H-Lv2b2 PA mesurements

[in3otd]

Hello,

I've done some measurements on the H-L v2b2 PA output;

here is the PA output vs. the TxPGA gain (controlled using the Spot value in Quisk):

results are quite similar to the stand-alone PA measured here. As known, the full driver output is not needed to have 37 dBm (5 W) out, so the PA is overdriven when using the maximum gain settings.

Next graph is the PA efficiency, same conditions as above:

Last graph is the fundamental and first few harmonics output for TxPGA gain setting of 8, which gives (more than) 5 W out up to 30 MHz:

All the above data were taken with the PA output transformer center tap connected to the on-board 9.4 V supply via a current meter, so the actual results for the usual case when the center tap is soldered directly to the power supply may be slightly different.
I need to check how well output transformer works at low frequency, I recall it got more than warm at full power at 1.8 MHz on the test PA.

73 de Claudio, IN3OTD / DK1CG

[Graeme Jury]

Hello Claudio,

Many thanks for the hugely useful graphs and information. I am interested to know which version of the PA you are using for the tests i.e. which transistors etc. I note that the amplifier will be run at more than +37 in the final system and probably at around +38 dBm depending on the filter board losses. In my own case my filter board is going to be the same size as the full size V2 HL board (the size with the tabs on) and will have the Rx and Tx filters on the board with M7 diodes switching the Tx and PE4259's switching the Rx filters so the losses will be higher than with relays but your graphs show there will be no problem with power and Quisk allows equalization by band as I expect other radios will too so I am unlikely to need to go higher than step 12 even at 30 MHz.

I really appreciate the work you are doing in this area and your whole web site data is a goldmine.

73, Graeme ZL2APV

[John Williams]

Graeme,

Will be anxious to follow your progress...

John

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Steve Haynal]

Hi Claudio and Group,

Thanks as always for your measurements. This project really benefits from your detailed work.

How hot were the AFT devices when you were running around 10W output? Do you think heat can be properly dissipated at that level? Are the harmonics relatively worse at 10W versus 5W?

I've been experimenting with various tweaks to the PA and driver. First, I've been using separate power supplies for the driver and the PA. U12 has been powering the driver after cutting the trace to pin 8 and connecting that to DB10 pin 1. R91=11.5K and R102=1.6K produces 10.15V. U8 is powering the PA with R19=16.2K and R46=1.8K for 7.9V. Second, I've reduced the gain of the driver by setting R55 to 120 Ohms. I see max output at the low power output of ~17dBm and the PA never goes higher than 39dBm with full TxDAC drive.

I'm considering on finalizing on one of these options for an assembled run. I'd appreciate your input on what you think is best.

Option 1:

* Include U8 for driver and PA at 9V tor 9.44V. 

* Add option to disconnect U8 from PA and use the fused 13.8V supply instead for PAs using TO-220 devices.

* U12 would never be used and may be removed.

* Since U8 is populated, less for a builder to do to add AFT05-based PA

* Higher voltage (>7.5V) for AFT05 devices may produce less harmonics but greater danger of running too hot, dissipating too much power

* Lower voltage of driver may mean worse harmonics, but my tests show this not to be significant, especially with R55=120 Ohms

* Higher voltage (>7.5V) for AFT05 devices means poorer efficiency

Option 2:

* Use U12 always for driver at 10V

* Builder adds U8 and associated parts if PA is to be AFT05-based, more for the builder to do 

* U8 can be run at lower voltage between 7.5 and 8V to provide safer conditions for AFT05-based PA

* Lower voltage for PA means better efficiency, less heat at PA

* Lower voltage for PA means potentially worse harmonics, IMD, but my tests show this is not significant

* Higher voltage driver may mean cleaner signal

* Less load on U8 which is running hot during TX for me

* Lower voltage for U8 means worse efficiency at power supply, but better efficiency at PA should be overall win

Option 3:

Option 1 or 2 but with AFT05-based PA stuffed on the assembled board. This would essentially eliminate the TO-220 option (it would require removing parts or doing a special build). This would be the least work for a builder as they would only have to install T3 and use a proper enclosure for thermal dissipation to enable TX. I have been very pleased with the AFT05 devices so far.

Option 4:

* Lowest cost build, this would only be good for RX and VNA use

* Least risk for assembly house doing something wrong

* U8, U12 and associated parts not stuffed. PA builders would have to add appropriate power supply

* Driver not stuffed, instead IAMP output can be jumpered from DB6 to DB21,DB25 for VNA use. Builder must install jumper and T1 for VNA use

* Versa Clock not stuffed, instead use 38.4 oscillator directly with AD9866 at X3. Phase noise may be slightly worse. Builder desiring more complicated/cleaner clocking would have to remove X3 and add Versa Clock.

* Slow ADC and digital POT, not stuffed but instead added by builder if needed

* Still one firmware, various configuration selected by jumpers CN8, CN9 and CN10

73,

Steve

KF7O

[in3otd]

Hello Steve,

How hot were the AFT devices when you were running around 10W output?

I do not really know. Since I was expecting that during the test the devices would have to dissipate quite some power I used a very high flow fan directed to a small metal plate soldered on the PCB border, near the finals. The board was almost blown away by the airflow, hi.
 

Do you think heat can be properly dissipated at that level?

Maybe. From the graph below you can see that for a TxPGA setting of 8 (the one giving more than 5 W up to 30 MHz), the peak overall PA power dissipation is slightly above 7.5 W, so 3.25 W per device. It's only slightly higher when overdriving the PA.

The datasheet says that the device thermal resistance, junction to case, is 4.1 °C/W so if we were able to keep the devices case at room temperature  the junctions temperature will rise just by 13 degrees. Equivalently, we can say that if we allow a junction temperature increase of 100 °C (a bit extreme, but the junction temperature can go up to 150 °C, with a lower MTBF) the overall thermal resistance can be almost 31 °C/W, leaving more than 26 °C/W for the case-to-ambient part. Of course this will leave no margin for increased power dissipation due to bad SWR and does not take into account the other heat sources in the H-Lv2 but we should be able to reach quite less than 26 °C/W of case-to-ambient thermal resistance. "How?" is the part that needs to be figured out, hi. I'll open another thread for the H-Lv2 thermal management.
 

Are the harmonics relatively worse at 10W versus 5W?

Yes, but not too much, see graph below for the maximum TxPGA setting:
 

I've been experimenting with various tweaks to the PA and driver. First, I've been using separate power supplies for the driver and the PA. U12 has been powering the driver after cutting the trace to pin 8 and connecting that to DB10 pin 1. R91=11.5K and R102=1.6K produces 10.15V. U8 is powering the PA with R19=16.2K and R46=1.8K for 7.9V. Second, I've reduced the gain of the driver by setting R55 to 120 Ohms. I see max output at the low power output of ~17dBm and the PA never goes higher than 39dBm with full TxDAC drive.

I'm considering on finalizing on one of these options for an assembled run. I'd appreciate your input on what you think is best.

[snip]
As usual, the best option depends of the definition of "best"... Option 4 if you'd like to go for the lowest cost; Option 3 IMHO is the best overall since it will suit both people wanting an (almost) ready made RTX and those wanting to experiment a bit and fit the RD15 HVF1 PA. Removing SMD components is easier than solder them, hi, and even if some traces/pads are lifted around the on-board PA while reworking it, it should be not too difficult to repair as all the connections are on the outer layers.
I'm not sure I agree with all the points listed for Option 1 and 2 regarding efficiency and power dissipation, need to think more about that.

[Steve Haynal]

Hi Claudio,

I doubt if we will ever have a good and exact analysis of the PA thermal dissipation for an opensource project of this small scope. I think we will just have to estimate. The thermal dissipation configuration in the build notes has worked well for me. Just 3 screws and a small copper strip. The copper strip may not even be doing much given the limited area of the transistor tab it touches and the relative roughness of both surfaces. I will try to do some more stress testing.

I look forward to hearing your thoughts on overall efficiency and thermal dissipation versus PA power supply levels.

73,

Steve

KF7O

[James Ahlstrom]

Hello Steve and Group,

I am back from some great skiing in Utah, USA!  And now it is snowing heavily at home in New Jersey.

On Monday, March 13, 2017 at 1:04:18 AM UTC-4, Steve Haynal wrote:

I'm considering on finalizing on one of these options for an assembled run. I'd appreciate your input on what you think is best.

Option 1:

 

The options vary in whether U8 and U12 are included.  Let's look at the costs.  Prices are Digikey small quantity and I didn't add a cost for assembly.  So prices are rough but indicative.

U8 is the 9 volt power supply for the AFT05 finals.  U8 costs $1.98 plus 11 small parts for a total of $6.71.

U12 is the 10 volt supply for the driver.  U12 costs $0.45 plus 4 small parts for a total of $0.97.

My thinking is that U12 should always be included because it is cheap.  This eliminates J20.  It requires a board change to connect Vin (pin 8) to the input supply.  Now that I am looking at it, J28 should never be installed because it connects VPA (9 volts) to FPGA pin 42.  This change eliminates all U12 build options.

U8 is more of a problem because it is a bit expensive.  It is required for the AFT05 finals but not for the TO-220 devices.  But even for TO-220 devices, it is a spare high current supply for experiments, and is a useful feature of HL2.   So my thinking is that U8 should always be included too.  But that also requires a board change.  If U8 is mounted, I am not sure we can jumper J6 to provide 12 volts even if the enable (pin 2) is false.  And there is no way to provide 12 volts to the finals and still have 9 volts for other purposes.  We would need to add more jumpers to choose the 9 or 12 volt supply, and another jumper for the enable pin.  This should not be difficult, and enables the user to choose AFT05 or TO-220 parts easily.  I don't believe there are many hams who are comfortable with adding or removing 0603 parts.

