Re: Completed 50 Watt PA, but...

Dennis Rieger

Mitchel,

I tried to operate my 50 watt PA on WSPR mode at full power. Both IRF510's failed. Hans reminded me that this amp is designed for CW with around 50% duty cycle. It is not designed to operate at 100% duty cycle in WSPR mode. Section 6.4 of the assembly manual (version 1.00k, page 49) discusses thermal considerations for this amp. Good thing it is easy to replace the IRF510's.

Dennis
KK5DB

Re: Audio "click" during knob rotation

Johan Bodin

Yes, that's true in the 16-bit world. 73 + 88 = FB is true in the world of hex :)

jerry@... wrote:

0xffff + 1 = 0

Re: Analyzing a PA

ajparent1/kb1gmx

Generally Ohms law is the rule.  Add RF to that and its true
but the numbers are now real and imaginary.   Generally back
of the envelope numbers set the ballpark and from there it is
dealing with all the small numbers that result.

However for push pull you simply factor times two
for load impedance and total voltage swing..

Google is your friend and try to find old ANs from Motorola
(Granberg mostly).

The magic is in the design and construction of the transformers needed.

A drain loaded to 1.6ohms would be 3.2 drain to drain for push pull.
In push pull voltage swing doubled, current is kept the same, and
load impedance is twice that of single ended.  Sometimes it helps to
thing of gett from 50 ohms down to what is needed.

Generally the big issue is load impedance (resistive) need to be lower
than whats needed to deliver power without clipping (hitting the rails).
However in real systems that's even lower as the device near cutoff and
saturation will be non-linear.

Also gate drive (if not bipolar) is voltage across a large reactive
(mostly capacitive) load.  So for example a device with a GM
of 1 Seimens (some devices higher)  would need 2V on the
gate to swing the drain 2A.  Doable but transformers need to
be correct.

For a bipolar its current so you need enough current with a
.65V standing BIAS that is voltage stiff.  That usually means
a 1-3ohm input impedance for 100W devices.  Also means
the bias source needs to have a impedance of under .1ohms.
So with a large signal Hfe of say 10 for 10 amps out, you need
1A of drive.  Typical 12V Bipolar power device has an Hfe in the
range of 15-30.   Effective gain is usually about ~11db.

Common case is IRF510,  Run at peak current of 4a and standing
current of 100ma we can call that 4A PP current [an upper limit].
At nominal 13.8V (socalled 12V system) that is a peak power
of less than 52W, note the case can dissipate 43W to a massive
heat sink.  However holding to a more moderate power we have
a few things to consider.  With lower current say 2A and push pull
we have 2A PP at roughly 20V PP or about 40W (peak)  or about
28W rms and that is in the easy zone.  However to get 2A at 20W
we need 10 ohms load per device or 20 push pull and translate
that to 50 ohms.  Transformer is 50/20 or impedance translation of
2.5 which is non integer ratio of 1:1.58 or about 3/2 which is how
most get the 2 turn primary with 3 turns output.  Too many turns
and inductance kills bandwidth.  That assumes the linearity is
adequate to have each device go to cutoff and swing to near 10V.
That is often not true.  If we wanted the same result with 24V supply
its easier as same 2a and nearly 40V PP only needs a 25 ohm load,
transformer is still a pain and the worst case peak of 48V is near half
the device breakdown voltage (IRF510 is 100V).

In both cases transformers to get into the low ohms range from
50 and reverse that need to be built for current in the 10-20A range
and inductances fairly low (10 ohms at 1.8mhz).   Those are tough
to meet.  Bottom line is for 100W (RMS out) you need at least
182W DC input assuming 55% efficient.  That 14A so I^2R is now
an enemy requiring all of the traces to be good for over 15A  with
low resistance, both signal trances, transformers, and ground plane.

FYI most of those china amps are at best 100W input DC.  A few will
do 100W but at 28, 36 or even 50V.  I have a few that were given to
me.  Lets call the numbers very optimistic same for useful bandwidth.

Allison
------
College text was wrong by 1976  as Silconix and others were delivering
150W devices called VMOS, DMOS, and the LDmos is the later
day follow on.  By time you got to read it it was technically out of date
on those devices.  I know as I used a Silconix VN60 in TO3 case that
did a very nice 60W at 13.5mhz for a commercial use.

