Tony,

I use 74AC244. ?8 buffer gates parallel in my PA.

73

Barb WB2CBA

Evan, Ryuji, JZ, et al,

The attached zip file includes two QDX PA sims that JZ and I experimented with a few months ago. Evan, you may have already seem some version of this. After a lot of experimentation, I found no value in retaining L14 so it has been removed for these sims resulting in a voltage-fed system. At this point, my sims diverged from John's as he retained L14 to experiment with suppressing the L di/dt spikes at key up. My goal was to experiment with different transistors for power output and logic drive capabilities.

The first of these sims uses the existing 74ACT08 drive scheme incorporated in the QDX, i,e., two AND gates paralleled to drive each pair of FETS? (one AND gate driver for each FET). The second sim "deadbugs" another 74ACT08 in parallel to allow four AND gates to drive each pair of FETS, so double the drive capability, If you recall, the suggestion to deadbug a second 74ACT08 came up in the forum posts months ago so I decided to try it after noticing that the existing driver can't provide enough current to quickly switch any of the FETs under consideration.

I don't currently have an equivalent sim for the BS170. I can reconstruct the original BS170 sim if necessary but I don't see any value in continuing with this device.

Takeaways:

1. The TN0110 device produces more power output than the BS170 for the same AND gate drive used in QDX. The drain voltage drops to around 2V when the FET is on showing that a single AND gate per FET is insufficient to fully turn on the FETs. This results in unwanted power loss in the FETs and limits power output to the LPF.

2, Doubling the 74ACT08 AND gate drive produces even more power output and the drain voltage drops to nearly 0V when the FETs are on.This confirms my suspicion that the FET gates need a driver capable of stronger current source/sinking and proves out the suggestion made months ago to give this a try.

3. Someone mentioned an alternate octal buffer that can source and sink more current than the 74ACT08. If someone can remind me of that device number, I'll create a third sim to test this theory.

Regards,

Tony - AC9QY

Tony,

Thank you!? With a little bit of editing, I was able to run the simulation.

One quick question: why 1uh for the transformer?? My measurements of the actual QDX transformer are closer to 8uh.? This also agrees with calculations based on the number of turns.

Again, thank you for sharing!

73

Evan

AC9TU

On Sun, Sep 24, 2023 at 8:55?PM Tony Scaminaci <tonyscam@...> wrote:

Evan, Ryuji, JZ, et al,

The attached zip file includes two QDX PA sims that JZ and I experimented with a few months ago. Evan, you may have already seem some version of this. After a lot of experimentation, I found no value in retaining L14 so it has been removed for these sims resulting in a voltage-fed system. At this point, my sims diverged from John's as he retained L14 to experiment with suppressing the L di/dt spikes at key up. My goal was to experiment with different transistors for power output and logic drive capabilities.

The first of these sims uses the existing 74ACT08 drive scheme incorporated in the QDX, i,e., two AND gates paralleled to drive each pair of FETS? (one AND gate driver for each FET). The second sim "deadbugs" another 74ACT08 in parallel to allow four AND gates to drive each pair of FETS, so double the drive capability, If you recall, the suggestion to deadbug a second 74ACT08 came up in the forum posts months ago so I decided to try it after noticing that the existing driver can't provide enough current to quickly switch any of the FETs under consideration.

I don't currently have an equivalent sim for the BS170. I can reconstruct the original BS170 sim if necessary but I don't see any value in continuing with this device.

Takeaways:

1. The TN0110 device produces more power output than the BS170 for the same AND gate drive used in QDX. The drain voltage drops to around 2V when the FET is on showing that a single AND gate per FET is insufficient to fully turn on the FETs. This results in unwanted power loss in the FETs and limits power output to the LPF.

2, Doubling the 74ACT08 AND gate drive produces even more power output and the drain voltage drops to nearly 0V when the FETs are on.This confirms my suspicion that the FET gates need a driver capable of stronger current source/sinking and proves out the suggestion made months ago to give this a try.

3. Someone mentioned an alternate octal buffer that can source and sink more current than the 74ACT08. If someone can remind me of that device number, I'll create a third sim to test this theory.

Regards,

Tony - AC9QY

I’m not sure Evan. JZ sent me the design so I’ll let him answer this. I would like to sync up our models on a regular basis, especially with the correct inductor values.

