Amplifier inductor question

[Analyser]

So, on to his RD70 attempt. He's done his best with the matching on this amp to make it broadband however due to the low impedances it's still quite critical and therefore with this design I'd be impressed if he manages 70W across the whole band, if at all. Most other designs I've seen using these 12V Mitsubishi FETs drive the RD70 with an RD15 using the extra power to compensate, however that's swings and roundabouts with some of the designs and you've got to watch that bias current. In short, with this design, your mileage may vary - expect to do a bit of "tweaking" to get the power you want.

Totally agree. The reason you don't see more 12v high power transistors is the low impedance needed at the output for matching. A two stage output match is unable to provide low enough Q for a broadband match at this voltage and power level.
Realistically the best way to get more power output that people have been asking for is to put a couple of these devices in push-pull.

[thewisepranker]

Regarding single sided pcb, it is not usable at RF, and for sure not to be used on higher power

I don't agree. Most of the stuff inside the kit that the cable TV companies fit in your house is single-sided, uses leaded components and works up to 1.2 GHz. Some cheap splitters don't even have a PCB inside.

Band II isn't particularly high frequency and single-sided is adequate for most things.

[Maximus]

My basic understanding is that the underside acts as a capacitive ground plane and simply alters interaction with the components and stabilises things.

A lot of the old pro gear never had double sided PCBs or if they did was limited to a certain amount/effectiveness.


Sent from Tapatalk

[OgreVorbis]

It seems that getting high power with 12v is not easy. Maybe I should go with a tuned narrowband design. I don't have enough knowledge to design a whole new push-pull type amp myself.

Here is a translation from a German who built this:

The Notune concept of Radio Free Berkeley I just do not attempt it. With the components values ​​of FRB no more than 30 watts were rauszuholen, although he moved to 8A.
Either I've got something wrong made the replica, or what I believe rather that specified
Capacitor values ​​in the output matching are partially incorrect.
Comparing the assembly instructions of FRB BOM with the values ​​in the platinum print, then contradict several positions. Also in the output circuit. And just as accurate capacitor values ​​are entscheident for a functioning Notune Matching. Try After 2 days, the values ​​for now are correct that from the bill, or the layout of pressure, I gave up and a conventional output network drangesetzt.
By many Soldery the board therefore sees already pretty messed up from.
So as I have now lying there 60Watt output at 14V / 7A and 5W input measured.
However, the thing is so damn narrowband. Only 1 Mhz frequency change and he bagged on 30-40Watt and has to be retuned.
So I'll have a bit of time to use for the output network. In addition, also lack the harmonic filter, so should my power measurement of 60Watt even slightly too large.
But I was not yet so reticent because the current heat sink is too small and it's glow within minutes on. So he still gets a fat heatsink with fan, be OWF, more input, more voltage (15-16V think he should tolerate). And then, of course, fine tuning on Spekki.
Oh nochwas. 70Watt sprint no gimmick. I got me a few times the fingertips singed when I so a few Pikos in output only as hanging over it once just wanted without soldering.
Never again. 70Watt feel not good.
The current 150pF directly on the drain are also damn hot, then drift enormously in the capacity and power sags. Similarly, the yellow trimmer in the main are good warm. In the long run will not survive. So here should also Micas purely because of the temperature stability.
Perhaps one or the other is surprised that these experiences are new for me, but this is my first self made about 50W. I Confess openly here. Endstufenbau was previously not really my thing. Had been bought just finished Pallets.
But now I have power licked and not give up until the thing runs perfectly.
Harmonics u. Spurious free and maybe even notune.
Some may wonder why such a rebellion, there are enough proven concepts with more power and also to a more favorable price / performance ratio.
The answer is simple: My premise was: It has to run to the car battery at 12V.
Since then there is no better alternative.

Have a nice Sunday
Broadcaster

[Analyser]

Poor fella, sounds like he had a nightmare. Soon those yellow trimmers wil melt and short circuit ending in a nastier mess. And putting 5w in to get 60w out isn't right either, that's the gain of a bipolar!

I agree you should try a narrowband design, this will give the best gain and efficiency.

[Shedbuilt]

Poor fella, sounds like he had a nightmare. Soon those yellow trimmers wil melt and short circuit ending in a nastier mess.

