|
One more thing on this. Tony AC9QY and I concluded that using a 100 nF capacitor to bypass the center tap of T1's primary to ground, combined with replacing L14 with a 2 ohm resistance, was a good thing.
The inductor L14 adds no value in this circuit, and even contributes a dangerous LdI/dT spike at the end of transmission. The 2 ohm resistor in its place offers some additional protection against low-Z high SWR, which is the high current case. The PA self-throttles in response to excessive current. The loss of power under a 50 ohm load is insignificant.
The combination of high transistor breakdown voltage and high current protection makes the PA nearly bulletproof regarding SWR. Simulation demonstrates that nicely.
JZ KJ4A?
On Fri, May 12, 2023, 6:51 AM John Zbrozek < jdzbrozek@...> wrote: Hello Tony,
Another Tony. AC9QY, and I have been running LTSpice simulations of the QDX PA using TN0110 transistors. They are similar to the Microchip TN0106 transistors that you have been testing,? differing mainly in that they have a 100 volt drain breakdown specification vs. 60 volts.
The simulation results look extremely encouraging, even? on 10m, but we had some questions about the adequacy of the real drive level available to the gates.
Learning of your actual results in real hardware is fabulous! Tip of the hat for trying that! Bravo!
If you are an LTSpice user and interested in running the simulation I would be delighted to send you the .asc file for it.
73, JZ KJ4A?
I have been experimenting with different transistors for the PA in my HB and LB QDXs. We have been having heavy rains the past two days and my antenna is providing some bad mismatches of up to 2.8 to 1. I used this condition to test several transistors for
robustness and have good results to report. The high SWR conitions I used were caused by load impedances less than 50 ohms as measured at the transmitter end of the coax with my nanovna.
First off, the BS170s do not tolerate these kinds of mismatches and blow quickly. But I have been unable to blow up either the TN0106 or VN0606 transistors. Note that these two both have reverse pinout with respect to the BS170 so the flat face has to be mounted
upwards such that it is pressed against the washer instead of the ground plane. This does not seem to bother the transistors and they run cool with a normal 50 ohm load. My units are wired for 9 volts and that is how I did my testing. I operated the VN0606
for an hour at 2.5 to 1 SWR. While the unit gets warm, it does not fail. The TN0106 behaves similarly but I seem to get more drive into the low Z loads and the transistor heats up a bit more. AT 50 ohms, the two transistors behave similarly.?
Neither QDX has any other modifications from stock. No diodes or capacitors are added. Just the transistors have been changed.
The transistors were ordered from Mouser.
I also tried out the 2N6660. This is a higher power device in a TO-39 can. While it works, it provides much lower output, even into 50 ohms (on the order of 1 watt). I tried using just two devices but that made even lower power. These are expensive and the
QDX does not provode enough drive for them.
The 10 meter performance of both the TN0106 and VN0606 seems to be better than the BS170.
I would be interested in hearing of peple trying these on their 12 volt units.
Tony
AD0VC
I have only had my HB for about a month and I have blown a couple of PAs due to matching issues. That aside, this QDX is FANTASTIC! The design is remarkable and well executed. I love the electronic aspects of the hobby so I always end up fiddling with things
but I would be proud of this design if it were mine. Just amazing what was done for the price. This thing makes me happy! I have ordered another for the low bands.
Tony
AD0VC
Dave et Al,
I have used the first prototype High Band since I designed the filters last year. I used at the home QTH and then took it to QRP Labs HQ for assessment and ultimate implementation. I have used that and another HB almost continuously. I¡¯ve operated with
it portable in Prague, Rome and Istanbul with various wire antennas out hotel windows on fishing rods and wires on lead sinkers slingshotted into local trees. I used to use a small 12V PS with a variable buck down and meters but now, due to noise reduction
measures, I run all my QDXes and RaspberryPIs on Lithiums, 2 in series that start at 8.4V and I usually change them and recharge when somewhere at 7.3V. RasPis I use a 30,000mAH power bank.?
The one thing I am very careful about is antenna matching. I used to use LC matches with plastic printed spherical variometers as the variable L. I now use FRI-matches with 3D printed large toroids. I also have a switch that puts a LED swr bridge in line
and a switched LED current monitor antenna wire.?
The first and original HB QDX is still going strong! It was built for 9V but usually operates at an average of 7.8V.?
Careful management of the antenna match is the most critical to long QDX life!
73
Ross
6
?
Hi Dave, all
The washer isn't the heatsink. The heatsink is the PCB. The washer and bolt are just the mechanical means of bolting the flat side of the transistors to the PCB.
As others have said... Thermal paste doesn't harm but doesn't make a lot of difference, either. Be careful too because some pastes are electrically conductive and that won't go well. Make sure you don't use that.
