make it work perfectly, but most guys will probably want to use an open loop ricebox or equivalent as a driver. This KW linear will need to be at least -28 dB 3rd IMD or better to be "reasonably" clean when operating wide band AM on the bands. (+- 5-6 KHz)     (Most 100W riceboxes are about -30 dB 3rd... some inexpensive SS linear amps can be as marginal as  -25 dB 3rd)    As you know, signal purity is always dictated by the weakest link in the chain.  For example, if the driver is -40dB 3rd (clean) and the amplifier is -20 dB  (poor) then the resultant output is less than -20dB, like ~ -19dB 3rd, etc..

T

[w9jsw]

For this to be driven from some SDRs we will need another amp stage. My Hermes-Lite 2, for instance, can put out 5W CW. Going to need 6db of gain, I believe, to get to 5W carrier. Anyone have an idea of components? I was thinking a class A single MOSFET solution, perhaps the RD16HHF1?

That may in turn require a tuned input to the main amp to avoid lots of harmonics being passed. Thinking of using 24V for the control voltage for the relays/bias and also use that for Vcc for this amp.

Or is there another way?

John

[steve_qix]

Steve,

Free online tone gen for two-tone test:

https://onlinetonegenerator.com/binauralbeats.html


I'm excited to build one. But we need to test a little further, past harmonic and voice yallo tests. Let's verify that these $3 FQA 11N90 devices are suitable for clean linear service using an  IMD test. Most common linear devices cost $100+ each.  The MRF150 is now over $100, so I am cautious.    A 700/1900Hz two-tone test at 1500 watts pep will tell the story. An AM and/or LSB test would suffice.   This will be the first thing the ARRL wants to know when they consider this concept for an article.

A closed loop SDR driver with Pure Signal (pre-distortion) will make it work perfectly, but most guys will probably want to use an open loop ricebox or equivalent as a driver. This KW linear will need to be at least -28 dB 3rd IMD or better to be "reasonably" clean when operating wide band AM on the bands. (+- 5-6 KHz)     (Most 100W riceboxes are about -30 dB 3rd... some inexpensive SS linear amps can be as low as  -25 dB 3rd)    As you know, signal purity is always dictated by the weakest link in the chain.  For example, if the driver is -40dB 3rd (clean) and the amplifier is -20 dB  (poor) then the resultant output is less than -20dB, like ~ -19dB 3rd, etc..

T

I have not as of yet measured the IMD.  I figure if it's at least as good as a the typical, good sounding AM transmitter that is operating today, it's good enough for the purpose.  The linear should not generate any more products than the standard AM transmitter, and it is likely to generate fewer unwanted products.  I do get reasonable triangle wave modulation performance.

The goal of the project is NOT to be able to replace a linear amplifier costing many thousands of dollars, but to have performance as good as, or better than, a reasonable sounding AM transmitter running similar power, and to do it at relatively low cost and relatively simple construction.   I believe that currently, the project meets that goal. 

With respect to bandwidth containment, the amplifier/Flex combo with the Flex set to a 6kHz high end compares favorably with my class E rigs running a 6kHz DSP audio filter, when I observe the modulation on a remote spectrum display.  With both transmitters, there is some stuff observable outside of the 12kHz total passband, but those products are way down in level.

[K9MB]

1:What is the calculated Drain impedance for the amplifier?
Can you use the old rule of thumb V^2/2*Pout that we used for BPT amps?
As a single ended amp that would make it about 10 ohms, but this is a form of push-pull -right?  So the impedance has to be higher, but not sure about how to figure it...
I was trying to run the network on Jim Tonne’s ELSIE program that allows one to tune the components to see where the matches are optimum, but I need to know the input impedance to the RF output network you are using.

Edit:
I ran some figures on ELSIE and it indicates that given the values on the schematic, the output impedance into the network is about 12 ohms. Does that seem right?

[K9MB]

2. I want to make it multi-band and I am womdering why one could not use a 15UH roller inductor?

[K9MB]

 Steve began with 5 devices per leg and now has 6 per leg.
What is the practical limit of expanding further to say- 8 or 10 per leg?
Would the output transformers begin to heat up due to losses with this? Please advise.

[w1jto]

A sequencer is ABSOLUTELY necessary !   This will facilitate an orderly transition between receive and transmit, and visa-versa.

Steve

Very interested in sequencer. Does a commercial equivalent exist ?.

John

[K9MB]

[
This is an interesting thread.
Steve, I just have a couple of questions:
1. You spec 80-85 vdc as the voltage used. Is there any reason why one could not go to 100vdc in this circuit and reduce current to keep dissipation down?
I have some nice 70vac @14amp transformers. Could put a variac in, if needed, but they will give about 100vdc output with full input AC voltage.