I am still thinking about the options 1 to 4 and will post again in a day or two.

Jim

N2ADR

[Steve Haynal]

Hi Jim,

Welcome back! I calculate $4.92 from the BOM for the 7.5 to 9 V power supply, not $6.71. $0.02 per SMT pin is a good rough estimate for assembly cost.

 

J28 is there as U12 pin 5 can support the high voltage, and FB15 is used to disconnect any high voltage from the FPGA. This is in place to be able to use the FPGA pin as a spare by connecting a wire to the FPGA-side pin of the FB15 footprint.

I also have thought about jumpering the output of U8 so that it can be used independently from the PA. I need to think more about this.

0603 parts are difficult to work with. I plan to try and make any part that a user might change 0805.

73,

Steve

KF7O

[Steve Haynal]

Hi,

Two more thoughts on this. Many of the HL2 parts were chosen because they are inexpensive and readily available in China. For example, the ST1S10 is typically in the $0.50 to $0.70 range on www.aliexpress.com even in small quantities. Also, Claudio pointed out that if we don't need to synchronize the switches (he saw very little power supply noise) then we can use the ST1S40. This is a little bit less expensive from US distributors but I can't find it on www.aliexpress.com.

73,

Steve

KF7O

[James Ahlstrom]

Hi Steve,

On Monday, March 13, 2017 at 1:04:18 AM UTC-4, Steve Haynal wrote:

I'm considering on finalizing on one of these options for an assembled run. I'd appreciate your input on what you think is best.

The best option depends on the target market.  "Market" may be an odd word for a free project, but I am a businessman and that is how I think.  In my mind the market is the same as the SoftRock market.  In particular, the market will not assemble any SMD parts, not even 1206 or SOIC.  It will demand an almost completed unit that can be put on the air as a QRP transceiver in short order.  The market can do simple mechanical construction, solder TO-220 pins, solder wires in through-hole pads and add jumpers to pin headers.  I have a prejudice for simplicity, as I have found that it is easier to make things simple rather than have to endure endless customer support issues.

Supporting both AFT05 and TO-220 finals is troublesome.  I presume the drive requirements are different, so the driver gain will have to be changed by replacing an 0603 resistor.  The power must be jumpered properly if both are supported.  Steve is reporting good progress with heat sinking the AFT05 devices.  If the heat sinking proves successful I vote for option 3, an AFT05 build with U8 (power amp supply) installed.  U12 (driver supply, 10 volts or so) is optional but cheap, and including it may add flexibility.

If the heat sink issue can not be solved, I vote for option 3, but with TO-220 finals.  This also applies if the AFT05s can not survive high SWR, for example, from an unconnected antenna.

This is a good time to think about battery operation.  We want to operate with nearly discharged batteries at a minimum voltage of Vmin, and U8 needs 15% headroom.  If we need 8.5 volts for the AFT05s, then Vmin is 8.5/0.85, or 10.0 volts.  Then U12 must be 10.0 - 0.5, or 9.5 volts for 500 mV dropout.  Then both U8 and U12 go out of regulation at the same voltage.

There should still be enough flexibility so that more advanced users can solder and unsolder jumpers for different finals, transverters or VNA use.  The finals can be turned off by setting their bias to zero, but we will need a jumper to set the U8 enable pin to false.

At this point HL2 is not a stand-alone QRP transceiver because it lacks Tx and Rx filters.  I believe that Graeme is working on an all-band filter.  I think we should also offer a simpler 2 or 3 band option similar to SoftRock.  My box is 15 cm long, so it is possible to make a 5x10 cm two-sided board to slide into the same slot as the HL2.  That should be enough for a few bands.  We should try to standardize on an enclosure so people can start designing companion boards.

Jim

N2ADR

[James Ahlstrom]

Hello Steve and Group,

On Monday, March 13, 2017 at 1:04:18 AM UTC-4, Steve Haynal wrote:

Second, I've reduced the gain of the driver by setting R55 to 120 Ohms. I see max output at the low power output of ~17dBm and the PA never goes higher than 39dBm with full TxDAC drive.

Now that my HL2 is working, I can make some measurements.  With R55 at the original 75 ohms and a signal at 7.2 MHz, it is necessary to  reduce the Tx Level (0 to 255) to 100 to achieve a 5 watt output level.  But if the user uses a Tx Level of 255, the finals are severely over driven.  So I changed R55 to 120 ohms, the same as Steve.  Here are the Tx Levels needed to achieve 5 watts:

40 meters   220

15               230

10               255

Once filters are installed, we may fall short of 5 watts on higher bands.

Here are the driver voltage gains at several values of R55:

R55              Gain         Gain dB

  75                 9.8                 0.0

100                 7.6                -2.2

110                 7.0                -2.9

120                 6.5                -3.6

I don't want users to over drive the finals, but maybe we should use 100 or 110 ohms and then reduce the power by specifying a Tx Level lower than 255.

On a different note, the peak-to-peak drive at each AFT05 gate is 2.3 Vpp.  The driver needs about 3 volts of headroom before clipping, so the driver can operate from a 5.3 volt supply or higher.  The driver works better at higher voltage, and at the moment I don't see a reason to go below a 9 or 10 volt supply.  But we could if we need to.

Jim

N2ADR

[Graeme Jury]

Hello All,

If the goal is to get 5 watts out of the HL2 then it should include the filters also. For a straight relay switched LP filter the loss would be typically .3 dBm and the 120 ohm resistor would be fine. If however solid state filter switching and antenna changeover is utilised we would be looking at a total loss of typically 1 dBm and would require 6.3 watts out to meet this so the 100 ohm resistor would be a good choice in this case with a little headroom as well.

Using Quisk with my HiQSDR I am using Tx Levels of around 120 to drive my amplifier to 100 watts and consider setting the level which does not cause spurs to be a mandatory setup requirement although I appreciate some users will not see it that way and go for as much power as they can get and we need to protect the community from them by limiting the output to the linear capability of the finals.

There seems to be a trend to using the AFT05's but it seems to me that many of the problems they suffer like getting rid of heat and running right at maximum power would simply go away with TO220 fets. I am wondering if the AFT's offer some big advantage I don't see?

73, Graeme zl2apv

[James Ahlstrom]

Hello Group,

I am measuring the current in the AFT05 finals by cutting their power trace and inserting an 0.1 ohm current sense resistor.  While setting the bias I noticed that the drain current started to increase at a bias setting (0 to 255) of 200.  That makes the adjustment a bit coarse since only 55 settings remain.  The change from setting 231 to 232 changed bias from 95 to 104 ma.  Perhaps this is not important.  But maybe we should set the voltage at U15 pins P0A and P1A to 1.8 volts (or a bit more) with a resistor voltage divider.  The minimum gate threshold voltage is 1.8 volts, and the max is 2.6 volts.

I am not sure where we stand on biasing TO-220 devices, or if we still intend to support them.  The gate threshold of the RD06HHF1 is 1.9 volts (good) but the max is 4.9 volts (bad) and so we need more voltage.

Jim

N2ADR

[in3otd]

Hello,
some solutions for this were discussed in another thread, see e.g. here. Having finer steps for the bias is certainly a good thing, even if maybe not so important for the RF performances; at least we will use the full potentiometers range and the users would not wonder why nothing happens over most of the range, hi.

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Graeme,
I agree it would be nice to have 5 W after the TX filters; IIRC we didn't find a cheap diode working well as TX switch down to 1.8 MHz and I tend to think that relays may be more suited here.

I think that the major advantages of the AFT05MS003N devices are the low cost and the fact that they are SMDs so they can be assembled on the board with the other components. The heat dissipation seems not to be a major problem at present, at least for SSB usage.
The PA with the RD15HVF1 I tried (see unfinished page here) could provide some more power but the gain vs. frequency variation was quite higher.

73 de Claudio, IN3OTD / DK1CG

[Graeme Jury]

Hello Claudio,

The diodes were a bit disappointing at 1.8 MHz except for the M7 which may be brand dependent and the ones I have here work quite well. I am mindful that some may want to work the LF bands and would need to be switching through relays for that although probably would have a separate filter board for that purpose.

Another issue somewhat loosely related to power output and measurements is should a changeover relay be incorporated onto the filter board or should the Rx and Tx ports be brought out on their own sma connectors together with the changeover relay signal? If the input of any following linear is separate Tx and Rx and output from HL was single connector from a changeover relay then a relay would need to be added to switch the Linear inputs and of course if it was a single Rx/Tx input Linear with HL being separate Tx and Rx we have the same issue. None of these are show stoppers but need to be considered as well as on board strapping options together with the inclusion of the relay or not.

I have built a few amplifiers in the 5 watt bracket using SOT-89 packages and getting rid of heat was always a headache for me. I finally got one going that did not thermally destruct by soldering a strip of copper foil as used for repousse work and dressing it to the side of the case and clamping it with some 4mm thick aluminium bar - hence my hesitancy but I am encouraged by the great results you guys are getting and am not so fearful now. Your reasons for the AFT05MS003N's are valid and really good ones.

73, Graeme zl2apv

[in3otd]

Hello Jim,
regarding the TX output variation with frequency, currently there is about 1 dB of ripple in the output level due to the filter at the TxDAC output; would it make sense to (try to) modify the filter to have less ripple or even a flat response in the passband?. IIRC, the main driver for the current filter response was to reduce the image at fs-ftx when transmitting in the 10 m band, now that fs is a little higher we could maybe afford a filter with slightly less rejection there.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Claudio,

On Tuesday, March 21, 2017 at 5:32:54 PM UTC-4, in3otd wrote:

Hello Jim,
regarding the TX output variation with frequency, currently there is about 1 dB of ripple in the output level due to the filter at the TxDAC output; would it make sense to (try to) modify the filter to have less ripple or even a flat response in the passband?