Personally I have found the QDX PA reliable and durable as have many others. A lot of failures have posthumously been attributed to sustained chronic SWR mismatch or acute mismatch during auto tuner use, supply spikes etc., over-voltage, including
from fully charged 12V-nominal batteries.
Remember also that on any forum such as this one, you will tend to hear far more frequently from people with a problem, than those without. You tend to get the mis-impression consisting of the proportion of people with problems, out of the collection
of people with problems. The actual proportion of people with problems, out of the total of all users, is a lot lot lower (and difficult to quantify).
73 Hans G0UPL
On Apr 29, 2023, at 8:45 AM, Dave Sergeant G3YMC < dave@...> wrote:
Humour aside Razvan does have some valid points. I have not yet put my
QDX into transmit mode but already have concerns about the design of
its PA stage. The requirements for not exceeding 12V (in a 12V HB
version) or daring to transmit into anything other than the religious
50 ohm load is an indication that the design is maybe just a tad close
to the limits for comfort. It almost made me laugh when I attached the
little washer to the rounded side of the BF170s asking myself if I had
mounted them upside down. Surely you should put the heatsink on their
flat side I asked.
In the desire to make the thing as small and cheap as possible it seems
common sense has been overlooked. A couple of TO220 case devices on a
slighly bigger board and a bit more metal to heatsink them would have
made a far more robust PA and removed the need for all these
discussions on this reflector.
No criticism of Hans of course, the QDX is an incredible little rig,
but as an engineer in my professional life I would have not done it
that way.
73 Dave G3YMC
On 28 Apr 2023 at 15:15, DL2ARL wrote:
you could also sprinkle over the BS170s some Water out of the Holy
Fountain of Lourdes from
southwestern France. It would help as much as thermal paste, but it
will have the big advantage of
not doing no harm and cost less.
|
I came to the same conclusion after I burned my PA...
I have installed 2x1Ohm in place of L14, 100n+10n from the T1 center tap do ground. Still have around 3.5W out.
Now the hot spot is T1, I wonder if wiggling with windings can improve something. On the other side less loss on the T1 may equal higher load to the BS170 team.
73
Pawe? SO8FM
|
Pawel, T1 is running hot? Is it hot on all bands? Surprised...JZ KJ4A?
On Sat, May 13, 2023, 11:26 AM Pawe? Hadam < pawel@...> wrote: I came to the same conclusion after I burned my PA...
I have installed 2x1Ohm in place of L14, 100n+10n from the T1 center tap do ground. Still have around 3.5W out.
Now the hot spot is T1, I wonder if wiggling with windings can improve something. On the other side less loss on the T1 may equal higher load to the BS170 team.
73
Pawe? SO8FM
|
Yes, all bands, QDX connected to 50Ohm dummy load.
Originally wounded for 12V operation, now for 9V. Still looks like it is the first component getting warm. It depends how we can define hot, I did not measured. I could still keep my finger on it.
I already took it out, I will try to wind it again. I want to try 2 cores configuration, something like in one of the mcHF mods. Heat is not big issue but If I could further reduce transistors load keeping the same output it would be nice.
Pawel SO8FM
|
Pawel,
If you unwind T1, inspect for a possible shorted turn. There must be a good reason why T1 is getting hot. Good luck with your experiments...JZ
On Sat, May 13, 2023 at 5:03?PM Pawe? Hadam < pawel@...> wrote: Yes, all bands, QDX connected to 50Ohm dummy load.
Originally wounded for 12V operation, now for 9V. Still looks like it is the first component getting warm. It depends how we can define hot, I did not measured. I could still keep my finger on it.
I already took it out, I will try to wind it again. I want to try 2 cores configuration, something like in one of the mcHF mods. Heat is not big issue but If I could further reduce transistors load keeping the same output it would be nice.
Pawel SO8FM
|
Hello John The? TN0110 sounds like a bullet-proof upgrade for the PA transistors. I¡¯m curious to see it¡¯s performance, especially on 10m.
-- Sverre
LA3ZA
http://la3za.blogspot.com
|
Hello Sverre, Tony S., AC9QY, gets the credit for finding that one after a deep dive into? search tools and many datasheets.
It appears to offer improvement in nearly every way except for input capacitance. I was delighted to see Tony AD0VC have success with it's 60V sibling even on 10m. In the worst case of maximum transistor gate capacitance and IC5 2x output current at spec minimum it likely would not work on 10m. Someone willing to play the statistical odds might again meet with success on 10m. For a QDX-LB there would be no performance problem.
Tony S.? AC9QY is now working to find a solution to that gate capacitance problem. Wish him good luck!
I was also delighted to see Pawel's comments about the success of the throttling scheme to protect the PA against low Z / high SWR. Nice!
73, JZ KJ4A?