Allison
-------------------------------
No private email, it goes to a bit bucket due to address harvesting

Re: QCX-mini failure on new build

Timothy Freeze

Hi Mont,

I have a cheap logic analyzer I bought on a whim.  No idea how to use it! 🙂

I just received a complete 80 M kit, built the display, and put the header pin in upside down.  I'm so dumb.  I had to remove the 10 pin header and resolder it in place properly.  Some of the thru hole copper came out, butit appears to work ok.  I finished putting it together and tried it in the QCX, well it gives the same result.  Sometimes ???or other characters and a flashing cursor to the right.  So I'd guess that means the main board is bad?

I'm thinking of just sucking it up and order another 20 M kit,  trying again and hoping for the best.  I loaded the complete firmware to another CPU and it didn't make any difference.   I haven't tried the new kit's cpu but I suspect I'd get the same result.

The twisted tabs on the LCD are not bent over.  They don't appear to be touching anything.

Here's the two displays, one connected with jumpers to the 10 pin header.  The other just on the QCX.

I'm sorry to be taking up so much of your time.

I'd be fine with testing the more complicated display on breadboard rails if that the thing to do.

Kindest regards,

Tim

From: QRPLabs@groups.io <QRPLabs@groups.io> on behalf of Mont Pierce KM6WT <de.km6wt@...>
Sent: Wednesday, June 23, 2021 12:34 PM
To: QRPLabs@groups.io <QRPLabs@groups.io>
Subject: Re: [QRPLabs] QCX-mini failure on new build

On Wed, Jun 23, 2021 at 06:19 AM, Timothy Freeze wrote:
I have a couple of blue 16x2 displays from my Aurdino collection.  Should I replace the display with one of them?

The 2nd one, with the 16 pin header will do nicely.

BEFORE we go on, there is one possibility that should be considered...  in constructing the mini, one of the twisted tabs under the LCD that hold it together has to be bent over to avoid hitting the alignment trimmer pots...  sometimes doing this causes the LCD issues, as it has internal contacts held together by pressure...  usually it causes missing rows of dots.  But, just wanted to mention it here.  Not sure what would happen if the tab broke off, or create too much pressure...

One test I would do, but it's just because I'm crazy...
I would hold the LCD with the 16 pin header up against the display board's LCD, with all the pins touching the corresponding display board LCD pads.  Then power on the mini.  Two display boards can work in parallel.
BE VERY CAREFUL THOUGH NOT TO SHORT PINS 1 & 2....
(Ok ok, bad idea... don't do this crazy stunt...)

Another alternative would be to use dupont breadboard jumpers, male-to-female, tack solder the male ends to the LCD pads on the display board, and plug the female ends onto the LCD with the 16 pin header.  Important:  you only need to do the 6 pins from each end.  The middle 4 pins are not used.  The two LCDs will work in parallel.
If you see same data on both, then data is corrupted (missing bits) before it gets to the LCD.
Could be weak or bad pins on the processor.

If you see good data on the attached LCD, then the display board's LCD must is probably bad...

A third approach, which is probably more reliable, but a bit more complicated, is to use a small breadboard with power rails.

Use jumpers to attach ground and VCC to the breadboard power rails.
Use jumpers to attach the LCD D7-D4, EN, RS, pins to their corresponding header pins on main board.
then for the other pins, duplicate the circuitry from schematic onto the breadboard.
LCD pin 16 can just be tied to ground.
LCD pin 15 needs R48, a 560 (+/- 200ohms, value not super critical, determines brightness of backlight)
LCD pin 1 and 5 also tied to ground.
LCD pin 3 needs a 3.3K to ground
LCD pin 3 also uses the wiper on 22k Pot between ground and VCC to control contrast
you can try different values of fixed resistors between VCC and pin 3

This would be a lot simpler I guess if you had another mini, or a friend close by with a mini, and could swap components back and forth to identify where it's failing...

This puzzle is a definite challenge....

You don't by chance have a Logic Analyzer?  One like this from Amazon (click here) would do.

73
km6wt

Re: Audio "click" during knob rotation

jerry@tr2.com

Sharc? Whazzat? IIRC the DSP we were using was a TMS320C31. We didn't
use any of the DSP features, it was just a good cheap CPU.

They sent me to a 1-week seminar to learn about it. In the class on
assembly language I remarked - "Of course, all real programming is done in
C". Ironically, I wound up doing my project in ASM.

Another cool thing I got to do - this was around 1987 - was to write
a program that converted a Compaq 286 into a recording oscilloscope.
That one really brought home the power of circular buffers - sized to a
power of 2. The CPU had segment registers, and each segment was max 64K.
In my sample-gathering interrupt routine, I used a 64K circular buffer.
That got me the wraparound for free - 0xffff + 1 = 0. Also - zero minus one
= 0xffff.