Tony

On Sun, Sep 24, 2023 at 11:11 PM Evan Hand <elhandjr@...> wrote:

Tony,

Thank you!? With a little bit of editing, I was able to run the simulation.

One quick question: why 1uh for the transformer?? My measurements of the actual QDX transformer are closer to 8uh.? This also agrees with calculations based on the number of turns.

Again, thank you for sharing!

73

Evan

AC9TU

On Sun, Sep 24, 2023 at 8:55?PM Tony Scaminaci <tonyscam@...> wrote:

Evan, Ryuji, JZ, et al,

The attached zip file includes two QDX PA sims that JZ and I experimented with a few months ago. Evan, you may have already seem some version of this. After a lot of experimentation, I found no value in retaining L14 so it has been removed for these sims resulting in a voltage-fed system. At this point, my sims diverged from John's as he retained L14 to experiment with suppressing the L di/dt spikes at key up. My goal was to experiment with different transistors for power output and logic drive capabilities.

The first of these sims uses the existing 74ACT08 drive scheme incorporated in the QDX, i,e., two AND gates paralleled to drive each pair of FETS? (one AND gate driver for each FET). The second sim "deadbugs" another 74ACT08 in parallel to allow four AND gates to drive each pair of FETS, so double the drive capability, If you recall, the suggestion to deadbug a second 74ACT08 came up in the forum posts months ago so I decided to try it after noticing that the existing driver can't provide enough current to quickly switch any of the FETs under consideration.

I don't currently have an equivalent sim for the BS170. I can reconstruct the original BS170 sim if necessary but I don't see any value in continuing with this device.

Takeaways:

1. The TN0110 device produces more power output than the BS170 for the same AND gate drive used in QDX. The drain voltage drops to around 2V when the FET is on showing that a single AND gate per FET is insufficient to fully turn on the FETs. This results in unwanted power loss in the FETs and limits power output to the LPF.

2, Doubling the 74ACT08 AND gate drive produces even more power output and the drain voltage drops to nearly 0V when the FETs are on.This confirms my suspicion that the FET gates need a driver capable of stronger current source/sinking and proves out the suggestion made months ago to give this a try.

3. Someone mentioned an alternate octal buffer that can source and sink more current than the 74ACT08. If someone can remind me of that device number, I'll create a third sim to test this theory.

Regards,

Tony - AC9QY

On Mon, Sep 25, 2023, 4:35 AM John Zbrozek <jdzbrozek@...> wrote:

Evan, Tony,

"Correct inductor value" is not a simple?question!

The inductance of a core based transformer will depend not just on the number of turns and type of core. Core materials tend to show declining permeability with increasing frequency, and so a given transformer winding will show much less inductance at 28 MHz than at 14 MHz for example.

To further complicate matters, transformers assembled with pairs of wires, parallel? or twisted, or with coaxial cable, exhibit transmission line properties. Inductances (including leakage inductance), and capacitances become absorbed into the properties of the transmission line and determine its characteristic Z.

So what to do in an LTSpice simulation?

That depends on what questions your simulation is to answer. For many purposes, especially where the coupling factor K is known to be near perfect unity, the precise inductor values used are not as important as are the ratios between them. Values that are within a range extending about a decade will produce similar results.

When K is lowered, leakage inductances develop within the model, proportionally to the assigned values. Precision becomes more important.

JZ

On Mon, Sep 25, 2023, 1:00 AM Tony Scaminaci <tonyscam@...> wrote:

I’m not sure Evan. JZ sent me the design so I’ll let him answer this. I would like to sync up our models on a regular basis, especially with the correct inductor values.

Tony

On Sun, Sep 24, 2023 at 11:11 PM Evan Hand <elhandjr@...> wrote:

Tony,

Thank you!? With a little bit of editing, I was able to run the simulation.

One quick question: why 1uh for the transformer?? My measurements of the actual QDX transformer are closer to 8uh.? This also agrees with calculations based on the number of turns.

Again, thank you for sharing!

73

Evan

AC9TU

On Sun, Sep 24, 2023 at 8:55?PM Tony Scaminaci <tonyscam@...> wrote:

Evan, Ryuji, JZ, et al,

The attached zip file includes two QDX PA sims that JZ and I experimented with a few months ago. Evan, you may have already seem some version of this. After a lot of experimentation, I found no value in retaining L14 so it has been removed for these sims resulting in a voltage-fed system. At this point, my sims diverged from John's as he retained L14 to experiment with suppressing the L di/dt spikes at key up. My goal was to experiment with different transistors for power output and logic drive capabilities.