I don't think the TX output variation is a problem.  The issue is output flatness after the finals and the Tx low-pass filters, not at the output of the TxDAC.  I have never built an amp that was flat at the antenna connector.  Setting a suitable Tx Level for each band equalizes the output in software and that is the solution.

A slight issue remains in that the Tx Level has 0.5 dB steps, so we can only adjust to within 0.25 dB.  I don't think that is a problem either.  But we could use Tx Level as a coarse control and adjust the Tx signal level to make up the difference.

Jim

N2ADR

[Steve Haynal]

Hi Graeme,

I favor the slightly lower resistor to provide some headroom. 

I don't think dissipating heat is a problem with the AFT05s. Also, we went through a lot of effort to get the drive level "right" for John's amplifier. What we are doing now with the AFT05s is a similar process and I think we will settle on a good configuration. Claudio has reported that even when driven to close to 10W, the AFT05's distortion did not increase too much.

The advantages I see for the AFT05s are:

* They are available from Mouser or Digikey. The Mitsubishi devices require adding another supplier.

* Even with the additional power supply, the AFT05s devices are less expensive than a 2 transistor Mitsubishi solution. Given the price of the power supply is significantly less in China, the savings are greater.

* They are surface mount devices. This means there is less special assembly required.

* Assuming the PCB/Enclosure slot thermal dissipation is enough, the final assembly (1 screw vs 2 bolts on the side) is simpler.

* This pushes thinking and experimentation into some new territory. I'm a bit tired of all the Mitsubishi-based amplifiers out there.

Regarding an earlier post proposing an Arduino on the filter board, this will require an additional programming step and increase costs. I have used Atmel microcontrollers in other projects. Is 16 bits of output from an I2C bus expander not enough? Software will have a general way to program each bit.

73,

Steve

KF7O

[Steve Haynal]

Hi Jim and Claudio,

Yes, the details Claudio worked out in the previous post he linked to is the direction I'd like to go. This will also allow us to have smaller steps and better resolution from the digital pot.

73,

Steve

KF7O

[Steve Haynal]

Hi Graeme,

My current thinking is for the changeover relay to be on the main board and the filters shared for RX/TX on the 5cm extension board. I'm still leaning towards the series arrangement discussed several months ago, with (1 of 3) HPF and a selected LPF inline. If we use SMT inductors for the higher frequencies, I think we should have enough room.

73,

Steve

KF7O

[James Ahlstrom]

Hello Group,

I am looking at bandpass filters instead of a highpass and lowpass combination.  The bandpass solution is a simple 3-pole filter for each band.  I have never seen bandpass filters used this way, but I plan to give it a try.  It may be OK at 5 watts.

Jim

N2ADR

[John Williams]

My Superband design used bandpass. See my PA github clone for the design and values. It was derived from the Peaberry design.

John W9JSW

--

[James Ahlstrom]

Hello John,

Very interesting!  I see you are using 4-pole filters, and have combined pairs of bands.  With bandpass filters, we can install the HL2 T/R relay K2 as usual, and just switch in a filter between the HL2 connector pads for RF2 and an external antenna connector.  This would be very simple.  It would protect both the Tx and Rx path.

Do you have any measurements for filter loss etc., or any experiences to share?  Is 4-pole necessary, or might 3-pole work?

Jim

N2ADR

[John Williams]

Yes, Jim. We have some runs. The issue we had was filter roll-off, mainly due to pcb layout (I was pretty new at it). I reworked the board a few times but never re-ran the filters as we moved on to the all band PA. At this point, I can share the plots with you but think they will confuse the issue, since they were evolutionary. I think it best for you to do a 2 layer simple layout to test them yourself. Should not cost very much to get boards from Elecrow to experiment with. The issue we had with the 2 band approach was due to the single FET PA and suppressing the second harmonic of the lower band. Rolloff was not very sharp. 60/40 and 30/20 were problematic. With a push-pull PA this will be much easier to overcome. We also had issues with the spurs on 12/10M but those have been now resolved.

Some Elsie data attached... If only we can get the real data to agree with elsie!

John

[James Ahlstrom]

Hello John,

I will look at this.  The Elsie data looks great, but I can hardly ever get a filter to agree with Elsie either.

Jim

N2ADR

[Graeme Jury]

Hello Jim,

There was a huge amount of work done last year by John, Glenn VK3PE and myself on bandpass filters. As John has pointed out we were working with the constraint of a single ended PA and had difficult requirements to meet for second harmonic suppression. The mesh filters we designed worked well and it may save you a lot of time to review some of the posts we made as we published some performance data.

The bandpass filters will work well and are easy to design with stable and reproducible results with standard components. The reason for moving away from these is to give the flexibility of running a second receiver which may be octaves away and separate hi pass lo pass filters will allow this to occur. It will require a rethink on filter switching from the Quisk etc. perspective as the bandwidth will need a low end and a high end signal and if there are 3 or 4 HP filters they will require a band button each and interlock function so the HP can't be higher than the LP.

As a review perhaps you might like to look at some of the following posts ...

Lo Pass 7 MHz filters - Last few posts most relevant and can be viewed here
30/20 mesh filters - First 10 or so posts and last post and can be viewed here
A collation of mesh filters from Glenn. This is a very important work and should be a big help to you. View it here

An original Elsie design for a 17/15 M mesh filter to test how actual vs Elsie design pans out
The schematic can be seen here and the Elsie plot is here
17M/15M mesh filter as swept by Glenn VK3PE seen here As you can see a pretty good match.

The bandpass filters worked quite well with smd inductors and paid a price of about .5 to 1 dB insertion loss in band only. Out of band the performance was similar to toroids but with 5 watts TX I would be strongly leaning to T37's. We did find that the design was not too critical and winding the toroids to the stated number of turns gave a quite close result so there was really only the tedium of winding them and no requirement for a well equipped lab to adjust after after building the filter.

I still lean to separate HP/LP filters and my experiments suggest that LP needs 160, 80, 60/40, 30/20, 17/15. I would leave the 30 MHz LP filter in circuit as it is easy to design for low insertion loss and low SWR. The HP filters need 160, 80, 40, 30 17. Again I would leave 160 in circuit as a floor filter and switch the other 4 filters. These are the filters which work best for me but may not be the general case as I have very strong broadcast on both AM and 80 MHz FM giving volts on my antenna.

I believe it is important that we agree on what the filtering is so that it can be worked on in unison as John, Glenn and I did. Together we made progress as a team that we never would have done in a sensible time frame individually. As the board is an add on other versions beside the official one can grow although they will be constrained by the signals from the radio software. I don't really care what the official design is as I will work on it until it is completed and if it is not meeting my needs I will simply do my own version later.

73, Graeme zl2apv

[pascal.v...@gmail.com]

Qrp-labs offers BPF https://www.qrp-labs.com/bpfkit.html - I tried one but probably not well adjusted, I got a 3dB loss that didn't allow to use it to transmit.
Best 73

[James Ahlstrom]

Hello Pascal,

Thanks for the link.  The bandpass filters you referenced are coupled resonators and due to the high Q they have large losses.  We are considering Butterworth bandpass filters with larger passbands.  These may work.

Jim

N2ADR

[James Ahlstrom]

Hello Graeme,

Thank you for the very detailed references.  Now I will have some homework to do!  I am surprised that SMD inductors worked.  What size parts did you use?

I agree that there can be several filter boards.  I was only thinking of a simple three or so band version, like SoftRock.  It would have all through hole parts and be easy to build.  The builder would have to decide on what bands were needed.  Fancier versions are very welcome.

Jim

N2ADR

[Steve Haynal]

Hi Pascal,

I recently purchased two of these filters also. My understanding is they are only for RX, no TX. Jim is thinking in terms of TX also. But for RX, I purchased them to try and measure how much improvement if any on RX do we see with a BPF. From BPF filters discussed on this list, I've seen data showing that BPFs work as designed, but I'd like to see specific data on if and by how much BPFs improve DR2/IP2 (2nd -order IMD) for the Hermes-Lite. This data is important to me to help decide what type of filtering is best. BPFs limit skimming on multiple bands, but if the improvements to DR2/IP2 are significant enough, then some may be worth it. If Graeme, Glen, Jim and other filter builders could provide this data, that would be great too.

73,

Steve

KF7O

[Graeme Jury]

Hello Steve,

Thanks for the info on the AFT05s and yes I am convinced that they will do the job just fine now. The thinking behind the microcontroller is that it will do the job of the 2 I2C bus expander chips in a single chip. As I was considering using PE4259's on Rx and M7 diodes on Tx, the microcontroller would switch them directly and no relay driver like ULN2003 would be needed so nearly all the board real estate would be available for filtering. I was also looking to produce the auxilliary output for a following linear including the PA bias control and TX/RX relay switching signals and was going to program the appropriate delays in the micro. Don't worry too much about this filter board as it is my personal design (although I will of course share it) and I will put my initial effort into the filter design that the group agrees on. I also like Jim's idea of a beginners filter with 3 bands or so and through hole parts. I agree that as high density surface mount board would seem formidable to an inexperienced constructor.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

Just to be clear, the schematic lists either the MCP23017 or MCP23008. The MCP23017 is a single chip I2C bus expander to 16 bits. It sells for $1.24. What is the least expensive Atmel device with 16 or more outputs?

Another project I am working on has DigiKey preprogramming our Atmel microcontrollers. Programmin or reprogramming these devices in circuit can be difficult for some users.

Based on layouts I was experimenting with earlier, I think we can do 5 LPFs and 2 or 3 HPFs on the 10cm by 5cm expansion card. I would prefer that over a 3 band card.

