Topics

QSX PA board

Hans Summers
 

Hi Simon 

I changed the subject as this is a worthy topic in its own right.

I hope to write up some details of the PA board in the next couple of days. This PA board is so GOOD that it will be made available also as a standalone 10W HF Linear PA kit. It is ready for production and should be the first of the QSX modules ready for sale.

The requirement for expensive Mitsubishi RD16HHF1 etc transistors in a decent amp is an absolute misconception. In reality they are very very expensive and not necessary in HF applications such as this.

This amplifier is proud to use two inexpensive IRF510 transistors. A great deal of attention has been paid to a symmetric and compact layout for best performance. 

I am very grateful to Allison KB1GMX for her advice, encouragement, support and testing during the development of this Linear PA.

More details soon but briefly:

1. Two IRF510 in push-pull, two BS170 in push-pull as the driver.

2. Gain 26-28dB, better than 2dB gain flatness from 2-30MHz. Still only 4dB down even at 6m, and respectable even to 4m.

3. Can easily achieve 10W across 160-10m using 12V supply

4. 80m band tested harmonic levels 2nd -38dBc, 3rd -31dBc. Thats BEFORE a low pass filter is added - addition of the LPF will result in a very clean output.

5. Very large heatsink for the 10W power rating. Ran continuous 100% duty cycle at 10W output for 1 hour without failure or degradation

6. Tested to 20V supply voltage; tested open load; tested shorted; tested with 20W 100% duty cycle for 10 minutes... all with no failure or degradation

7. Quite easily delivers a lot more than 10W too...

Above results are for the PA module alone, not integrated within the QSX. I think it is very unlikely that that you would be able to replace the PA module in the QSX with anthing that could deliver as good performance, even assuming you are willing to spend a lot more money.

Anyone who thinks IRF510s aren't good for the job, will need to reconsider their view after this.

73 Hans G0UPL



On Sun, Aug 19, 2018, 09:28 Kelly Jack <kellyjack1968@...> wrote:
Hi all,

Some observations:
Looking at the PA board - it shows traces leading to components for two of the leads with no trace visible for the other lead of the PA transistors. If this other lead goes to ground the pinout suggests that the PA transistors may be from the IRF510 family and not of the RD16HHF1 Mitsubishi type.
The PA board looks much smaller than a WA2EBY PA board and I'm wondering what layout/parasitic issues it may present, particularly at higher frequencies.
The PA board does look completely self contained though giving the option of any other PA board of your choosing.
Given the lack of details given, it appears this part of the circuit is still in development.

I don't know enough about bandscopes and what inputs they need but I am wondering whether there will be software onboard that enables hooking up a separate (no microcontroller) bandscope module either DIY or a later QRPLabs kit.

Really looking forward to this project.

Regards



Simon
VK3ELH












Kelly Jack
 

Hans,

Well I misread the lack of details matter!

Those specs, robustness and package size are amazing - and without the need for unobtanium (or even just expensive) parts  - truly impressive.

And decent at 4m and 6m!  Cherry on top.

I feel the qsx is going to make the unit numbers of the qcx look very small in comparison.

Regards 


Simon
VK3ELH 




Hans Summers
 

Hi Simon

Still on the back foot here... after the "YOTA leaks" hi hi. So details take time to sort out. Back from S.Africa on Thursday but I'm dealing with an episode of chronic bronchitis, my Mother visiting from England, 4 days of public holiday next week, and hundreds of backlogged emails accumulated due to my reduced attention during my week in S.Africa. Not to mention, having missed my 10-year wedding anniversary and two weddings while I was away... I owe some family time! So bear with me for a bit  ;-)

73 Hans G0UPL 


On Sun, Aug 19, 2018, 11:45 Kelly Jack <kellyjack1968@...> wrote:
Hans,

Well I misread the lack of details matter!

Those specs, robustness and package size are amazing - and without the need for unobtanium (or even just expensive) parts  - truly impressive.

And decent at 4m and 6m!  Cherry on top.

I feel the qsx is going to make the unit numbers of the qcx look very small in comparison.

Regards 


Simon
VK3ELH 




ajparent1/KB1GMX
 

Simon Et al,

I've been keeping very quiet on this.  This amp proves if you treat high gain devices
properly it behaves exceedingly well.  The IRF510 is a high gain device that works
at VHF.  Therefor it is a high gain VHF device.

That said things like full length leads are examples we see and also poor RF
power amp practice.  The leads if anything need to be as short as practically
possible.  Also input and output need to be opposing not intermixed.  