The first of these sims uses the existing 74ACT08 drive scheme incorporated in the QDX, i,e., two AND gates paralleled to drive each pair of FETS? (one AND gate driver for each FET). The second sim "deadbugs" another 74ACT08 in parallel to allow four AND gates to drive each pair of FETS, so double the drive capability, If you recall, the suggestion to deadbug a second 74ACT08 came up in the forum posts months ago so I decided to try it after noticing that the existing driver can't provide enough current to quickly switch any of the FETs under consideration.

I don't currently have an equivalent sim for the BS170. I can reconstruct the original BS170 sim if necessary but I don't see any value in continuing with this device.

Takeaways:

1. The TN0110 device produces more power output than the BS170 for the same AND gate drive used in QDX. The drain voltage drops to around 2V when the FET is on showing that a single AND gate per FET is insufficient to fully turn on the FETs. This results in unwanted power loss in the FETs and limits power output to the LPF.

2, Doubling the 74ACT08 AND gate drive produces even more power output and the drain voltage drops to nearly 0V when the FETs are on.This confirms my suspicion that the FET gates need a driver capable of stronger current source/sinking and proves out the suggestion made months ago to give this a try.

3. Someone mentioned an alternate octal buffer that can source and sink more current than the 74ACT08. If someone can remind me of that device number, I'll create a third sim to test this theory.

Regards,

Tony - AC9QY

Hi Tony

Real world testing, it is easy to see the clear decrease in power output and efficiency if you omit L14 (the 10 turn FT37-43 choke). The difference in performance is notable and significant. The fact that your simulation produces a different result indicates to me that something isn't quite right there. I suspect accurately simulating the output transformer may be critical, a simple coupled coil model may not suffice.?

The point about driver level is also quite interesting. Back when developing the original QCX I experimented with this. In my first prototype I had an entire ACT00 logic chip driving the three parallel BS170s with all four gates in parallel. Later I needed to implement other features in the transceiver and for some of those, taking some of the NAND gates for other purposes would be useful or necessary. I wanted an inverter for the TX signal and also to be able to gate both the CLK2 signal to the PA and the signal generator. The end result was that I only had a single gate left, to drive all three BS170 in parallel. I found that a little alarming, because there are three BS170s in parallel in QCX and they are Class E, and fast clean switching is very important. So of course then I had to decide whether to add another ACT00 chip or perhaps add something some place else to free up gates again. To make that decision I made detailed power and efficiency measurements for 1, 2, 4 gate cases. I also tried one gate per BS170 gate (3 gates used). I found that there was an increase in power output and efficiency when going from 1 gate driver to 2. But it was quite a slight small improvement. There wasn't any significant improvement in going from 2 to 4. Driving the gates one by one (one gate per BS170 separately) actually made things slightly worse; for whatever reason, paralleled gates seem to work better. Accordingly with these observations in hand, I felt happy sticking with the single gate drive for QCX. Then when it came to QDX where:

1. There are 2 gates per driver not one;

2. There are 2 parallel transistors not 3;

3. It isn't Class E so it's a little less critical;

I felt entirely comfortable with 2 gates driving 2 parallel BS170s on each side.?

So I haven't tried doubling up a piggy back ACT08 but it would be an interesting experiment for someone; personally it would surprise me if doubling up made a significant or even noticeable improvement.?

73 Hans G0UPL

I have attached a PDF of some work I did over the past few days regarding core loss modeling for my QDX. The result data actually covers 80-10 meters but I was looking at my high band unit when doing the work.

Maybe the SPICE modelers among us will be able to make use of it.

Tony

AD0VC

---

From: [email protected] <[email protected]> on behalf of John Z <jdzbrozek@...>
Sent: Tuesday, September 26, 2023 12:27 PM
To: [email protected] <[email protected]>
Subject: Re: [QRPLabs] QDX Experimental PA Simulation

?

Hi Tony, Hans et al,

As I had posted earlier, modelling the output transformer for QDX/QMX
is tricky business!

Proper modeling of the output transformer is complicated at the
outset. There are parasitic elements, frequency dependent
permeabilities and couplings , losses, etc.