73,

Steve

KF7O

[Graeme Jury]

Hello Jim,

Here is a typical smd inductor as used in John's board and tested by Glenn. I have run 10 watts through these and they did not fry but I would not use them myself in Tx service without a lot more testing for things like through loss, IMD, heat etc. They work fine in Rx service and the extra loss over toroids can't be heard in practice but can be seen in Tx service.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme and Jim,

There has been much good discussion and research in the past regarding SMD inductors suitable for 5W TX. The RS-HFIQ used SMD inductors. Please see this post for a complete list of the inductors used in the RS-HFIQ at 5W.

73,

Steve

KF7O

[Graeme Jury]

Ahh I missed that only one expander chip needed. Yes an atmega326 will cost $US1.40 incl shipping from China so it is a more expensive option. As the HL will provide the I2C clocking the internal oscillator will do the job as nothing else is time critical, but yes again you have made a point.

73, Graeme zl2apv

[Sid Boyce]

I am working with Steve Wilson (KA6S) on a project using an ODROID-C2
and the IO Pi Plus board (2x MCP23017's) base for pihpsdr used with HL
v1.44 and John's 5W PA.
John provided the OC config for the filter switching.

I currently only have the 600 PPR VFO encoder connected and working OK.
In a few days I plan to wire in the other encoders and the switches.

Steve is evaluating the new Asus Tinker Board which if as advertised has
the same GPIO configuration as the Pi 3, would be a drop in replacement.

The goal is a compact rig based on HL v2.0.
73 ... Sid.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[in3otd]

Hello Steve,

I initially thought that the PA power dissipation would not be so much different between a 7.5 V and a 9 V supply but then I did some measurements to have some actual data and indeed there is a big difference, as you suspected.

I cut the VPA supply trace and used an external power supply to provide different voltages; I run some tests at 9 V and 7.5 V;

here is the power dissipation at 9 V vs. the TxPGA setting (similar to what posted before but not identical, see below)

and here is the power dissipation at 7.5 V, still vs. the TxPGA setting

For a power output around 37 dBm, a 7.5 V supply gives about 2 W less power dissipation! I double checked the calculations since I thought I did some mistake, hi.

This gives some ideas for the H-L v3: the VPA DC/DC converter output level could be controlled by the FPGA, so that when the drive level is lowered the PA (and driver) supply is lowered too and 9 V are used only for the max drive level. Maybe no real DAC/potentiometer could be needed and just some rough delta-sigma modulation of an output pin could be enough to inject an offset into the regulator feedback.

Regarding the output harmonics, the gain at 7.5 V is about 0.5 dB lower maybe also because I did not adjust the bias when changing the PA supply level. So for getting about the same power output, the TxPGA gain level had to be 1 step higher (TxPGA = 8 instead of 7) for the lower supply. Here is the power output for the two supply levels:

output power is practically the same but note that the frequency response is slightly different.
Here are the harmonics levels for the two cases

The higher supply produces less higher-order harmonics, but for the 2nd and 3rd the behavior is less clear.

The harmonics levels (and the power dissipation) vs. frequency are slightly different than the graphs posted previously; one of the reason could be that the previous measurements were done by lifting the PA output transformer center tap and measuring the current there; the tap bypass was done with just a single leaded 100 nF capacitor, while here all the bypass caps are in place as I cut the supply trace near the DC/DC converter.

I'll try do to also some measurements of the PA IMD; as seen before on the test PA, at 7.5 V is usually quite worse than at 9 V.

[Steve Haynal]

Hi Claudio,

Thanks for the measurements. It looks like some dissipation savings at 7.5V, but slightly higher 4th and 5th harmonics plus possibly slightly worse IMD. I am leaning towards separate power supplies for Vop and Vpa for the reasons I expressed earlier in this thread. There is an easy resistor combination for ~7.9V Vpa and ~10.1V Vop. One of the configurations I am considering might have the bias voltage on for the PA but the PA would be unpowered (Vpa==0) during low power RF output. I don't see a problem with this. Do you?

Yes, we should start adjusting Vpa based on the drive and desired output. :) The PA then starts looking like a class G or H amplifier.

73,

Steve

KF7O

[in3otd]

Hello,
here is the H-Lv2b2 PA measured IMD vs Pout,

at 7.5 V :

and at 9.0 V supply

the test_PA had a slightly lower IMD at higher power, maybe because the bias current was a little higher there.

BTW, if we will have PureSignal the IMD levels should of course improve quite a bit, even at 7.5 V. Do we need some particular coupling circuitry to have the right input level to the RX for this? I see there is of course quite some leakage from the PA out to the RX but I don't know if it will be enough.

Out of curiosity, I measured also the driver IMD:

and, as expected, its distortion is very low.

I also prefer separate power supplies for driver and PA, seems more flexible - even if a common supply may make sense if we go the "power tracking" supply route.

Having the PA with no drain supply but bias voltage on the gates should not be an issue.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Thanks for the measurements. What was the op amp supply voltage for the driver IMD measurements?

I'm not sure how much signal is needed for pure signal. I typically see my RX clip (with gain at +19dB) during TX so think that there is probably enough leakage. We can also feed some extra signal back via DB4 or DB16. We can vary the LNA gain (max of +20dB though) during TX to try an achieve a stable RX signal during TX. Let me know if you have any other ideas.

Regarding the TX inhibit you suggested, I think CN8, CN9 or CN10 can be run to the back edge of the board for this input, or we can use an input pin on front CN4. Let me know if you have a location preference or ideas regarding what the circuit should be.

73,

Steve

KF7O

[in3otd]

Hello,
for all the measurements the driver was supplied by the VPA converter, at about 9.3 V.

Also here the RX clip LEDs are on during TX, when the gain is not too low; I have actually never used PureSignal so I don't know if there are any particular constraints on the feedback level but I agree that if we can get the RX to clip that should be enough, hi.

I missed that there were already CN8/9/10 that could be used for the TXinhibit input; it should be ok even if they are not at the board edge and one has to run a wire to a nearby connector if this feature is needed. One point that's not clear to me is how it should be implemented by default: the TXinhibit pin should be grounded (TX enabled) when this feature is not needed (the majority of users, I think) and brought to a connector when this input is needed. So should we have a 0 ohm jumper to GND that those needing a TXinhibit can remove?

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

I was thinking we pull-up TX inhibit and a user only grounds this when desiring TX inhibit. It sounds like you think it should be the other way? User grounds TX inhibit when TX is allowed to be enabled?

I also have some thoughts regarding you other post, but will post them when I have more time.

73,

Steve
KF7O

[in3otd]

Hello Steve,
I was just "thinking HW" too much, either polarity will likely be fine. BTW, the Elecraft K3 and KX3 (and maybe others) allow to choose the polarity of the TX Inhibit input, we could also do the same. It might be a bit confusing but I think that as long as the default configuration is "TX enabled" that would be fine. Those needing the TX Inhibit function will have to read the manual, hi.
So in the end, no HW change is likely needed, this function can be assigned to CN8/9/10 by a user configuration and one will have to bring one of these inputs to their connector of choice.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Group,

A while ago John, Glenn and Graeme contributed designs and measured data for mesh filters to be used with the HL2.  I have been working on these, and I have some more data to contribute.

I have PCBs ordered from OSHpark that accommodate three filters switched by relays.  Each filter is two poles; two series LC resonators  and an intermediate capacitor to ground.  The filters were designed with Elsie, and the schematic is attached.  I built one with T37 toroids, and the second with 1812 SMD inductors.  Here are the in-band filter losses for the three bands 40, 20 and 10:

                                  40 meters              20 meters       10 meters

toroid T37-2                                               0.31 dB

toroid T37-1                  1.35 dB

toroid T37-2                  0.39

toroid T37-6                                                                      0.41 dB

Coilcraft 1812                                            0.71

Bourns 1812                 0.56

Bourns 1812                                                                     2.07

Coilcraft 1812                                                                   0.64

All filters are the same in that they have a design bandwidth of about 38% of center.  The Coilcraft inductors are wound on ceramic cores and are 1812CS series (eg. 1812CS-102XJLB), and the minimum Q is 60 at 50 MHz  The Bourns are wound on ferrite and are PM1812 series (eg. PM1812-1R0J-RC), and the minimum Q is 50 at 7.96 MHz.  Both are available from Mouser.

The 10 meter result was a surprise.  The Bourns inductor that was so successful at lower frequency had a 2 dB loss at 30 MHz and is completely unsuitable.  I have other measurements that show the Bourns have better Q at lower frequencies, and the Coilcraft are better at higher frequencies.  I don't know where the crossover is yet.   But proper measurement and selection of SMD inductors is critical for Tx filters, and no SMD inductor is suitable until it is proven to be so.

The toroids always have lower loss and higher Q; not a surprise.  The largest penalty in the table is 0.4 dB, which is 480 mW at the 5 watt level.  Although this is unwelcome, it is acceptable.  I ran the filters at 5 watts and my finger test showed the inductors were barely warm.  I had problems winding the toroids to the correct inductance, and measuring the inductance and Q.  I will write a separate post for that.  So I think SMD inductors should be considered for Tx, not just Rx.

John's filters had more poles and combined some bands.  My objective was sufficient attenuation at harmonics and a high return loss within the band.  With only two poles I could not make combining bands work, so each band will need its own filter.  John also designed for a single output transistor.  This design assumes push-pull finals with low second harmonics; thus it has fewer poles.

I measured the "filter" attenuation without the filters; that is, with the bare board.  At 10 MHz, it is -79 dB and at 100 MHz it is -61 dB.  Adding the relays (still no filters) changes this to -74 and -56 dB.  Since there is significant bleed through around the filters I did not think increasing the poles was worth it, but this could change.  It may be necessary to add a final low pass VHF filter like the HPSDR has.