On Sat, May 13, 2023, 11:59 PM Sverre Holm < svholm54@...> wrote: Hello John
The?TN0110 sounds like a bullet-proof upgrade for the PA transistors. I¡¯m curious to see it¡¯s performance, especially on 10m.
--
Sverre
LA3ZA
|
The TN0110 does not appear to be available from my usual suppliers (Mouser, Digi-Key). Mouser shows some expected in August. The capacitance spec for the TN0106 is about the same as the TN0110, 50pf. The ACT08 may switch slower but that can be offset by the
fact that the transistor needs less drive for a given drain current.
Tony
AD0VC
toggle quoted message
Show quoted text
Hello Sverre,
Tony S., AC9QY, gets the credit for finding that one after a deep dive into? search tools and many datasheets.
It appears to offer improvement in nearly every way except for input capacitance. I was delighted to see Tony AD0VC have success with it's 60V sibling even on 10m. In the worst case of maximum transistor gate capacitance and IC5 2x output current
at spec minimum it likely would not work on 10m. Someone willing to play the statistical odds might again meet with success on 10m. For a QDX-LB there would be no performance problem.
Tony S.? AC9QY is now working to find a solution to that gate capacitance problem. Wish him good luck!
I was also delighted to see Pawel's comments about the success of the throttling scheme to protect the PA against low Z / high SWR. Nice!
73, JZ KJ4A?
On Sat, May 13, 2023, 11:59 PM Sverre Holm < svholm54@...> wrote:
Hello John
The?TN0110 sounds like a bullet-proof upgrade for the PA transistors. I¡¯m curious to see it¡¯s performance, especially on 10m.
--
Sverre
LA3ZA
|
That's correct, Tony. The lower threshold?voltage of this family buys a lot of goodness! Your independent?find of these devices, turned into successful hardware, stands as proof.
Someone?here once suggested that in case of inadequate drive a clone IC5 could be soldered on top of the existing IC5, giving 4x the output of a single logic gate. Inelegant? Sure, but it probably would work. Not for trying by those with less-than-de luxe soldering tools!
JZ KJ4A
On Sun, May 14, 2023 at 4:41?AM mux_folder2001 < canthony15@...> wrote:
The TN0110 does not appear to be available from my usual suppliers (Mouser, Digi-Key). Mouser shows some expected in August. The capacitance spec for the TN0106 is about the same as the TN0110, 50pf. The ACT08 may switch slower but that can be offset by the
fact that the transistor needs less drive for a given drain current.
Tony
AD0VC
Hello Sverre,
Tony S., AC9QY, gets the credit for finding that one after a deep dive into? search tools and many datasheets.
It appears to offer improvement in nearly every way except for input capacitance. I was delighted to see Tony AD0VC have success with it's 60V sibling even on 10m. In the worst case of maximum transistor gate capacitance and IC5 2x output current
at spec minimum it likely would not work on 10m. Someone willing to play the statistical odds might again meet with success on 10m. For a QDX-LB there would be no performance problem.
Tony S.? AC9QY is now working to find a solution to that gate capacitance problem. Wish him good luck!
I was also delighted to see Pawel's comments about the success of the throttling scheme to protect the PA against low Z / high SWR. Nice!
73, JZ KJ4A?
On Sat, May 13, 2023, 11:59 PM Sverre Holm < svholm54@...> wrote:
Hello John
The?TN0110 sounds like a bullet-proof upgrade for the PA transistors. I¡¯m curious to see it¡¯s performance, especially on 10m.
--
Sverre
LA3ZA
|
Hans,
do you plan to create a watertight enclosure?
This should work...maybe even for your competitions...
Giuseppe Marullo
IW2JWW - JN45RQ
PS: always to make some PI4 take a bath in it and overclock...never dared also because of FC77 prices
|
Tonight it is raining again and the SWR on 30 meters is 3.2 to 1 with the antenna impedance on the low side of 50 ohms, 15.6 ohms to be more exact. I have my low band QDX fitted with the VN0606 transistor and ran it for 30 minutes under this condition with
no ill efects. QSOs with Italy (I am in the Denver, CO area) and reports of +05 from the west coasters.?
The meter read 1.9 watts but this is basically a voltmeter that assumes a 50 ohm load. If you apply the ratio of 50/15.6 (ie the SWR) to this you get an actual power being delivered of 6 watts. Pretty impressive if you ask me. It ran a bit warm but not too
bad.
Tony
AD0VC
toggle quoted message
Show quoted text
From: Anthony Casorso <canthony15@...>
Sent: Sunday, May 14, 2023 2:41 AM
To: [email protected] <[email protected]>
Subject: Re: [QRPLabs] 3 tips for not blowing the finals of the QDX #qdx
?
The TN0110 does not appear to be available from my usual suppliers (Mouser, Digi-Key). Mouser shows some expected in August. The capacitance spec for the TN0106 is about the same as the TN0110, 50pf. The ACT08 may switch slower but that can be offset by the
fact that the transistor needs less drive for a given drain current.