.Ross's move to a RWTST design for the 12 Volt radios further
complicates simulation. The 12V RWTST transformer is a hybrid between
a transmission line transformer and a conventional transformer, as it
has elements of both. It is neither fish nor fowl.

Then there is the issue of the? peculiar binocular transformer core.
Ross posted a model which treated the windings through each aperture
of the binocular as separate, non-coupling inductors. Yes, when wound
in a way which avoids linking the central post of the binocular, that
is what happens! It is key to the operation of the Stockton SWR bridge
incorporated in QMX.

More Controversy!

JZ












On Tue, Sep 26, 2023 at 1:52?PM Tony Scaminaci <tonyscam@...> wrote:
>
> Hi Hans,
>
> You make valid points, especially about modeling the output transformer. I’ll leave this question to JZ as he has more expertise in this particular area. I do think his model is fairly accurate though based on the spike and power output levels we measured early on. The spike level was close to what you and others measured on the bench.
>
> John initially suggested replacing L14 with a small value resistor on the order of 1-2 ohms as a way of eliminating the key-up spike. We also experimented with keeping L14 and using the bypass cap. Both of these alternatives resulted in eliminating the spike but I don’t recall what the effect was on power output. In this respect, you may be correct. These experiments were done with the BS170s and I took it one step further and tied the transformer directly to 12V which did result in a slight increase in power output. All of these experiments were run at ~28 MHz in an attempt to improve power output on 10 meters.
>
> At this point, I began searching for a better transistor with similar specs to the BS170 but with two notable improvements: lower Vgs threshold and higher Vds breakdown voltage. Microchip’s TN0110 is an excellent drop-in replacement for QDX radios. Simply flip them upside down and you’re in business. The 1V lower gate threshold is more suitable for a 5V drive level and the 100V minimum breakdown voltage is much more effective against the L14 spike voltage. As always, it’s best if the spike never exceeds 100V… this could be achieved by a slight reduction in L14’s inductance.
>
> The TN0110 is back in stock at Mouser. I encourage you to build a QDX with this device and compare it directly against a QDX using the BS170. You’ll be pleasantly surprised at the increased power output and your customers will experience a much lower failure rate of the PA. A win for everyone.?
>
> Regards,
>
> Tony
> AC9QY
>
>
> On Mon, Sep 25, 2023 at 12:25 AM Hans Summers <hans.summers@...> wrote:
>>
>> Hi Tony
>>
>> Real world testing, it is easy to see the clear decrease in power output and efficiency if you omit L14 (the 10 turn FT37-43 choke). The difference in performance is notable and significant. The fact that your simulation produces a different result indicates to me that something isn't quite right there. I suspect accurately simulating the output transformer may be critical, a simple coupled coil model may not suffice.
>>
>> The point about driver level is also quite interesting. Back when developing the original QCX I experimented with this. In my first prototype I had an entire ACT00 logic chip driving the three parallel BS170s with all four gates in parallel. Later I needed to implement other features in the transceiver and for some of those, taking some of the NAND gates for other purposes would be useful or necessary. I wanted an inverter for the TX signal and also to be able to gate both the CLK2 signal to the PA and the signal generator. The end result was that I only had a single gate left, to drive all three BS170 in parallel. I found that a little alarming, because there are three BS170s in parallel in QCX and they are Class E, and fast clean switching is very important. So of course then I had to decide whether to add another ACT00 chip or perhaps add something some place else to free up gates again. To make that decision I made detailed power and efficiency measurements for 1, 2, 4 gate cases. I also tried one gate per BS170 gate (3 gates used). I found that there was an increase in power output and efficiency when going from 1 gate driver to 2. But it was quite a slight small improvement. There wasn't any significant improvement in going from 2 to 4. Driving the gates one by one (one gate per BS170 separately) actually made things slightly worse; for whatever reason, paralleled gates seem to work better. Accordingly with these observations in hand, I felt happy sticking with the single gate drive for QCX. Then when it came to QDX where:
>>
>> 1. There are 2 gates per driver not one;
>> 2. There are 2 parallel transistors not 3;
>> 3. It isn't Class E so it's a little less critical;
>> I felt entirely comfortable with 2 gates driving 2 parallel BS170s on each side.
>>
>> So I haven't tried doubling up a piggy back ACT08 but it would be an interesting experiment for someone; personally it would surprise me if doubling up made a significant or even noticeable improvement.
>>
>> 73 Hans G0UPL
>>
>>
>
>