The capacitors were a bit of a problem.  With a 50 ohm load the voltage across the first capacitor is multiplied by the filter loaded Q of 3.  With an open circuit load I measured a voltage 5.8 times the input.  For 5 watts, the input is 22 volts peak, so that is 130 volts across the capacitor.  It is easy to get high voltage 5% capacitors, but they are only available in the 10% series values 10, 12, 15, 18 etc.  The intermediate 5% values 11, 13, 16 are unavailable or are expensive.  So I made the board with two pads for each capacitor and it is often necessary to use two parallel capacitors for each value.  I used the muRata GRM series NP0 630 volt 1206 capacitors because I had them on hand.

I don't consider either of the filter boards to be a final design.  The original idea was to make a simple board with through-hole parts that would be easy to build.  But the board requires two SOIC ICs so it can't be 100% through-hole.  And I had trouble winding and measuring the toroids.  I measured with my AADE L/C meter but I sometimes had to remove a turn from one toroid but not the other to get the correct inductance.  And builders may not have an L/C meter.

If the board were for me, I would not hesitate to use SMD inductors.  I think I can squeeze 5 filters on the board and I think the 1812 parts are easy enough to solder.  If the whole board is SMD, everything but the filter parts could be manufactured.  Then the user would only need to add relays and 1206 and 1812 filter components.  So I just don't know what to do next.

I realize that this design will not appeal to everyone, and that others are working on combination low pass / high pass designs.  With these, you can attach a multi-band antenna and receive on several bands at once.  Transmit still requires the correct filter.  But the bandpass design is simpler, and offers a custom low pass for each band with no cost or space penalty.  Whether this is important depends on the QTH.  Perhaps for field day, bandpass is better, while for a country location, hi/lo is fine.

Jim

N2ADR

[Graeme Jury]

Hello Jim and Group,

Thanks Jim for the post. I am really excited to see your project as I felt that what I am doing would not appeal to all and was going to be difficult to build due to having to use hand wound and calibrated toroids. Having a version which is straight forward is going to be good for the project. I expect other versions will filter out if I may use that expression.

A few comments; I was interested in how the smd inductors turned out for you. I wanted to try them out myself but the inductor values for my filters would not fit the standard smd values. It was difficult enough to manipulate the filter values to get the capacitors to fit. I also ran into the E96 values and voltage limitations and like yourself have had to parallel to keep to 630 volt units. I have ordered most of mine from China so hope they will be OK and will report to the group as I build them up. My intention was to use single pads and piggyback the other caps on top. I have used this method before and it works well even 0805's will piggyback on 1206's.

I have yet to measure the IMD generated from the switching diodes and have gone that way because the filters will need to be switched on the fly if I am transmitting say on 40M band but have the filters set for 80M Hi pass and 20 M Lo pass so I can skim those bands and of course I need to switch in the 40M LPF to transmit.

Steve if you are reading this, I will need some info from you about what is going to happen in the I2C control to the filters but have not bothered you at this stage while you are concentrating on getting beta3 out but when you are clear I will have several questions like will PTT be over I2C and so on.

I have included the nearly complete schematic. On the atmega328, A0 to A3 will be for filter control lines and there will be a selection for this or I2C input to enable it to be used with HLv1 or HLv2. Still a long way to go but at least I am now finally home and can get some useful time on it.

73, Graeme zl2apv

[James Ahlstrom]

Hello Graeme,

On Wednesday, June 14, 2017 at 10:31:47 PM UTC-4, Graeme Jury wrote:

I wanted to try them out myself but the inductor values for my filters would not fit the standard smd values. It was difficult enough to manipulate the filter values to get the capacitors to fit.

For the toroid version, I also  also adjusted the center frequency and bandwidth to get easy capacitor values.  But for the SMD inductor version, I adjusted for standard inductor values and secondarily for easy capacitor values.  I was often left with odd capacitor values.  It also means that each band needs a custom filter to juggle inductor values, capacitor values, pass band and return loss.  If I go with a toroid version, I will adjust inductor values to integral toroid turns.

I have also has good luck piggybacking capacitors.  I made two pads so filters could be manufactured.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

While building filters for HL2, I needed to measure the inductance and Q of toroids and 1812 inductors.  I have always used my AADE L/C meter, but it did not agree with my return loss bridge, and I didn't feel comfortable with the results.  So I bought some 1% inductors and capacitors and did some research.  After some misadventures Googling for Q-meter I finally found what I was looking for in "Experimental Methods in RF Design" on page 7.36.  You just connect an RF signal generator and detector together with an SMA "T", and connect a series LC from the center to ground.  If the capacitor is 1%, the dip in response at resonance gives the inductance.  What I didn't know was that the depth of the dip (the attenuation) is a direct measure of Q.  I hope your browser can render these equations:

A picture of my test fixture is attached.  The left SMA uses the variable capacitor and is used to measure Q at variable frequencies.  An 1812 inductor is shown.  The right SMA uses one of two 1% capacitors and is used for measuring inductance.  A toroid is shown.  Connections are made by tack soldering.  I am getting good results with this setup, except that my 1% 3.9 uH inductor measures 3.6% low.  I only have two 1% inductors, but I plan to get more and continue to experiment.

Meanwhile, here is the measured Q of the Bourns 1.0 uH 5% inductor that had high losses in the 10 meter filter.

Freq MHz        Q

  7.933             52

10.060             50

15.048             35

20.229             23

30.167             13

Other measurements show that the Coilcraft inductors have low Q at lower frequency, and the Bourns are preferred.  Both Coilcraft and Bourns are available at Mouser, but not Digikey.

I also discovered that my AADE L/C meter reads higher inductance for lower Q.  For example, it measures 4.338 uH for a toroid with high Q.  But if I add a 1 ohm resistor in series, it measures 4.545 uH, or +4.8%.  The Q with the resistor is 20 at a 750 kHz test frequency.  Remember that the AADE measures at 750 kHz or lower, but the test fixture can measure at the operating frequency.

I hope this is useful to someone.

Jim

N2ADR

[Steve Haynal]

Hi Jim and Graeme,

Thanks for your work on filters and the valuable data regarding the inductors and capacitors you experimented with. Jim, what are your thoughts on the Bourns SMD inductors for 160M and 80M filters? Will they do the job or will we need to hand wind for the lower frequencies? Also, I am very interested in how the fwd/rev power works out on your board. I assume you haven't tested that yet as the boards are on order?

Graeme, so far only the band select via the mcp23008 is via I2C. I was thinking PTT would always be via CN11. We can discuss additional PTT over I2C as there would be no changes to the HL2 board to support this.

Regarding bandpass filters, I have been noticing interesting behavior with my ZM-2 Z-match tuner again. I recently put a long wire antenna on the peak of my roof and tune it with this tuner. The tuning is very sharp, which is not always the case for things (fences, gutters) I've tuned with this tuner. When I have a sharp tuning like this, I can increase the LNA gain up to +36 dB before I see clipping. This is in a fairly noisy suburban environment. With other antennas, like my attic dipole, I see clipping in the upper teens to mid +20 dB range depending on the time of day. I usually run with a LNA setting of +19 dB, and assuming there are no benefits from increased gain in my environment (some would say I can run at even lower gain settings and achieve similar receiver results), this means I have roughly >2 bits (~6 dB per bit) of headroom when using the Z-match. With this configuration, I could get by with the 10-bit ad9865 and not notice much difference. In fact, I'm thinking a bit of injected dither might help with the ad9866. When I compare this antenna to my attic dipole, it "feels" very similar in terms of JT9/65 and WSPR spots. I haven't seen any second order IMD improvement due to this "preselector", but I haven't tried any systematic measurements. Where I am going with all this rambling is that I can see the practical value of RX bandpass filters to help with the limited 12-bits of the AD9866 given broadband noise in general, but have yet to see (and want to see) the practical benefits for IM2.

73,

Steve 

KF7O

[James Ahlstrom]

Hello Steve,

On Saturday, June 17, 2017 at 1:50:21 PM UTC-4, Steve Haynal wrote:

 Jim, what are your thoughts on the Bourns SMD inductors for 160M and 80M filters?

I have the Bourns parts on hand, so I will test them and report back.

 

Also, I am very interested in how the fwd/rev power works out on your board. I assume you haven't tested that yet as the boards are on order?

I have the boards and I have built and tested fwd/rev power.  So far I see that 5.6 watts produces 1.2 volts DC at the forward diode.  And I see that the reverse power null adjustment pot is not necessary, as the null is very broad.   Voltage at the reverse diode at 50 ohm load and 5 watts was 71 to 16 mV for max and min pot, with 0.1 mV at mid range.  This was on 20 meters.  But consider an antenna with an SWR of 1.2 corresponding to a return loss of 20 dB.  The reverse voltage would be 1/10 of the forward voltage, or 120 mV.  The diode forward voltage drop will dominate the measurement.

A good reference is "A Simple and Accurate QRP Directional Wattmeter" by Roy Lewallen, W7EL, QST Feb 1990.  He gets accuracy at low power by matching the SWR bridge diodes to the op amp compensation diodes, and choosing DC impedance levels carefully.  But the op amp compensation diodes are on HL2 and will not necessarily match the diodes in the SWR bridge.  We may have to live with an inaccurate relative SWR indication, although that would still be useful.  Or we could ask the user to calibrate the bridge and supply the correction in software.  I won't be able to work on this until I have a working HL2 with fwd/rev power included.

any systematic measurements. Where I am going with all this rambling is that I can see the practical value of RX bandpass filters to help with the limited 12-bits of the AD9866 given broadband noise in general, but have yet to see (and want to see) the practical benefits for IM2.

I might be missing something, but IM2 means that if you are listening on 14 MHz, you don't want to hear the sum of two signals on 6 and 8 MHz.  But bandpass filters will attenuate these signals and reduce the sum by twice the attenuation.  Whether this is a practical benefit depends on whether there really are any two signals below 14 MHz that add up to 14 MHz.