+1

And putting 5w in to get 60w out isn't right either, that's the gain of a bipolar

These seem to have fairly low gain for FETs. Gp is specified at 10.6dB @ 175MHz/70W, and about 3dB less at 520MHz/50W.

[OgreVorbis]

I agree you should try a narrowband design, this will give the best gain and efficiency.

Does anyone know how to do that? If I want to make a narrowband design for 88MHz, how do I take the s-parameters in the datasheet and turn that into a matching network? I never actually designed an amp before. Is it just a matter of doing a calculation to get the component values or what?

Thanks for the help.

[Analyser]

I agree you should try a narrowband design, this will give the best gain and efficiency.

Does anyone know how to do that? If I want to make a narrowband design for 88MHz, how do I take the s-parameters in the datasheet and turn that into a matching network? I never actually designed an amp before. Is it just a matter of doing a calculation to get the component values or what?

Thanks for the help.

I'll try and do something for you over the next few days.

[Analyser]

Here you go.



Please bear in mind this circuit is purely theoretical and as with all things RF it may probably need tweaking and playing with. I have tuned it for 98MHz so you will need to adjust input and output inductors for different frequencies.
For the input, if you're finding the trimmer is fully meshed then make the hairpin longer and/ or add some capacitance to ground. A similar process is followed for the output but the output coil is easier to tune by compressing or opening the spacing between turns.

Let us know how you get on!

[OgreVorbis]

Thanks for the help. I'll give it a try after I order a few parts.

Also, I think I'll use the bias circuit design from the original because it uses a 9V regulator. My plan is to run this from a car battery and as the voltage slowly lowers, I want the bias to stay the same so it doesn't drop as much power. What do you think?

[Analyser]

Thanks for the help. I'll give it a try after I order a few parts.

Also, I think I'll use the bias circuit design from the original because it uses a 9V regulator. My plan is to run this from a car battery and as the voltage slowly lowers, I want the bias to stay the same so it doesn't drop as much power. What do you think?

Use a 5V regulator instead, this will give you much more room to play with as the battery voltage gets lower.

[Shedbuilt]

Analyser Sir, Please don't take this the wrong way. It's probably me - probably missing something, but the value of L3 looks rather high ?
I get much the same theoretical values as you, for everything else in the input and output networks......

[Analyser]

Analyser Sir, Please don't take this the wrong way. It's probably me - probably missing something, but the value of L3 looks rather high ?
I get much the same theoretical values as you, for everything else in the input and output networks......

Could be, it's just one of those situations where you'd have to try it and see, if I were building the amp I would make L3 variable, with a 2t S18 or something.

I've just had a quick look at my calculations and you're correct, it's a bit high, it should probably be in the region of 25.8nH, which is based on a Coss of 162pF, rather than the 120pF I had before.
You're probably not talking about that amount of error though, I suspect you've come up with something like 9.5nH which is based on matching to 0.7-j0 at 98Mhz?

[Shedbuilt]

Analyser Sir, Please don't take this the wrong way. It's probably me - probably missing something, but the value of L3 looks rather high ?
I get much the same theoretical values as you, for everything else in the input and output networks......

Could be, it's just one of those situations where you'd have to try it and see, if I were building the amp I would make L3 variable, with a 2t S18 or something.

I've just had a quick look at my calculations and you're correct, it's a bit high, it should probably be in the region of 25.8nH, which is based on a Coss of 162pF, rather than the 120pF I had before.
You're probably not talking about that amount of error though, I suspect you've come up with something like 9.5nH which is based on matching to 0.7-j0 at 98Mhz?

Thanks. Yes, you're right. It was COSS I missed. Based on the Zo data for 135MHz and 175MHz, I estimated something like 0.7 +j0.4 I think. So I guess you resonated the COSS, and added the resultant inductance to L3 ?

[Analyser]

Analyser Sir, Please don't take this the wrong way. It's probably me - probably missing something, but the value of L3 looks rather high ?
I get much the same theoretical values as you, for everything else in the input and output networks......

Could be, it's just one of those situations where you'd have to try it and see, if I were building the amp I would make L3 variable, with a 2t S18 or something.

I've just had a quick look at my calculations and you're correct, it's a bit high, it should probably be in the region of 25.8nH, which is based on a Coss of 162pF, rather than the 120pF I had before.
You're probably not talking about that amount of error though, I suspect you've come up with something like 9.5nH which is based on matching to 0.7-j0 at 98Mhz?