The WA2EBY amp one I also have is laid out with better care than many.
Most that I know building it add additional z-leads or vias and wrap top
ground to bottom to attain a good RF power layout without parasitics.
The most common with even lower power amps is high RF current
ground loops.  For a moment, an amp that runs at say 2A DC may
have many more amps of RF current flowing due to the requisite
low impedance required to obtain power.

Examples are OZ1PIF 500W 6M amp, my copy of one module at
225W(and lower voltage) and many single device amps for
6M QRP use.  (QRP being 10W SSB!)

Bottom line is with rare exception MOSFETS (trench, LD, T, and Hexfet) all have
VHF capability its where the ability to switch fast comes from.  It is the parameter
that helps make switch mode power supplies efficient.   MOSFETS have no
parameter like Bipolar types where their gain (FT) decreases with increasing
frequency.  However things like leads and package are important.

For those that whinge about large capacitance. You do not want go there.  Some
of the legendary RF parts have very large input and output capacitance.

All that said we have a part of choice if only for wide availability widespread use
and low cost.  But like all high gain devices that run at VHF layout is everything.

So Hans took the challenge and deigned as it it were a VHF device with VHF layout.
the result is stellar performance.

Design power 10W 3-30mhz  Its more like 2 to 70mhz.
Power out that exceeds design. amp is capable at 13.8V of more than 15W
and works well at 20V (note capacitor voltages) producing upwards of 25W. 

Gain is fairly flat, less than 2DB from 2-30mhz,  4db down at 50mhz and 
8db down at 70mhz.  The power Gain the 26-28db (50ohm).  Give it
the needed drive it produces power.  Power out is a product of gain and
available drive so with reasonable drive you get the expected linear power.

The low second harmonic output is a product of an physically balanced and
electrically balanced design.  Third harmonic is a artifact of push pull designs
but is low as its clean.

Thermal performance.  Large heat sink is critical to a IRF510 amp, the
TO220 case is a limiting factor.  Mechanical design can make the best
of it.  That allows the designer to obtain a good part of the 43 watt
dissipation the device is rated for.

A extended test run was conducted by accident, forgot it was on.
It got hot but didn't self destruct or become unstable even with the 
heat sink at 150F and the devices hotter.  Heat is a reality of linear
amplifiers.  The ability is to manage and withstand it is a performance
area.  This is needed for modes like RTTY and FT8 as they are very
high duty cycle modes.

Stability is however a important parameter.  Mismatched loads and reactive
loads were used to abuse the amp.  For those used to IRF510s taking off
this must be a surprising change.   It very hot and have not yet killed
part.  How many have IRF510 designs in their hands that will run at
100ma, 200ma, or even 250ma and not go unstable?  

I'm still working with this amp but the only thing I've used that behaves
near as well are instrumentation block amps.  In once case it outran
a Qbit block amp.

Allison







geoffrey pike
 

IMD figures?
cheers
Geoff
GI0GDP

On Sunday, 19 August 2018, 16:31:32 BST, ajparent1/KB1GMX <kb1gmx@...> wrote:


Simon Et al,

I've been keeping very quiet on this.  This amp proves if you treat high gain devices
properly it behaves exceedingly well.  The IRF510 is a high gain device that works
at VHF.  Therefor it is a high gain VHF device.

That said things like full length leads are examples we see and also poor RF
power amp practice.  The leads if anything need to be as short as practically
possible.  Also input and output need to be opposing not intermixed.  

The WA2EBY amp one I also have is laid out with better care than many.
Most that I know building it add additional z-leads or vias and wrap top
ground to bottom to attain a good RF power layout without parasitics.
The most common with even lower power amps is high RF current
ground loops.  For a moment, an amp that runs at say 2A DC may
have many more amps of RF current flowing due to the requisite
low impedance required to obtain power.

Examples are OZ1PIF 500W 6M amp, my copy of one module at
225W(and lower voltage) and many single device amps for
6M QRP use.  (QRP being 10W SSB!)

Bottom line is with rare exception MOSFETS (trench, LD, T, and Hexfet) all have
VHF capability its where the ability to switch fast comes from.  It is the parameter
that helps make switch mode power supplies efficient.   MOSFETS have no
parameter like Bipolar types where their gain (FT) decreases with increasing
frequency.  However things like leads and package are important.

For those that whinge about large capacitance. You do not want go there.  Some
of the legendary RF parts have very large input and output capacitance.

All that said we have a part of choice if only for wide availability widespread use
and low cost.  But like all high gain devices that run at VHF layout is everything.

So Hans took the challenge and deigned as it it were a VHF device with VHF layout.
the result is stellar performance.

Design power 10W 3-30mhz  Its more like 2 to 70mhz.
Power out that exceeds design. amp is capable at 13.8V of more than 15W
and works well at 20V (note capacitor voltages) producing upwards of 25W. 