2. The 1.5 ohm resistors on the source leads- are they a special low inductance thick film or just 5 watt thick film resistors with spiral windings?
The implied question is whether inductance matters at this low resistance and if so- would 3-4.7 ohm 2 watt resistors be better?
Same question on the 4.7 ohm input resistors.
3. If one went to a 4x4 setup (16 FETs) how will that affect the shunt caps?
4. Can this be bandswitched by tapping a larger output inductor and using-say- a 1000pF at 10KV vacuum variable and using vacuum relays?

[/quote]
Answer 1:  You probably could increase the voltage, but I have not tested at anything higher than 85VDC.  I'm currently using a 70V transformer secondary, and I get about 82V.  The filter capacitor is probably too small for the current.

Answer 2:  The 1.5 ohm resistors are 3  4.7 ohm resistors in parallel.  The resistors are metal film resistors (low inductance).  You could probably get away with a 2 watt resistor, but a 3 watt resistor would be better.  

Answer 3:  I haven't tested.  But, the shunt cap is a non-critical value.  I had 1500pF around, so I used them :-)  I have not experimented with other values.

Answer 4:  You could probably tap down on the inductor, but I would not short it out.  This will create a LOT of heat !  

Hope this helps !!
[/quote]

It does-thank you very much. 73, Mike

[K9MB]

I have been doing a little research on literature relating to the use of Power MOSFETs in linear applications.
I have only read articles quickly, but discovered that the important thing about linear applications is to keep the devices in the SOA (safe operation area). This is important because while generally- FETs are seen to have a positive temperature characteristic, that only holds up to certain temperature and then it inflects past zero and become more and more negative. This is similar to Bipolar amplifiers run in linear mode. In BPT amps, the emitters are ballated to compensate for increased bias current by lowering base bias. As current devices, this works well within limits.
In FETs, which are designed mostly for digital operation, the SOA area is fairly narrow, though I have links to extended SOA devices in this post at the bottom.

Two things can be done to mitigate the chances for getting devices into the deadly area above the SOA graph, according to articles listed below.

1. Keep Drain to source voltages as low as possible and then extra current can be drawn by the devices and still stay inside the SOA zone

2. Use as many devices in parallel as is practical to spread the current among a lot of devices and stay well within the SOA zome for each.

Steve has done the second thing by putting 6 devices in parallel in each leg. I wonder if that can be expanded to 8 or 10 in a leg to share it more widely?

Steve has specified that he uses 80-85 volts on the Drains and has stated that he has not looked at higher voltages.  My initial instinct was to push that voltage higher, but knowing that increased Drain voltage does far more than increased current in a device to push a device above the SOA zone, I now wonder if dropping the Drain voltage down to 50 volts might make more sense, even if it means increasing the current to still get the power output desired?

Any comments or discussion in this matter may improve the ling term success of efforts to build a stable and successful Linear amp that may be counted on to perform well with few failures due to thermal runaway.

The Extended SOA devices are available at Mouser and Digikey. They are more expensive than the FQA11N90s, but may be more reliable long term.

Ideas?

https://linearaudio.nl/sites/linearaudio.net/files/Fairchild_Cabiluna_AN-4161-TrenchMOS-SOA-Spirito-linear-mode_09-Dec-2013_PDF.pdf

http://www.kerrywong.com/2016/10/08/linear-mosfet-and-its-use-in-electronic-load/

https://www.infineon.com/dgdl/Infineon-ApplicationNote_Linear_Mode_Operation_Safe_Operation_Diagram_MOSFETs-AN-v01_00-EN.pdf?fileId=db3a30433e30e4bf013e3646e9381200

https://www.richardsonrfpd.com/docs/rfpd/Microsemi%20How%20to%20Make%20Linear%20Mode%20Work.pdf

https://www.microsemi.com/product-directory/mosfet/732-linear


https://www.power-mag.com/pdf/feature_pdf/1330614332_Renesas_Feature_Layout_1.pdf


Extended SOA MOSFETs

https://www.digikey.com/en/product-highlight/i/ixys/linear-l2-mosfets

https://www.mouser.com/new/ixys/ixys-mosfet-with-fbsoa/?gclid=CjwKCAiA4rGCBhAQEiwAelVtiyT7KvrhEjLXpdXkGL0M349GYKFuKcdLlwbUmQzu-AxqHT8iR65IeRoCzLsQAvD_BwE


Here is the SOA graph for the FQA11N90C
The DC Zone drops off sharply at all currents above about 30 volts, but if you keep well under the maximum current, it can be within the parameters.
Writers say that the zone may be much smaller than shown in graphs, so keeping voltages down and spreading current can keep ome well away from the dangerous edges.
Steve’s design seems to fit within the graph at 80-85 volts, but is not far from the edge, so more devices and not raising voltages will move one leftward and down in the region of safety and lessen chances of meltdown...