Thanks. Yes, you're right. It was COSS I missed. Based on the Zo data for 135MHz and 175MHz, I estimated something like 0.7 +j0.4 I think. So I guess you resonated the COSS, and added the resultant inductance to L3 ?

Yes, exactly.

[Shedbuilt]

Yes, exactly.

Thanks Analyser

[Shedbuilt]

Analyser, I've been wondering whether I should ask this. I fully agree that any theoretical RF design is likely to need optimisation, and empirical tweaking, but since this part of the thread arose from a question about theoretical design, I hope it's not inappropriate. Here goes. By adding inductance to L3, aren't you treating COSS as a series capacitance ?

[Analyser]

Analyser, I've been wondering whether I should ask this. I fully agree that any theoretical RF design is likely to need optimisation, and empirical tweaking, but since this part of the thread arose from a question about theoretical design, I hope it's not inappropriate. Here goes. By adding inductance to L3, aren't you treating COSS as a series capacitance ?

That's correct, Coss is a parallel impedance, therefore you need to convert it to series before it can be tuned out via the series inductor.

[Shedbuilt]

Analyser, I've been wondering whether I should ask this. I fully agree that any theoretical RF design is likely to need optimisation, and empirical tweaking, but since this part of the thread arose from a question about theoretical design, I hope it's not inappropriate. Here goes. By adding inductance to L3, aren't you treating COSS as a series capacitance ?

That's correct, Coss is a parallel impedance, therefore you need to convert it to series before it can be tuned out via the series inductor.

Apologies for more questions, and please feel free to correct me if I'm wrong, or if I'm looking at it completely wrong (if you have the time and inclination), but if we look at the parallel impedance (using Zo as 0.7 +/- j0 for convenience), it looks like 0.7R in parallel with something like 130pF. If we perform a parallel to series conversion, at 98MHz, the series equivalent gives a resistance just under 0.7R, in series with almost 42nF. If we want to absorb 42nF by series resonance (again at 98MHz), we get something like 63pH - rather than 20ish nH. This would mean we could use the figures calculated from Zo alone - pretty much unchanged (rather than adding 20ish nH to L3) ?
I guess - if we wanted to, there are other ways to deal with COSS. One might be to make the output match a pi-match, and absorb COSS into the input capacitance (ie subtract it from the capacitance at the input of the pi-match) ?
Another option could be to absorb COSS by parallel resonance (with a parallel inductor; possibly L2) ?

[Analyser]

Apologies for more questions, and please feel free to correct me if I'm wrong, or if I'm looking at it completely wrong (if you have the time and inclination), but if we look at the parallel impedance (using Zo as 0.7 +/- j0 for convenience), it looks like 0.7R in parallel with something like 130pF. If we perform a parallel to series conversion, at 98MHz, the series equivalent gives a resistance just under 0.7R, in series with almost 42nF. If we want to absorb 42nF by series resonance (again at 98MHz), we get something like 63pH - rather than 20ish nH. This would mean we could use the figures calculated from Zo alone - pretty much unchanged (rather than adding 20ish nH to L3) ?
I guess - if we wanted to, there are other ways to deal with COSS. One might be to make the output match a pi-match, and absorb COSS into the input capacitance (ie subtract it from the capacitance at the input of the pi-match) ?
Another option could be to absorb COSS by parallel resonance (with a parallel inductor; possibly L2) ?

You're absolutely right, I've just checked the maths and your workings are correct; due to being lazy I used an app to calculate the resonant inductor and managed to get about an extra 15nH. It still resonates with 162pF but the resistance is higher.
You might ask why 162pF and not 130pF for Coss? Well, you need to apply a scaling factor (I generally start with 1.2x Coss but you'll see anywhere from 1.1 to 1.5 times Coss in literature) before working out the resonant inductor. I read the data sheet at 135pF at 12.5v and so that times 1.2 is 162pF. This is one of the reasons you need to make L3 variable.

There are lots of ways to compensate for output impedance. If you want to do a pi match the drain capacitor actually has the effect of increasing the ouput impedance (a tiny LC match where the L is so small it can be ignored or is in the transistor lead itself).

I might have a go at building one of these amps for the guy that asked for a schematic; it's not really fair to put 2-3 versions of a theoretical design and expect someone with limited experience to build all of them and be a ginuea pig. When I get some time later I'll update the diagram and post it up here. Thanks for spotting the error!