Jim

N2ADR 

[Graeme Jury]

Hello Jim and fellow experimenters,

That information is very valuable Jim and in line with what I have found. I wonder if you have come across the spreadsheet done by Jaques Audet VE2AZX available here which can be used in association with his QEX article here. This method is very accurate and shows how toroid inductances vary with frequency etc. I do find a few anomalies but the results are within ham shack experimental range. His whole web site is a mine of information. I use a test jig made from a variable capacitor with 2 bnc connectors and a copper strip soldered between them. The scale of the capacitor is calibrated in pF and the gangs can be switched in or out with slide switches. I previously had a single gang version and this was the best as the larger capacitor introduces some unwanted reactances but is still a useful tool. The best tool for this kind of measurement is a small board made by Claudio which is simply a 50 ohm stripline with a pair of sma connectors each end. I solder a low inductance 150 pF capacitor to ground and connect my inductors to this and the centre of the stripline. I have attached a picture of both units.

73, Graeme ZL2APV

[James Ahlstrom]

Hello Graeme,

Thank you for the references and pictures.  Your stripline seems better than my T at high frequencies because of the controlled impedance.   My T will have more stray impedance.  But I think I am OK at HF up to 30 MHz.  Your comments give me more confidence that the method I am using gives valid results.

Jim

N2ADR

[James Ahlstrom]

Hello Steve and Group,

I tested the Bourns inductors for Q.  The 160 meter filter uses 12 uH inductors, and the Q is 58.  The 80 meter filter uses 6.8 uH inductors, and the Q is 73.  So far, so good.

But then I decided to power test the 40 meter filter that uses 3.9 uH Bourns inductors.  I used my Pennywhistle HPSDR amp and Quisk.  Power was 9.5 watts because I wanted to stress test the inductor.  One of the Bourns inductors failed.  The inductance changed to 1.12 uH.

Next I tested the 20 meter filter that uses 2.2 uH Coilcraft inductors at 12.5 Watts.  The inductors survived, but were very hot to the touch.

I know that others have used SMD inductors as small as 1206.  Did you test at high power, and did you experience any failures?

Jim

N2ADR

[Glenn P]

Hi gents

RS-HFIQ project uses SMD inductors in the Tx LPF side. It's a nominal 5W rig though.

For interest, here are the type used:

Glenn - vk3pe

	RS-HFIQ   Tx LPF inductor types
	line:-
	54	3	L2, L7, L8	FIXED IND 1.8UH 900MA 160 MOHM	BOURNS	CC453232-1R8KL	MOUSER	652-CC453232-1R8KL
	59	1	L3	IND 560NH 460MA 1.1 OHM	ABRACON	AISC-1210-R56J-T	MOUSER	815-AISC-1210-R56J-T
	62	3	L4, L5, L10	IND 270NH 630MA 400 MOHM	ABRACON	AISC-1210-R27J-T	MOUSER	815-AISC-1210-R27J-T
	63	2	L6, L11	IND 180NH 700MA 300 MOHM	ABRACON	AISC-1210-R18J-T	MOUSER	815-AISC-1210-R18J-T
	64	1	L9	IND 470NH 490MA 800 MOHM	ABRACON	AISC-1210-R47J-T	MOUSER	815-AISC-1210-R47J-T

[Glenn P]

there are 5 LPF in this rig.

[James Ahlstrom]

Hello Glen,

Thanks for the info on the low pass filters.  I can't seem to find a schematic.  Are the LPF Chebyshev or elliptic?

Jim

N2ADR

[in3otd]

Hello Jim,
for schematics and BOM see the links at the bottom of this page.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Jim,

There was some discussion RS-HFIQ filters on this list in the past: https://groups.google.com/d/msg/hermes-lite/2ARSvdvlEfQ/VmaSZ0OCBwAJ

Jim Veatch has been very helpful and open about his designs in the past. I will forward hist contact info to you privately. You can ask him about heat and various issues when using such devices. He did mention to me that he was not 100% happy with the 80M filter. That may be why there is no 160M on that radio. It is what led me to think that we will have to use some sort of hybrid with wind-your-own for 40M (or 80M) and below but SMT for higher frequencies.

The EMRFD group may have reports of filters with SMT devices.

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

The FWD/REV circuit is lifted from the Alex opnHPSDR project. See the Alex manual under the documents link. The schematic and BOM are at the very end of the manual. John Williams was the first to start experimenting with it and the HL. There was some prior discussion in this thread.

My understanding of IM2 is as you describe. I've just never seen it happen yet in practice. I agree that a traditional bandpass filter will help if IM2 is occurring. Whether or not IM2 matters day to day use is relevant to the design process of a budget radio. If it does matter, than you need narrower filters to block the 6 and 8 MHz when using 20M. If it does not really matter in practice and is mainly a lab room creation, then you may only need to block parts of the spectrum where there are many strong signals, for example AM or lower HF on very active nights. The latter is a simpler filter requirement and still allows for multiband skimming. I prefer the latter, but would like to justify it with proper measurements.

73,

Steve

KF7O

[James Ahlstrom]

Hello Steve,

On Tuesday, June 20, 2017 at 2:18:41 PM UTC-4, Steve Haynal wrote:

The FWD/REV circuit is lifted from the Alex opnHPSDR project. See the Alex manual under the documents link.

That is the tandem match circuit.  I didn't use that as it requires two transformers and I was short of room.  I am using a version of the Bruene  circuit.  It uses one transformer to sample the current, and a resistor divider to sample the voltage.  Other versions use a capacitor divider.  I won't know whether it really works until I can test it with the final HL2.

Jim

N2ADR

[James Ahlstrom]

Hello Claudio,

On Tuesday, June 20, 2017 at 1:10:54 PM UTC-4, in3otd wrote:

for schematics and BOM see the links at the bottom of this page.

Thanks for the link.  It was very useful.

We should be careful about copying the RS-HFIQ filters.  Their radio is a direct conversion type like Softrock, and shows a strong response at odd harmonics.  So if we receive at 7 MHz, we need to reject 21 MHz.   So the radio has bandpass filters in the receive path.  They also protect the analog mixers.  The HL2 only requires rejecting the higher Nyquist zones above 30 MHz and the filter for that is on the HL2 main board.

The transmit path goes through the receive filters, then through the power amp, then through Chebyshev or Cauer filters to the antenna.  These filters could be useful for HL2, but are low pass and not bandpass.

Jim

N2ADR

[James Ahlstrom]

Hello Glen,

Thanks for the inductors.  I took a close look at the 1.8 uH Bourns.  It is wound on ferrite like my PM1812, but has a higher current rating of 900 ma.  My PM1812 in 3.9 uH is rated 330 ma.  But the CC453232 tolerance is 10% instead of 5%, and the minimum Q is 10 instead of 50.  It is used in a Chebyshev LPF and I doubt it would work in a bandpass filter.

The other inductors are 1210 and wound on ceramic, not ferrite.  And their values are 180 nH to 1.8 uH.  My bandpass filters need values 1.0 to 6.8 uH.  The bandpass filters are more difficult because the voltage, current and losses are multiplied by the Q of 3.

Did you or John or Graeme run your mesh coupled bandpass filters at power?  And did any inductors die?

Jim

N2ADR

[Graeme Jury]

Hello Jim,

The bandpass filters with the smd inductors were on the receive filter board and fed via PE4259 switches which will not withstand 5 watts so I have not tested at power. If you like I can whip the 2.2 uH inductors out and build a 30/20M filter and smoke test it for you. I think I have some suitable 630 volt capacitors as the ones in circuit are 50 volt types which are fine for Rx only.

73 Graeme zl2apv

[James Ahlstrom]

Hello Group,

I am trying to figure out why my 3.9 uH Bourns PM1812 inductor failed.  The data sheet does not specify an RF power limit, but it does specify a max DC current of 330 ma.  The DC resistance is 0.9 ohm, so the DC power is 98 mW.  This is very small, so the assumption is that the limit is not temperature, but saturation in the ferrite.  Five watts at 50 ohms has a peak current of 447 ma, and that exceeds the DC maximum.  And then we must multiply by the Q of 3 so the AC peak current is about 1.3 amps.  And I was running almost 10 watts.  Goodbye PM1812.

Glenn gave the part number of the 80 meter filter inductors in the RS-HFIQ.  It is a Bourns CC453232 high current ferrite inductor with a 900 ma max DC current.  This handles the 447 ma (probably) but only in a Chebyshev or Cauer filter, not the 1.3 amps in a bandpass filter.

We can compare the Coilcraft 1812CS 3.9 uH inductor.  The specified DC current is only 240 ma, but this is based on a temperature rise of 15 degrees.  There is no ferrite to saturate as it is wound on ceramic.  So chip inductors wound on ceramic can handle power until they melt.

Choosing a chip inductor is turning out to be a lot more complicated than I thought.  The ceramic ones don't saturate, but the best Q is around 50 MHz and Q goes down fast at lower frequencies.  Ferrite inductors have good Q at 8 MHz, but no good spec on power handling.  And coupled resonator bandpass filters need much better inductors that regular low pass types.  And bandpass filters need inductors from 1.0 to 6.8 uH.  Low pass filters need smaller inductors from 0.18 to 1.8 uH so are not really comparable.

I am starting to think that chip inductors will never work in my bandpass filters, and I must use toroids.  But I will continue to measure and study them for use in high pass / low pass filters.