Tony
AD0VC
Hello Sverre,
Tony S., AC9QY, gets the credit for finding that one after a deep dive into? search tools and many datasheets.
It appears to offer improvement in nearly every way except for input capacitance. I was delighted to see Tony AD0VC have success with it's 60V sibling even on 10m. In the worst case of maximum transistor gate capacitance and IC5 2x output current
at spec minimum it likely would not work on 10m. Someone willing to play the statistical odds might again meet with success on 10m. For a QDX-LB there would be no performance problem.
Tony S.? AC9QY is now working to find a solution to that gate capacitance problem. Wish him good luck!
I was also delighted to see Pawel's comments about the success of the throttling scheme to protect the PA against low Z / high SWR. Nice!
73, JZ KJ4A?
On Sat, May 13, 2023, 11:59 PM Sverre Holm < svholm54@...> wrote:
Hello John
The?TN0110 sounds like a bullet-proof upgrade for the PA transistors. I¡¯m curious to see it¡¯s performance, especially on 10m.
--
Sverre
LA3ZA
|
About the T1 getting hot:
/g/QRPLabs/topic/qdx_t1_heat_thermal_design/91641546
|
Vincent et al, Revs 1,2 and 3 use 330 ohm resistors at R2, R4. The referenced power calculation of 338 mW per resistor would be valid and excessive heat would occur. Revs 4 and 5 reduced the value of these resistors to 33 ohms. The current through?each is approx 38 mA. The power dissipated?then is only 48 mW per resistor, well within spec.
JZ
|
On 15/05/2023 09:51, John Z wrote: Revs 1,2 and 3 use 330 ohm resistors at R2, R4. The referenced power calculation of 338 mW per resistor would be valid and excessive heat would occur. John, The surprising thing is that this does not seem to be causing many failures. It was noticed long ago, some did modify theirs but the subject does not surface very often. (The rev 4/5 upgrade was a big change, not easy to copy.) I guess excessively hot components will have a shorter life. 73 Alan G4ZFQ
|
Alan,
Glad to hear it has not been a big problem. Those resistors may be getting some protection from the thermal mass in the neighborhood. If T1 is taking up some of the heat and then losing it, that helps. I think the risk is more with the thermal cycling and fracture of solder joints than of the resistors.
JZ
On 15/05/2023 09:51, John Z wrote:
> Revs 1,2 and 3 use 330 ohm resistors at R2, R4. The referenced power
> calculation of 338 mW per resistor would be valid and excessive heat
> would occur.
John,
The surprising thing is that this does not seem to be causing many
failures. It was noticed long ago, some did modify theirs but the
subject does not surface very often.
(The rev 4/5 upgrade was a big change, not easy to copy.)
I guess excessively hot components will have a shorter life.
73 Alan G4ZFQ
|
On Mon, May 15, 2023 at 05:51 AM, John Z wrote:
Revs 1,2 and 3 use 330 ohm resistors at R2, R4....
On the earlier QDXs I have I changed them to "805" sized 330 ohm SMD resistors, 1/2W. They fit within the pads. I can't remember if I set them on edge or just flat. --Al WD4AH
|
John,
it may not be obvious, but substituting a 2 ¦¸ resistor for L14 changes
the principle of operation of the Class D PA. As
originally designed, it is a current switching PA, and
requires a parallel-tuned output network. As modified it has
become a voltage switching PA, and requires a series-tuned
output network. Since the original QDX PA design was for a current
switching PA, this may explain why some have observed the output
filter getting hot.
These two classes are noted in the RSGB Handbook (12th edition,
p10.6) and Device Evaluation for Current-Mode Class-D RF Power
Amplifiers to be found at
at figures 3.4 and 3.5. [There are also many IEEE papers on class
D, but these seem to be behind the usual IEEE paywall.]
L14 is employed to act as a current source. Thus when the output
filter sees 50 ¦¸ resistive, a similar load will appear at L14, but
when things are reactive, there is the potential for undesirable
voltages at the PA MOSFETs. (A simple measure of SWR says nothing
as to how reactive the load might be.)
All that is not to say that a re-designed PA stage might not be
more resilient against non-50 ¦¸ loads, but it is not a simple
change. The original design has a great deal going for it. (And
even valve (tube) PAs can flash-over with too much SWR.)
73, Stay Safe,
Robin, G8DQX
On 12/05/2023 12:12, John Z wrote:
toggle quoted message
Show quoted text
One more thing on this. Tony AC9QY and I concluded
that using a 100 nF capacitor to bypass the center tap of T1's
primary to ground, combined with replacing L14 with a 2 ohm
resistance, was a good thing.