- Jerry

On 2021-06-23 11:29, Julian N4JO wrote:
Thanks for that story, Jerry. I know exactly where you're coming from
on the subject of optimization.
My story (and I'm not going to try to one-up you!) also includes
assembly language programming - 8086/87. It was a machine tool control
project that brought me over here to the 'States in 1982 for my
employer, Gettys Mfg., who made servo motors and drives, and a strange
control system called a "tracer/duplicator" which, when attached to a
milling machine, would copy models into real metal by means of tracing
over the model with a probe, and controlling servos to move a milling
cutter over a raw metal blank. Obviously we were eventually replaced
with CNC, but in our heyday our system was used extensively in the
automotive industry centered around Detroit/Windsor. and throughout UK
and Europe.
The project was canceled in '83 when the company was bought out, I
stayed here anyway; offshoot company started by original team to
continue the work; I rejoined that team in '89 I think it was, some
years after I'd earned my Green Card in '85, and I got involved with
the Mk II version, which is where the fun - and my story - starts.
capability), so I nailed all our existing code into a C framework. To
do that I had to remap the register usage and function call structure
to match the method the compiler used; quite a task, but at the end of
the day I could now write new C functions to be called from the
assembly language code under the C framework!
But yes, there was a lot of optimization to do. Our processor
complement increased from 2 x 8086/87 to 3x 286/287 on Intel Multibus
I, which was not much of an improvement because of bus/shared memory
access constraints. The REAL boost came when we graduated to a pair of
486DX, which had the math co-processor built in. That reduced the time
for the math co-pro context switch down from some 93 processor cycles
to just a handful. Given the amount of floating point math we need
(two and three dimensional trig), that made a heckuva difference. I
reduced our mainline interrupt loop time from 10ms down to 1.25 ms if
I recall, and the beast was still begging for more work to do!
Fun times.
Was your TI DSP the Sharc? I did some work on that while I was
software manager at Tech 80, up here in Minneapolis. We got bought out
and things went south, but it was fun while it lasted. --
Julian, N4JO.
------
[1] https://groups.io/g/QRPLabs/message/67672
[2] https://groups.io/mt/83622776/243852
[3] https://groups.io/g/QRPLabs/post
[4] https://groups.io/g/QRPLabs/editsub/243852
[5] https://groups.io/g/QRPLabs/leave/10312280/243852/1190336630/xyzzy

Re: Analyzing a PA

jerry@tr2.com

Whups, that's "current equals Power / Voltage".
100W / 9.05V = about 11A.

To pull 11A out of a 9V source, we need a load impedance of

The impedance ratio ( for one side of the center tap ) is 50/.82, or 61.
The turns ratio is square root of 61.
Let's call it 64 for a turns ratio of 8.
...but this is only one half of the primary.
So - final answer - turns ratio of 4. I think.

- Jerry

On 2021-06-23 08:56, jerry@... wrote:
All,
I have a sudden interest in calculating transformer ratios for
solid state power amplifiers. Mostly, because I have a Chinese kit
linear I'd like to interface to my QRP radios.
Not sure I remember how to do this, but here goes:
* Suppose I want to deliver 100W with a power supply of 13.8V. What
turns ratio do I need for the output transformer?
We have two devices in push-pull. Each device delivers 100W, but
only half the time. I'll do it for one device.
I'll assume that my device can swing its output voltage to within a
half volt of ground and Vdd. So its total available swing is 12.8V,
or 12.8V PEAK. Assuming a sine wave, the RMS
voltage is the peak voltage divided by the square root of two,
Here's where I get a little confused. For a single-ended amp, I
would say "Peak-to-peak swing is 12.8V" - peak voltage 12.8 divided by
2, rms = peak divided by squareroot-of-two. But for a push-pull
stage, each transistor delivers one ALTERNATION of the output wave.
So I guess I don't have to put in that divide by two to get from
peak-to-peak to peak.
So - we have a device delivering 9.05V (rms). To deliver 100W,
( for half the time ), current = voltage / power, or 50/9.05, or 5.52
Amps. And the desired load impedance would be R = E/I, or 9.05/5.52,
or
For the transformer, we have a center-tapped primary, with 1.6 ohms
impedance on each side of the center tap, or 3.2 ohms total. So our
impedance ratio is 50ohms / 3.2ohms, or 15.6. Let's call it 16.
Turns ratio is the square root of impedance ratio. So that gives us a
turns
ratio of 4.
Forty years ago, I could literally do this stuff in my head...
I was just reading a footnote in my 1976 "Devices and Circuits"
college textbook:
"* Most FETs have power ratings under 1 W, although a few are
available with ratings as high as 10 W."
- Jerry KF6VB

Re: Desoldering

Chuck ke9uw

I would also like to mention the Hakko desoldering gun. Heated tip with power vacuum pump inside. Works incredibly well. I desoldered a display 16 pins in 16 seconds. Pricey but you only have to buy one in your hobby career.