Jim

N2ADR

[Graeme Jury]

Hello Jim,

With T37 toroids I have run them up to 30 watts successfully but in Chebyshev or Cauer filters only. A possible solution for your smd inductors is to go away from the mesh filters and cascade a Hi/Lo pass filter for your bandpass. The tradeoff is that you will need one extra inductor possibly and unfortunately all the inductors will have different values making it very difficult to juggle filter parameters to get standard inductor values. The capacitors will not be so bad with series or parallel combinations. Electronics is no exception to life - everything is a tradeoff :-)

73, Graeme zl2apv

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[James Ahlstrom]

Hello Graeme,

On Wednesday, June 21, 2017 at 2:41:35 PM UTC-4, Graeme Jury wrote:

The bandpass filters with the smd inductors were on the receive filter board and fed via PE4259 switches which will not withstand 5 watts so I have not tested at power. If you like I can whip the 2.2 uH inductors out and build a 30/20M filter and smoke test it for you. I think I have some suitable 630 volt capacitors as the ones in circuit are 50 volt types which are fine for Rx only.

Thanks, but a smoke test is not needed.  I think the way to test ferrite inductors is to test with and without the filter at 1, 2, 5, 10, 20  watts, and look to see whether 2-tone IMD and harmonics increase with the filter in-line.  That would indicate non-linearity in the ferrite.

I thought someone used mesh coupled bandpass filters at power in Tx.  Was it Glenn or John?  Am I the first?

Jim

N2ADR 

[John Williams]

I used bandpass filters borrowed from the Peaberry design in the Superband board. Very small run of cards and never did solve issues with sharpness of the filters as we then moved to the multiband tx board with lpf filters.

--

[James Ahlstrom]

Hello Graeme,

On Wednesday, June 21, 2017 at 3:52:21 PM UTC-4, Graeme Jury wrote:

A possible solution for your smd inductors is to go away from the mesh filters and cascade a Hi/Lo pass filter for your bandpass.

Yes, exactly right.  I will keep measuring parts, both torroids and SMD.

Jim

N2ADR 

[Graeme Jury]

I used the mesh filters Jim but with T37 toroids which I did run run up to 30 watts with no sign of heating but I did not do any IMD testing. I had sent the group a plot from quiskVNA of one of the filters but as far as I know you are the first to try smd inductors in TX mesh filters. I will be doing the same non linearity testing on my Hi/Lo pass filter set as you propose as I am concerned about non linearity from the switching diodes although the testing Claudio did and has published on his web site indicate they will be pretty good.

73, Graeme

[Steve Haynal]

Hi Jim,

Sorry, I didn't notice you had switched to the Bruene circuit. The circuilt I've been considering is the tandem match circuit. You can wind the two coils on a single binocular core to save space as discussed here: https://groups.google.com/d/msg/hermes-lite/HyG6kNSas58/IktZjxKsAwAJ

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

My interest in the RS-HFIQ has been mainly the LPF using SMT inductors for the TX path, not the bandpass filters that are also included in the RX path.

Thanks for investigating SMT inductors for HP, LP and BP filters. I think there is some missing information there, at least for me.

73

Steve

KF7O

[Steve Haynal]

Hi John,

One of you PA/filter boards had the option to use 0805 inductors in the TX LPFs, IIRC. Did anyone try those?

73,

Steve

KF7O

[James Ahlstrom]

Hello Group,

On Wednesday, June 21, 2017 at 3:09:30 PM UTC-4, James Ahlstrom wrote:

>> And then we must multiply by the Q of 3 so the AC peak current is about 1.3 amps.

After looking at some computer models, I see this is incorrect.  The current is not multiplied by the Q, the voltage is.  The current is unchanged.  But the peak current was 600 ma, still greatly in excess of the 330 ma DC current limit.

Jim

N2ADR

[James Ahlstrom]

Hi Steve,

On Saturday, June 24, 2017 at 1:21:27 AM UTC-4, Steve Haynal wrote:

My interest in the RS-HFIQ has been mainly the LPF using SMT inductors for the TX path, not the bandpass filters that are also included in the RX path.

Me too.  I have those parts on order and will test them.

Jim

N2ADR 

[John Williams]

Hi Steve,

Not that I am aware of. Perhaps Glenn or Graeme will recall.

John

[irbsu...@yahoo.co.uk]

Hi Steve,
I have an RF-HFIQ board and have done some tests looking at the Tx harmonics which are not bad. Anything in particular you would like measured?
Andrew
G4XZL

[Sid Boyce]

Same here John,
Both the ones I received use T37 cores.
73 ... Sid.

>> <mailto:hermes-lite...@googlegroups.com>.

>> For more options, visit https://groups.google.com/d/optout.
>

> --
> You received this message because you are subscribed to the Google
> Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to hermes-lite...@googlegroups.com

> <mailto:hermes-lite...@googlegroups.com>.

> For more options, visit https://groups.google.com/d/optout.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[Steve Haynal]

Hi Andrew,

Given Jim pointed out problems with heat/current, I'd be interested in how much heat you experience, especially on the lower bands. Also, what filter losses are you seeing? I am surprised and a bit skeptical that it achieves adequate harmonic suppression with only 5 LPFs and a single transistor PA richer in second harmonics. You and others worked hard to achieve that with John's PA. What attenuation is being achieved for 2nd and 3rd harmonics on every amateur band? We can look at the harmonic output from the HL2beta2 and see if these filters will be adequate.

73,

Steve

KF7O

[irbsu...@yahoo.co.uk]

Hi Steve,
No problem, I will look at taking the measurements later this week, or possibly early next. Regarding heat, all I can so at the moment is that the PA gets very hot very quickly, need to get a temperature probe on it. It runs in class A, not quite sure what the standing current is supposed to be.
Andrew
G4XZL

[James Ahlstrom]

Hello Andrew,

On Saturday, June 24, 2017 at 3:30:05 PM UTC-4, irbsu...@yahoo.co.uk wrote:

I have an RF-HFIQ board and have done some tests looking at the Tx harmonics which are not bad. Anything in particular you would like measured?

I am curious about whether the harmonics are all at -43 dBc, and how much the image from the Tx mixer is suppressed.

Jim

N2ADR 

[James Ahlstrom]

Hello Group,

I got the SMT inductors used in the RS-HFIQ except for the 180 nF unit.  I tested the Q at the frequency of each filter.  Here are the results:

Inductance   Mfg.      Part No.                   band, Q

1.8 uH    Bourns      CC453232-1R8KL    80 meters  Q=30 ; 60 Q=27 ; 40 Q=22

560 nH   Abracon    AISC-1210-R56j-T    60  Q=19 ; 40 Q=22

470 nH   Abracon    AISC-1210-R47J-T   30 Q=26 ; 20 Q=29

270 nH   Abracon    AISC-1210-R27J-T   30 Q=23 ; 20 Q=27 ; 17 Q=28 ; 15 Q=29 ; 10 Q=31

Now we can make Elsie files for each filter and estimate the performance.  For filters with two different inductors, I used an average Q, and for the 180 nF inductor I assumed a Q of 32.  The Elsie files are attached.

I am focused on whether the inductors may overheat, not on the stop band performance.  For that, see the Elsie graphs.  Briefly, the 10/12, 15/17 and 20/30 filters have a 0.3 or 0.4 dB loss.  The 40/60 filter has a poor return loss.  And the 80 meter filter has a loss of 1.8 dB.  At five watts, a loss of 1 dB is 1.03 watts.  See the graphs.

The Abracon units are ceramic core, and the Q increases with frequency.  The Bourns is ferrite, and the Q decreases with frequency.  So far it looks like the SMD inductors may work at higher HF, but not at 80 and 160 meters.  Three of the filters are Chebyshev (or similar) and two are Cauer.  Different filter typologies use different inductance for the same band.  So the topology interacts with the inductor availability and Q, and makes things complicated.  None of these parts has anywhere near the Q of a toroid.  I will email Jim Veatch and ask him about his experience with these filters.

Jim

N2ADR

[Graeme Jury]

Hello Jim and Group,

I have collected up my designs for my HP and LP filters and made a catalog of each which I have attached to this post. I have built several of them and they have all come out very close to the design so anybody wanting to play with filters can do so using these designs with confidence. In each case I used T37 toroids and have decided to stay with them in spite of the disadvantages of size, adjustment and winding tedium In favour of less passband loss, higher power handling and ability to adjust inductance exactly to that required. In practice there was no need to adjust the turns but a little improvement can be gained if you do. I got most of my capacitors from Ali Express and they averaged out at around $US4.00 for 100 but are messy to find values and I had to go to many suppliers but I did get 630V 1206 COG by hunting around. A few sizes were not obtainable and I had to go to Element14 at around $US5.00 for 10.

Jim, I have built a mesh filter for 30/20 in both SMD inductors and using T37 toroids. There was little difference in the stopband regions but the passband loss is higher in the SMD version. I will do some plots with quisk_vna tomorrow and post them so you can see the difference.

Hope the designs are useful.

73 de Graeme zl2apv

[James Ahlstrom]

Hello Graeme,

Thank you for the valuable information.  I am finishing up my work on SMD inductor filters.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

My 40 meter mesh filter with 3.9 uH Bourns PM1812 inductors failed at 10 watts.  So I replaced the inductors with higher power Coilcraft 1812FS-392JLB  inductors.  These are specified on the data sheet as filter inductors and they have magnetic shielding.  Unfortunately the Q at 7 MHz is only 30.  I measured the transmission in-band as -1.25 dB.  I tested them at 2.5 watts output after the filter, and they were too hot to touch.  So I can not use them in a mesh filter.  My other inductor choices have a similar Q or worse.  Too much power is being dissipated in the inductors.  Back to the drawing board.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

My failed 40 meter mesh filter used a 3.9 uH SMD inductor wound on ferrite.  But lower valued inductors are wound on a ceramic core, so I decided to test one of those.  I built the RS-HFIQ 15/17 meter filter that uses 270 nH inductors in a 1210 SMD package.  I previously measured the Q of these inductors as 29 at 21.144 MHz.  I modeled the filter in Elsie, and the predicted transmission at 21.2 MHz was -0.46 dB.  Interestingly, the transmission at 10 MHz is -2.4 dB, so this is an optimized filter and not the usual 0.5 dB Chebyshev.