The inductor L14 adds no value in this circuit,
and even contributes a dangerous LdI/dT spike at the end of
transmission. The 2 ohm resistor in its place offers some
additional protection against low-Z high SWR, which is the
high current case. The PA self-throttles in response to
excessive current. The loss of power under a 50 ohm load is
insignificant.
The combination of high transistor breakdown
voltage and high current protection makes the PA nearly
bulletproof regarding SWR. Simulation demonstrates that
nicely.
JZ KJ4A?
On Fri, May 12, 2023, 6:51 AM
John Zbrozek < jdzbrozek@...>
wrote:
Hello Tony,
Another Tony. AC9QY, and I have been running
LTSpice simulations of the QDX PA using TN0110
transistors.
They are similar to the Microchip TN0106
transistors that you have been testing,? differing mainly
in that they have a 100 volt drain breakdown specification
vs. 60 volts.
The simulation results look extremely
encouraging, even? on 10m, but we had some questions about
the adequacy of the real drive level available to the
gates.
Learning of your actual results in real
hardware is fabulous!
Tip of the hat for trying that! Bravo!
If you are an LTSpice user and interested in
running the simulation I would be delighted to send you
the .asc file for it.
73, JZ KJ4A?
I
have been experimenting with different transistors for
the PA in my HB and LB QDXs. We have been having heavy
rains the past two days and my antenna is providing
some bad mismatches of up to 2.8 to 1. I used this
condition to test several transistors for robustness
and have good results to report. The high SWR
conitions I used were caused by load impedances less
than 50 ohms as measured at the transmitter end of the
coax with my nanovna.
First
off, the BS170s do not tolerate these kinds of
mismatches and blow quickly. But I have been unable to
blow up either the TN0106 or VN0606 transistors. Note
that these two both have reverse pinout with respect
to the BS170 so the flat face has to be mounted
upwards such that it is pressed against the washer
instead of the ground plane. This does not seem to
bother the transistors and they run cool with a normal
50 ohm load. My units are wired for 9 volts and that
is how I did my testing. I operated the VN0606 for an
hour at 2.5 to 1 SWR. While the unit gets warm, it
does not fail. The TN0106 behaves similarly but I seem
to get more drive into the low Z loads and the
transistor heats up a bit more. AT 50 ohms, the two
transistors behave similarly.?
Neither
QDX has any other modifications from stock. No diodes
or capacitors are added. Just the transistors have
been changed.
The
transistors were ordered from Mouser.
I
also tried out the 2N6660. This is a higher power
device in a TO-39 can. While it works, it provides
much lower output, even into 50 ohms (on the order of
1 watt). I tried using just two devices but that made
even lower power. These are expensive and the QDX does
not provode enough drive for them.
The
10 meter performance of both the TN0106 and VN0606
seems to be better than the BS170.
I
would be interested in hearing of peple trying these
on their 12 volt units.
Tony
AD0VC
I
have only had my HB for about a month and I have
blown a couple of PAs due to matching issues. That
aside, this QDX is FANTASTIC! The design is
remarkable and well executed. I love the electronic
aspects of the hobby so I always end up fiddling
with things but I would be proud of this design if
it were mine. Just amazing what was done for the
price. This thing makes me happy! I have ordered
another for the low bands.
Tony
AD0VC
Dave et Al,
I have used the first prototype High Band since
I designed the filters last year. I used at the
home QTH and then took it to QRP Labs HQ for
assessment and ultimate implementation. I have
used that and another HB almost continuously. I¡¯ve
operated with it portable in Prague, Rome and
Istanbul with various wire antennas out hotel
windows on fishing rods and wires on lead sinkers
slingshotted into local trees. I used to use a
small 12V PS with a variable buck down and meters
but now, due to noise reduction measures, I run
all my QDXes and RaspberryPIs on Lithiums, 2 in
series that start at 8.4V and I usually change
them and recharge when somewhere at 7.3V. RasPis I
use a 30,000mAH power bank.?
The one thing I am very careful about is
antenna matching. I used to use LC matches with
plastic printed spherical variometers as the
variable L. I now use FRI-matches with 3D printed
large toroids. I also have a switch that puts a
LED swr bridge in line and a switched LED current
monitor antenna wire.?
The first and original HB QDX is still going
strong! It was built for 9V but usually operates
at an average of 7.8V.?
Careful management of the antenna match is the
most critical to long QDX life!
73
Ross
6
?
Hi Dave, all
The washer isn't the heatsink.
The heatsink is the PCB. The washer and bolt
are just the mechanical means of bolting the
flat side of the transistors to the PCB.
As others have said... Thermal
paste doesn't harm but doesn't make a lot of
difference, either. Be careful too because
some pastes are electrically conductive and
that won't go well. Make sure you don't use
that.