Gain is fairly flat, less than 2DB from 2-30mhz,  4db down at 50mhz and 
8db down at 70mhz.  The power Gain the 26-28db (50ohm).  Give it
the needed drive it produces power.  Power out is a product of gain and
available drive so with reasonable drive you get the expected linear power.

The low second harmonic output is a product of an physically balanced and
electrically balanced design.  Third harmonic is a artifact of push pull designs
but is low as its clean.

Thermal performance.  Large heat sink is critical to a IRF510 amp, the
TO220 case is a limiting factor.  Mechanical design can make the best
of it.  That allows the designer to obtain a good part of the 43 watt
dissipation the device is rated for.

A extended test run was conducted by accident, forgot it was on.
It got hot but didn't self destruct or become unstable even with the 
heat sink at 150F and the devices hotter.  Heat is a reality of linear
amplifiers.  The ability is to manage and withstand it is a performance
area.  This is needed for modes like RTTY and FT8 as they are very
high duty cycle modes.

Stability is however a important parameter.  Mismatched loads and reactive
loads were used to abuse the amp.  For those used to IRF510s taking off
this must be a surprising change.   It very hot and have not yet killed
part.  How many have IRF510 designs in their hands that will run at
100ma, 200ma, or even 250ma and not go unstable?  

I'm still working with this amp but the only thing I've used that behaves
near as well are instrumentation block amps.  In once case it outran
a Qbit block amp.

Allison







ajparent1/KB1GMX
 

IMD is a ways off as I have to graft it to to a radio!
I can do two tone with discrete signals but that is not what is of interest.

In a general sense if the device is linear and not run to 1db
compression then the IMD will be good.

It helps that this amp runs well with high bias so its more class
AB or AB1 (100-200ma device) which really helps.

Compared to the whole lot out there running IRF510s
at 50ma or hitting compression due to current or voltage
limiting it is expected to be much better.

Devices like RD16HHF want 200-300ma for reasonable IMD as well.

Allison

Clint Sharp <cjaysharp@...>
 

I've played with switching FQP13N10 FETs in PAs and found them to be capable of lots of power for very little money so I was of the opinion that they could be made into a good PA but I found they could also be difficult to tame, I managed to let the smoke out of a few while experimenting with bias, VDss limits etc. for instance and I know when you get it wrong they produce all sorts of nasties and/or explode in spectacular fashion so, while I was sure they could be made to behave and work well, I wasn't able to get them to so I'm fascinated by the work that's gone into the QSX PA and can't wait to get my hands on at least the PA and preferably the whole radio.

I was also confused about gate capacitance as I'd noticed that some 'proper' RF FETs have much higher gate/source capacitance in the data sheets, it seems logical to me that it should be lower in a switching FET as the main challenge is in the driver design, it has to be able to sink and source current fast to get the device out of the linear region as fast as possible to minimise dissipation, in a FET designed for linear operation it's not quite as important? 


On Sun, 19 Aug 2018 at 18:08, ajparent1/KB1GMX <kb1gmx@...> wrote:
IMD is a ways off as I have to graft it to to a radio!
I can do two tone with discrete signals but that is not what is of interest.

In a general sense if the device is linear and not run to 1db
compression then the IMD will be good.

It helps that this amp runs well with high bias so its more class
AB or AB1 (100-200ma device) which really helps.

Compared to the whole lot out there running IRF510s
at 50ma or hitting compression due to current or voltage
limiting it is expected to be much better.

Devices like RD16HHF want 200-300ma for reasonable IMD as well.

Allison



--
Clint. M0UAW IO83

No trees were harmed in the sending of this mail. However, a large number of electrons were greatly inconvenienced.

Kees T
 

Allison,

Yes, that is correct. I have found out that compression levels are only acceptable if you run higher quiescent current and Class AB1. That's also the only way to maintain linearity in the PA but it will require a BIG heatsink. Makes "digital modes" quite inefficient, power wise.....except for CW.

73 Kees K5BCQ

ajparent1/KB1GMX
 

Clint,

The gate source capacitance is a product of many parallel gates on many parallel devices
effectively.  The structures (LD, T, HEX) vary but that is the easy explanation.  So the way
you deal with it is being able to supply current and lots of it for an instant.  Its a challenge
and there are parts from Ixys and others to address that.  Its a trade for high switching
speed and very high power gains.  It also means the paths around the driver and
source/gate have to behave like a low impedance 300mhz RF circuit.