Re: Desoldering

Arv Evans <arvid.evans@...>

Had not considered flux as the dielectric in a homebrew capacitor...?

On Wed, Jun 23, 2021 at 1:57 PM Russ AB6MU <u.rusty@...> wrote:
You can find the a Flux Capacitor in November 5,1955.

Re: QCX: top row blocks on display, but CPU responds to ISP

Andrea Massucco IZ1IVA

Thank you gentlemen, unfortunately resoldering hasn't solved the issue. Since che CPU is working (as it's accessible via ISP), I'm tempted to blame the display assembly, perhaps the rubber contacts between the display and the PCB... if that's the case I fear I'd need a new display board.
--
73 de Andrea IZ1IVA - cq at iz1iva . net - http://iz1iva.net

Re: Desoldering

Russ AB6MU

I've had good success with solder wick and RA flux. I also use a solder sucker once in a while.

Re: Desoldering

Chuck ke9uw

Michael,
Everything I’ve found about it is just minimal flux.

Re: Desoldering

Russ AB6MU

You can find the a Flux Capacitor in November 5,1955.

Re: Desoldering

Arv Evans <arvid.evans@...>

On Wed, Jun 23, 2021 at 10:33 AM Podolsky A <mdjdmi@...> wrote:
But where do I find a working Flux Capacitor?

Re: NanoVNA and LPF?

Christopher Maness

He just does trial and error method in his video.  The only difference is doing it before you solder it in the board which I don’t like because the non zero chance of hosing up a trace during the desolder process.

KQ6UP

On Wed, Jun 23, 2021 at 10:40 AM Yury Krasouski <krasoffski@...> wrote:

Tune LPF is only half of the way.

You need to tune PA output by L4 to be 50 Ohms or at least close to this value and only than it makes sense to tune LPF with nano VNA.

Re: Shortage of Si5351A

Jim Sheldon

In keeping with several other replies to Pierre's topic here, I group I work with uses the 5351 in several of our projects but we don't use them in near the quantity that Hans @ QRP Labs does.

When they started to become scarce, I bought a few of the last ones available from Mouser, and when that batch got close to being used up, I went looking to see if there was another source.  I don't remember but I think Arv Evans over on the BITX20 reflector mentioned that the Chinese made an equivalent device called the MS5351M.  In searching for those to see if they were suitable, I came across a website called "Chipmall" (www.chipmall.com).  I checked them out and they actually had the MS5351M in stock, showing a quantity in excess of 5000.  Turns out I had to register as a customer before I could order, but after doing that, I took a chance and ordered 30 of them.  The price was less than \$1.00 USD, but their only shipping option was DHL at \$30 which was pretty expensive but after doing the math, \$0.77 USD each in 30 quantity - \$23.10 + \$30 = \$53.10 total cost for 30 of them delivered.  Now, divide \$53.10 by 30 and you get \$1.77 each which is not really that much more expensive than paying \$1.09 each + \$7.50 shipping from mouser (\$40.20 total for 30 and if you do the math on that, they come to \$1.34 each which is only 0.43 difference.)  That was well within range of keeping our close to break even point on target with no change in actual prices of the kit, we just dropped free postage and kept the kit price the same but added shipping & handling which actually saved a bit of money on our end as USPS has raised prices on shipping several times lately.

Now, once I got the initial order of 30 in, using one of our kits as the base, and finding an actual test socket for that SOT-10 format chip on Amazon, I designed and built a test set for the 5351.  Testing the MS5351M's against the Si5351a using the exact same parameters for both, and scope checking all 3 clocks, there turned out to be no significant difference between the two.  Even the I2C address of 0x60 was the same.

I'm posting this here so that anyone needing reasonable quantities of the 5351 chips can, if they like, obtain them at prices that aren't terribly elevated from what the Si5351a-b-gt was before the shortage.