Personally I have found the
QDX PA reliable and durable as have many
others. A lot of failures have posthumously
been attributed to sustained chronic SWR
mismatch or acute mismatch during auto tuner
use, supply spikes etc., over-voltage,
including from fully charged 12V-nominal
batteries.
Remember also that on any
forum such as this one, you will tend to
hear far more frequently from people with a
problem, than those without. You tend to get
the mis-impression consisting of the
proportion of people with problems, out of
the collection of people with problems. The
actual proportion of people with problems,
out of the total of all users, is a lot lot
lower (and difficult to quantify).
73 Hans G0UPL
On Apr 29, 2023, at 8:45 AM, Dave
Sergeant G3YMC < dave@...>
wrote:
Humour aside Razvan does have some valid points. I have not yet put my
QDX into transmit mode but already have concerns about the design of
its PA stage. The requirements for not exceeding 12V (in a 12V HB
version) or daring to transmit into anything other than the religious
50 ohm load is an indication that the design is maybe just a tad close
to the limits for comfort. It almost made me laugh when I attached the
little washer to the rounded side of the BF170s asking myself if I had
mounted them upside down. Surely you should put the heatsink on their
flat side I asked.
In the desire to make the thing as small and cheap as possible it seems
common sense has been overlooked. A couple of TO220 case devices on a
slighly bigger board and a bit more metal to heatsink them would have
made a far more robust PA and removed the need for all these
discussions on this reflector.
No criticism of Hans of course, the QDX is an incredible little rig,
but as an engineer in my professional life I would have not done it
that way.
73 Dave G3YMC
On 28 Apr 2023 at 15:15, DL2ARL wrote:
you could also sprinkle over the BS170s some Water out of the Holy
Fountain of Lourdes from
southwestern France. It would help as much as thermal paste, but it
will have the big advantage of
not doing no harm and cost less.
|
Robin,
Thank you for the linked material! It is a good addition to my library.
I'm very much aware of the current-fed operation of the original design and of the role of L14 as a current source. In the world of switch-mode power conversion this topology is known as a Current Fed Push Pull Converter. There is quite a bit of literature on it.
The problem for QDX is that the inductor L14 stores a good bit of energy and releases it all in a vicious LdI/dt spike at the end of transmission. No accomodation for that exists in the existing PA circuit, and so the PA transistors are at risk.
Earlier, Hans , working on a low 10m power issue, had experimented with the use of a 100 nF bypass capacitor at the center tap of L14'S primary, and reported good results with it. Several other users have adopted it as well.
Once that capacitor is deployed, L14 is rendered relatively harmless, but also completely moot. It's position could well be used for something else.
Regarding the output network on QDX, it is a classic two-stage Pi LPF. Is that more like a series LC or a parallel LC? Regardless, it seems to do its job and the amplifier delivers power whether it is operating in current fed or voltage fed mode.
The problem here is that even a purely resistive load that is less than 50 ohms is, in either operating mode, returned to the transistors as a low impedance and drain current elevates. It is not limited by L14. It doesn't take much to raise drain current to dangerous levels. A series resistance in place of L14 is intended to address that.
Regards! JZ KJ4A?
On Mon, May 15, 2023, 9:03 AM G8DQX list < list@...> wrote:
John,
it may not be obvious, but substituting a 2 ¦¸ resistor for L14 changes
the principle of operation of the Class D PA. As
originally designed, it is a current switching PA, and
requires a parallel-tuned output network. As modified it has
become a voltage switching PA, and requires a series-tuned
output network. Since the original QDX PA design was for a current
switching PA, this may explain why some have observed the output
filter getting hot.
These two classes are noted in the RSGB Handbook (12th edition,
p10.6) and Device Evaluation for Current-Mode Class-D RF Power
Amplifiers to be found at
at figures 3.4 and 3.5. [There are also many IEEE papers on class
D, but these seem to be behind the usual IEEE paywall.]
L14 is employed to act as a current source. Thus when the output
filter sees 50 ¦¸ resistive, a similar load will appear at L14, but
when things are reactive, there is the potential for undesirable
voltages at the PA MOSFETs. (A simple measure of SWR says nothing
as to how reactive the load might be.)
All that is not to say that a re-designed PA stage might not be
more resilient against non-50 ¦¸ loads, but it is not a simple
change. The original design has a great deal going for it. (And
even valve (tube) PAs can flash-over with too much SWR.)
73, Stay Safe,
Robin, G8DQX
On 12/05/2023 12:12, John Z wrote:
One more thing on this. Tony AC9QY and I concluded
that using a 100 nF capacitor to bypass the center tap of T1's
primary to ground, combined with replacing L14 with a 2 ohm
resistance, was a good thing.