For RF that means current at RF or power.  You have to treat the gate as a low impedance.
The problematic capacitance for a lot of MOSFETs is the gate to drain capacitance
that also results from the construction.  That can reach proportions that can be difficult
to handle.  Some call it Miller capacitance.  It represents negative feedback and if the
gate circuit is not low enough in impedance it can dominate and cause low gain or if the
gate circuit has enough reactance you get oscillation from  the phase shift.  Very often
that oscillation is fully destructive.  Typical reason why is that feedback path is resonant
and the gate voltage gets high enough to punch the oxide and death is immediate.
The voltage needed to do that is less than 20V.  

So feeding a MOSFET from a bias (DC) perpective is a its an open circuit with nanoamps 
of current.  At 14mhz its a 28ohm reactive load that the driver has to push power into and
maintain.  Many get that part wrong.  With the gain and bandwidth of the part failing to do 
that means its going to take off, and die.

If you look at RF circuits using RF MOSFETs (typically LDmos) you immediately see low
input impedance and usually a smith chart or S parameters that talk about that.
FYI the input impedance are typically in the same range as bipolar devices in of the
same power.

The FQP13N10 is a really nice part.

FYI switching MOSFETs are optimized for the RDSon or the ability to be saturated hard 
enough to look like a very fast mechanical switch.  Where a MRF150 not so much
but even then it can be!  An example a MRF140 amp the bias supply regulator failed
the the gate went to about 8-9V, drain current was over 20A just before the power
supply went away (fried) the mosfet however didn't even get warm or damaged!

Allison

Kelly Jack
 

Allison,

Not sure if this should be for you but do you know if there will be a writeup on the design of the PA circuit akin to the one Hans did for the QCX?

From what I've read I feel like this PA circuit will earn itself its own place in the annals of ham radio history and the design process and considerations would be of benefit to document (fully and in one place).

73



Simon
VK3ELH

Hans Summers
 

Hi Simon,

I will try to write it up nicely in the documentation. I think there isn't magic. 

Layout is very important. The PCB is compact and there is groundplane everywhere as much as possible, and wide traces. Symmetric as possible. Lead lengths are short, particularly the IRF510 leads which with only about 1mm between the component body and PCB traces. 

There's nothing new here...the importance of good layout is well known, just not often thoroughly practiced. 

Interestingly the 10W Linear PA uses only through-hole components - thereby also busting another myth that the short  (almost non-existent) leads of SMD components are a necessity in a PA for HF. 

The most important points:

1. Lowly IRF510s can do an excellent job at HF and even into VHF, if they are treated properly. Expensive Mitsubishi RD-whatsits are neither necessary nor desirable.

2. It can be done even with through-hole components. And 2-layer PCB. Low cost, easy to build!

3. Layout is critical. Take short-cuts here, expect to be punished, the RF will not forgive you. The techniques I used for the PA layout, are the same as I implemented on a quadband cellular GSM board (850/900/1800/1900MHz).

73 Hans G0UPL 

On Fri, Aug 24, 2018, 11:12 Kelly Jack <kellyjack1968@...> wrote:
Allison,

Not sure if this should be for you but do you know if there will be a writeup on the design of the PA circuit akin to the one Hans did for the QCX?

From what I've read I feel like this PA circuit will earn itself its own place in the annals of ham radio history and the design process and considerations would be of benefit to document (fully and in one place).

73



Simon
VK3ELH

Kelly Jack
 

Thanks Hans,

Every magic trick is unremarkable to the skilled magician, while the dilettante audience (me at least) looks on in wonder.

Looking forward to your writeup.

73


Simon
VK3ELH 



 

ajparent1/KB1GMX
 

Simon,

Hans did the work, for his radio design.  It would be out of place for me to do more
than comment on it.  

That being said he did the work!  By treating the device as a high performing
RF component and applied RF design rules.  That is how to get the most
from it.   Its also a very conservative design,  By not running wide open
and applying reasonable feedback you get predictable results.  Not magic,
no miracles, its applied electronics.  Nicely done too.

What works at near DC (say 3.5mhz)  is an abomination at 30mhz and flat
stops working at 50mhz.  This is especially true in low impedance circuits 
where even small amount of resistance and inductance become significant.
We do not have to go far to see bad examples.  Yet we keep seeing them
copied.   It strikes me odd no one says this is not working. 

Allison

ulidf5sf2002
 

Hello Hans,

have you meassured IMD3 and IMD5 on all band ?
I am curious about it.

Tanks
Uli
DF5SF

ajparent1/KB1GMX
 

Uli,

Why?  It will likely be better than some IRFxxx based
if only because its not run to -1db compression and 
it behaves well with high drain currents (100-300ma
bias per device) which is typical of most MOSFET amps.

Since its not a external power amp and part of the
radios power chain its hard to test standalone and have
valid results.  I'd like to test it but an exciter is not handy
at this point for two tone testing (is is an exciter
dependent number).

Allison