Jim, W0EB

Re: Audio "click" during knob rotation

Thanks for that story, Jerry. I know exactly where you're coming from on the subject of optimization.

My story (and I'm not going to try to one-up you!) also includes assembly language programming - 8086/87. It was a machine tool control project that brought me over here to the 'States in 1982 for my employer, Gettys Mfg., who made servo motors and drives, and a strange control system called a "tracer/duplicator" which, when attached to a milling machine, would copy models into real metal by means of tracing over the model with a probe, and controlling servos to move a milling cutter over a raw metal blank. Obviously we were eventually replaced with CNC, but in our heyday our system was used extensively in the automotive industry centered around Detroit/Windsor. and throughout UK and Europe.

The project was canceled in '83 when the company was bought out, I stayed here anyway; offshoot company started by original team to continue the work; I rejoined that team in '89 I think it was, some years after I'd earned my Green Card in '85, and I got involved with the Mk II version, which is where the fun - and my story - starts.

I had to add a lot of extra features (including adding real CNC capability), so I nailed all our existing code into a C framework. To do that I had to remap the register usage and function call structure to match the method the compiler used; quite a task, but at the end of the day I could now write new C functions to be called from the assembly language code under the C framework!

But yes, there was a lot of optimization to do. Our processor complement increased from 2 x 8086/87 to 3x 286/287 on Intel Multibus I, which was not much of an improvement because of bus/shared memory access constraints. The REAL boost came when we graduated to a pair of 486DX, which had the math co-processor built in. That reduced the time for the math co-pro context switch down from some 93 processor cycles to just a handful. Given the amount of floating point math we need (two and three dimensional trig), that made a heckuva difference. I reduced our mainline interrupt loop time from 10ms down to 1.25 ms if I recall, and the beast was still begging for more work to do!

Fun times.

Was your TI DSP the Sharc? I did some work on that while I was software manager at Tech 80, up here in Minneapolis. We got bought out and things went south, but it was fun while it lasted.

--
Julian, N4JO.

Re: Desoldering

Johan Bodin

But where do I find a working Flux Capacitor?

Re: Audio "click" during knob rotation

jerry@tr2.com

On 2021-06-23 07:01, Julian N4JO wrote:
ency) are only really achieved when the inputs are
[...] In high hit-rate situations like encoder interfaces,
the increased overhead of interrupt handler architectures actually
reduces the performance,
I once was given a project with a TI DSP. I was given the hardware,
and created the firmware system from scratch. It needed to count Ethernet
packets. At first, I architected an interrupt driven system - one interrupt
per packet.

Then I realized that there really was no "main program" for the interrupt
to interrupt, and went to a polling loop instead.

The processor was underpowered for what it had to do, and I
pulled out all the stops to get it efficient. Things like
* Writing the whole thing in DSP assembler.
* structuring "If" statements so they usually failed. ( If the processor took a
jump, it would blow the instruction queue )
* Hand-unrolling small loops
* obsessing over the 64-instruction cache
* Pushing and popping ONLY registers that were actually used, for any
particular subroutine. This one was VERY labor intensive, and resulted
in lots of big comments at the tops of subroutines. But pushes & pops were expensive.

In the end, my processor kept up with a constant blast of minimum-length packets.

- Jerry KF6VB

Re: NanoVNA and LPF?

Yury Krasouski

Tune LPF is only half of the way.

You need to tune PA output by L4 to be 50 Ohms or at least close to this value and only than it makes sense to tune LPF with nano VNA.

Re: NanoVNA and LPF?

howard winwood G4GPF

Thats how I did mine, just used some perf board and socket strips aligned with filter pins,
Surprising how easy it is and you can see the performance of the filter as you adjust.

On 23 Jun 2021, at 17:12, Christopher Maness <christopher.maness@...> wrote:

Yeah.  I think I will try to build something out of perf board with some headers for the coils.  The caps are pretty easy to desolder from the perf board.

I guess I could find the VNA instructions for sweeping a LPF on YouTube.  I have not looked for one yet.

KQ6UP

On Wed, Jun 23, 2021 at 9:08 AM N3MNT <bob@...> wrote:
Nothing special;  I used the adapter in the link below and installed wire sockets on an old LPF board from Hans.  The simply look at the filter cut off VS frequency and adjust the windings until the knee is at the lowest frequency you are operating.  I used the sockets so I would not have to solder and remove the torroids to adjust.  This method does induce some error, but it was negligible,