The inductor L14 adds no value in this circuit,
and even contributes a dangerous LdI/dT spike at the end of
transmission. The 2 ohm resistor in its place offers some
additional protection against low-Z high SWR, which is the
high current case. The PA self-throttles in response to
excessive current. The loss of power under a 50 ohm load is
insignificant.
The combination of high transistor breakdown
voltage and high current protection makes the PA nearly
bulletproof regarding SWR. Simulation demonstrates that
nicely.
JZ KJ4A?
On Fri, May 12, 2023, 6:51 AM
John Zbrozek < jdzbrozek@...>
wrote:
Hello Tony,
Another Tony. AC9QY, and I have been running
LTSpice simulations of the QDX PA using TN0110
transistors.
They are similar to the Microchip TN0106
transistors that you have been testing,? differing mainly
in that they have a 100 volt drain breakdown specification
vs. 60 volts.
The simulation results look extremely
encouraging, even? on 10m, but we had some questions about
the adequacy of the real drive level available to the
gates.
Learning of your actual results in real
hardware is fabulous!
Tip of the hat for trying that! Bravo!
If you are an LTSpice user and interested in
running the simulation I would be delighted to send you
the .asc file for it.
73, JZ KJ4A?
I
have been experimenting with different transistors for
the PA in my HB and LB QDXs. We have been having heavy
rains the past two days and my antenna is providing
some bad mismatches of up to 2.8 to 1. I used this
condition to test several transistors for robustness
and have good results to report. The high SWR
conitions I used were caused by load impedances less
than 50 ohms as measured at the transmitter end of the
coax with my nanovna.
First
off, the BS170s do not tolerate these kinds of
mismatches and blow quickly. But I have been unable to
blow up either the TN0106 or VN0606 transistors. Note
that these two both have reverse pinout with respect
to the BS170 so the flat face has to be mounted
upwards such that it is pressed against the washer
instead of the ground plane. This does not seem to
bother the transistors and they run cool with a normal
50 ohm load. My units are wired for 9 volts and that
is how I did my testing. I operated the VN0606 for an
hour at 2.5 to 1 SWR. While the unit gets warm, it
does not fail. The TN0106 behaves similarly but I seem
to get more drive into the low Z loads and the
transistor heats up a bit more. AT 50 ohms, the two
transistors behave similarly.?
Neither
QDX has any other modifications from stock. No diodes
or capacitors are added. Just the transistors have
been changed.
The
transistors were ordered from Mouser.
I
also tried out the 2N6660. This is a higher power
device in a TO-39 can. While it works, it provides
much lower output, even into 50 ohms (on the order of
1 watt). I tried using just two devices but that made
even lower power. These are expensive and the QDX does
not provode enough drive for them.
The
10 meter performance of both the TN0106 and VN0606
seems to be better than the BS170.
I
would be interested in hearing of peple trying these
on their 12 volt units.
Tony
AD0VC
I
have only had my HB for about a month and I have
blown a couple of PAs due to matching issues. That
aside, this QDX is FANTASTIC! The design is
remarkable and well executed. I love the electronic
aspects of the hobby so I always end up fiddling
with things but I would be proud of this design if
it were mine. Just amazing what was done for the
price. This thing makes me happy! I have ordered
another for the low bands.
Tony
AD0VC
Dave et Al,
I have used the first prototype High Band since
I designed the filters last year. I used at the
home QTH and then took it to QRP Labs HQ for
assessment and ultimate implementation. I have
used that and another HB almost continuously. I¡¯ve
operated with it portable in Prague, Rome and
Istanbul with various wire antennas out hotel
windows on fishing rods and wires on lead sinkers
slingshotted into local trees. I used to use a
small 12V PS with a variable buck down and meters
but now, due to noise reduction measures, I run
all my QDXes and RaspberryPIs on Lithiums, 2 in
series that start at 8.4V and I usually change
them and recharge when somewhere at 7.3V. RasPis I
use a 30,000mAH power bank.?
The one thing I am very careful about is
antenna matching. I used to use LC matches with
plastic printed spherical variometers as the
variable L. I now use FRI-matches with 3D printed
large toroids. I also have a switch that puts a
LED swr bridge in line and a switched LED current
monitor antenna wire.?
The first and original HB QDX is still going
strong! It was built for 9V but usually operates
at an average of 7.8V.?
Careful management of the antenna match is the
most critical to long QDX life!
73
Ross
6
?
Hi Dave, all
The washer isn't the heatsink.
The heatsink is the PCB. The washer and bolt
are just the mechanical means of bolting the
flat side of the transistors to the PCB.
As others have said... Thermal
paste doesn't harm but doesn't make a lot of
difference, either. Be careful too because
some pastes are electrically conductive and
that won't go well. Make sure you don't use
that.
Personally I have found the
QDX PA reliable and durable as have many
others. A lot of failures have posthumously
been attributed to sustained chronic SWR
mismatch or acute mismatch during auto tuner
use, supply spikes etc., over-voltage,
including from fully charged 12V-nominal
batteries.
Remember also that on any
forum such as this one, you will tend to
hear far more frequently from people with a
problem, than those without. You tend to get
the mis-impression consisting of the
proportion of people with problems, out of
the collection of people with problems. The
actual proportion of people with problems,
out of the total of all users, is a lot lot
lower (and difficult to quantify).
73 Hans G0UPL
On Apr 29, 2023, at 8:45 AM, Dave
Sergeant G3YMC < dave@...>
wrote:
Humour aside Razvan does have some valid points. I have not yet put my
QDX into transmit mode but already have concerns about the design of
its PA stage. The requirements for not exceeding 12V (in a 12V HB
version) or daring to transmit into anything other than the religious
50 ohm load is an indication that the design is maybe just a tad close
to the limits for comfort. It almost made me laugh when I attached the
little washer to the rounded side of the BF170s asking myself if I had
mounted them upside down. Surely you should put the heatsink on their
flat side I asked.
In the desire to make the thing as small and cheap as possible it seems
common sense has been overlooked. A couple of TO220 case devices on a
slighly bigger board and a bit more metal to heatsink them would have
made a far more robust PA and removed the need for all these
discussions on this reflector.
No criticism of Hans of course, the QDX is an incredible little rig,
but as an engineer in my professional life I would have not done it
that way.
73 Dave G3YMC
On 28 Apr 2023 at 15:15, DL2ARL wrote:
you could also sprinkle over the BS170s some Water out of the Holy
Fountain of Lourdes from
southwestern France. It would help as much as thermal paste, but it
will have the big advantage of
not doing no harm and cost less.
|
Andy,
People who do not accept the fact that TO-92 transistors dissipate the heat through the leads maybe know the thermodynamic model of those transistors. Some heat is dissipated through leads, but mainly through one lead only: the lead with the bond pad. The chip is mounted with silver filled epoxy on the pad of this lead. The others are just connected to the chip with a thin gold bonding wire with low heat transfer capacity. Furthermore, a good part of the heat is dissipated through the thermoset plastic encapsulation. This material is designed with specifications for thermal conductivity for the purpose of conducting thermal energy away from the chip. If you touch an overloaded transistor with your fingers, you will quickly change your mind.
73, Jakob
|
Hello guys
Just want to summarize my experiments with T1 transformer improvement. Please note all was tested with modified PA: 2Ohm resistor in place of L14 and 100nF+10nF capacitor from transformer tap to ground. 80-20m QDX. Dedicated 9V 1.5A power supply (gives 9.15V). I already blown my PA in the past so the transistors are not stock one but good quality.
My goal was to have as much output power as possible without excessing 1A PA supply current (leaving some safety margin would be good).
I have tested the following:
1. Dual core transformer (as in mcHF transceiver modification) - power drops significantly, especially on higher bands. I remember it highly improved the mcHF output but here it seems to have too much to push with little BS170.
2. Different number of turns: 2/2, 4/4, 3/4, 2/3. Some combinations gives improvement on some bands but usually makes other bands worse. Seems 3/3 is well optimized.
3. Use different wires: original kit wire is described as 0.6mm. I have measured 0.6mm but together with enamel. I tried 0.6 twisted wire, 8x0.08mm litz wire, 16x0.08mm litz wire, insulated teflon wire, single wire from CAT6 network cable, 50Ohm coaxial cable and 0.8mm enamel wire (measured without enamel). The last option gives about 0.2-0.3W more output power with the same PA supply current. Other options work as the original winding, worse, or improve some bands when ruin others.
It is really hard to fit 0.8mm wire on the core but is doable. It will need also some thinner wire at the end of the windings to solder transformer to the board.
The 0.8mm T1 gives me 3.5W-3.8W on all bands. The PA current was around 0.8A. I checked also replace 2Ohm resistor with 1.5Ohm, what results in 3.8-4.1W output with 0.92A PA supply current.
Looking from this perspective the design is really optimized, I bet Hans spent a lot of time on simulations and tests. If point-something Watt is in your opinion worth it then feel free. I am not sure how the bigger winding diameter will affect not modified PA - it may happen the current will grow too much and blow PA transistors. Waiting for the QMX manual to check if Hans has maybe figured out something robust :).
If you need single band only radio then QDX-M is something for you. I checked that you loose 0.1-0.2W only on the switching diodes.
From other perspective change from 3W to 4W is only 1.25dB, and 4W to 5W is less than one dB. You can sometimes gain more using better quality BNC-UHF adapter...
If I will find more time I will test thicker wire on the LPF or maybe adding some silver plated wire to the PA output tracks or look closer on the ground connection between PA and BNC...
Cheers
Pawe? SO8FM.
|
PE2V (Vincent)