I build this filter on a scrap of PCB on pads milled out with a rotary tool.  The measured transmission at 21.2 MHz was -0.95 dB, a bit higher than expected.  But remember that this is not necessarily due to power dissipation.  It can also be caused by impedance mismatch.  I then ran the filter for several minutes at 5 watts out.  The inductors were barely warm.  So it seems SMD inductors of modest Q and wound on ceramic are usable for the higher bands.

I am still not sure why ferrite inductors don't work.  Perhaps the ferrite is unsuitable at HF or the inductors are saturating.  There is little insight available from the data sheets.

Jim Veatch, WA2EUJ was kind enough to respond to my request for information.  His design goal was to have 5 watts output, not to minimize filter losses.  And he wanted to manufacture a completed unit, not a kit.  So toroids were out.

Jim

N2ADR

[Graeme Jury]

Hello Jim and Group,

I was thinking about the ferrite heating problem you had Jim and thought I would try inductors on standard 1/4" slug tuned formers. To my surprise they work very well and although Q is low and through losses higher than toroids, they have very good shape and out of band attenuation. As you are designing for a minimal number of filters there is probably space for these coils in a can to fit the board. The downside is that it would be necessary to tune each coil to the appropriate zero frequency but with slug tuning this is easy and even a receiver tunes to the zero frequency listening for the dip will do. I tested the filter at sustained 20 watts and there was the tiniest bit of warming. Maybe not what you are looking for but it will work. Great on the breakthrough with the ceramic inductors at Higher frequencies.

73, Graeme zl2apv

[James Ahlstrom]

Hello Group,

When cascading high/low pass filters it is well to keep in mind that they interact.  For example, a 40 MHz Butterworth 3-pole roofing filter has an input impedance of 100 -13J at 29 MHz, and will affect a 10 meter filter ahead of it.  Even at 21 MHz, the input impedance is 66 +3J.  We need to model the whole filter chain for cascaded filters.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

I am trying to figure out what types of filter would be useful for HL2.  So I modeled 5 different filters designed for 10.1 MHz and compared them.  The filters are low pass Cauer, Chebyshev, mesh coupled, Pi and Tonne opt.  The Pi filter is two pi networks in series, and the Tonne filter is an optimized low pass by Jim Tonne as designed by his software.  Making the filters "the same" is a judgement call.  The filters have different shapes, but I am trying to get similar attenuation at harmonics and similar return loss in-band.  I modeled with Elsie and assumed a Q of 30.  Here are the results.

Filter               In-band transmission  Inductor value

Chebyshev                      -0.94 dB        910 nF

Mesh                               -0.99            2700

Cauer                              -0.50              820, 620

Pi                                     -0.88             620

Tonne                              -1.16              510, 430

The Cauer seems to have a bit less loss that the other types.  I thought the mesh coupled would have greater loss because it is two resonators, but this is not the case.  But the mesh coupled uses a much larger inductor value that the other types.  The inductor can be reduced by using end matching capacitors to ground, but I didn't try this, as it adds parts.

The available SMD inductors with ceramic cores show falling Q as the inductance increases.  Switching to ferrite cores at higher inductance increases Q but I haven't found SMD inductors that work.

So what can we conclude?  The mesh coupled filter is problematical because it needs a much larger inductor value.  But it is fine with toroids, and since it is bandpass, it needs no high pass roofing filter to get rid of 1 MHz broadcast signals.  The Cauer looks good, but it requires specific inductor values that may not be available.  And I have found it requires stricter tolerances.

I think that at least a single high pass roofing filter (maybe better said as a "flooring" filter)  will be desirable with a 2 or 3 MHz cutoff to get rid of AM broadcast.  But such a filter will affect the design of subsequent filters.  Of course, this HP filter is only needed on Rx, so if we put the T/R relay on the filter board, the filter requirements are simpler, but the switching becomes more complicated.  BTW all this makes be wonder how anyone operates on 160 meters right next to AM broadcast.

The RS-HFIQ Tx filters are a mixture of Chebyshev-type and Cauer-type, so it looks like Jim designed a custom filter for each band group based on what inductors were available.

I realize that the ideas I am presenting are a bit half-baked, but they help my understanding, and I hope they will be useful to others.

Jim

N2ADR

[Graeme Jury]

Hello Jim,

Very interesting experiments with even more useful results. Using a mixture of filter types to get appropriate inductors is a novel and practical idea. The big disadvantage of Cauer and Tonne filters is that different values are needed for the inductors which makes it hard to design from the limited range of the devices available but if a pair can be found the Cauer gives the best result. You have alluded elsewhere that the Tonne is not a good candidate for switched cascaded filter sets as it has a slight bandpass characteristic. I have included plots of the filter set that I am currently using on my HiQSDR on the output of my home brew 10 watt amp. Again I am sorry but they are only Toroids used there but they do give an indication of how the filter set performs and that there is no worries about wild input or output impedances as the S11 plots show.

I just have to mention Jim that your gift to the ham radio community of your Quisk VNA has enabled me to do work with filters that I could never have achieved any other way and I am very grateful.

73, Graeme zl2apv

[James Ahlstrom]

Hello Graeme,

Thank you for the kind words, and I am glad you find quisk_vna useful.  I look forward to making it work with HL2.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

I built and tested the RS-HFIQ 80 meter filter.  It uses 1.8 uH Bourns CC453232 ferrite inductors.  The Elsie model has a transmission at 3.5 and 4.0 MHz of -1.68 and -1.95 dB, but I measured -0.88 and -1.22 dB.  At 3.75 MHz, the impedance is 39 -14J, the return loss is -14 dB and the SWR is 1.5.

I ran the filter at 5 watts, and the inductors were a bit warm.  At 8 watts, they were hot.  I thought the increase in temperature might indicate non-linearity, so I tried to check for amplitude compression and IMD increase.  But that was difficult because the varying power level changed the gain and IMD in the amp chain.  But I did not see anything alarming with the filter.  So I conclude that this 80 meter filter can handle 5 watts.

So why does this ferrite filter work and my 40 meter filter fail?  This inductor is 1.8 uH, is rated at 900 ma, and is used in a Chebyshev-type filter at 3.75 MHz.  My inductor is 3.9 uH, is rated 330 ma, and is used in a mesh-coupled filter at 7.15 MHz.  Some inductors work and some don't.

Since the 80 meter filter works, it gives me more confidence that all HL2 filters can be SMD.  But I don't think we can just copy the RS-HFIQ filters because I think they are optimized for the second harmonic.

Jim

N2ADR

[Steve Haynal]

Hi Jim,

Thanks for all your experiments. You say below, "But I don't think we can just copy the RS-HFIQ filters because I think they are optimized for the second harmonic." Even if they are optimized for the second harmonic, won't third harmonic suppression still be sufficient? Can you elaborate a bit more on your concerns?

73,

Steve

KF7O

[in3otd]

Hello,
I've simulated the RS-HFIQ TX filters response, to have an idea of the harmonics attenuation we could expect. Inductor Q was not modeled as it should not have a significant effect in the stopband. The 40m/60m filter response looks a bit strange to me, there may be a mistake somewhere.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Claudio

On Saturday, July 15, 2017 at 1:02:11 PM UTC-4, in3otd wrote:

The 40m/60m filter response looks a bit strange to me, there may be a mistake somewhere.

The zeros in the passband are at 10.8 and 13.1 MHz, and that does seem odd for a Cauer filter.  I assumed they were moved to the second harmonics of the 60 and 40 meter band, which are at 10.8 and 14.3 MHz.

Jim

N2ADR 

[James Ahlstrom]

Hello Steve,

On Saturday, July 15, 2017 at 11:17:04 AM UTC-4, Steve Haynal wrote

suppression still be sufficient? Can you elaborate a bit more on your concerns?

When designing low pass filters, for example Chebyshev or elliptical, the classic design requires a limited amount of ripple from zero to the cut off frequency.  If we relax that requirement we can get better attenuation in the stop band.  That is the point of Jim Tonne's optimized filters.  We don't really need limited ripple below the band, so we optimize and alter the classic designs.

In general, we can optimize filters by adjusting the Ls and Cs to get better objectives, namely better return loss, or better 2nd harmonic or 3rd harmonic suppression.  But, of course, they trade off against each other.  An elliptic filter is a case in point.  If you put a passband zero at the second harmonic, it may very well be that the 3rd harmonic suffers.  In a Chebyshev, if you ask for more 2nd harmonic suppression, the 3rd harmonic will probably be fine, but the return loss will suffer.  A bad return loss means the AFT05s will not be operating into 50 ohms, although they may still be OK with the load they have; or maybe not.

The RS-HFIQ has a single-ended output, and I assume second harmonic suppression was a problem.

Jim

N2ADR

[James Ahlstrom]

Hello Group,

I am designing filters for HL2, but I am not sure about how much harmonic suppression I need.  I think I need up to 21 dB second and up to 26 dB third harmonic suppression.  But it depends on the band.  Also, if we use SMD inductors with their higher losses, we need more than 5 watts from the finals.  So do my numbers look OK given about 7 watts out from the AFT05s?  How much for each band?

Jim

N2ADR

[Graeme Jury]

Hello Jim,

I don't have numbers for the harmonic output of the AFT05 amplifiers and probably we won't get any final plots until the 5 current beta boards are built and running. I would guess similarly to you that -20 dB at 2nd harmonic would be OK due to the push pull amplifiers already giving some suppression so the 3rd and 5th harmonics will be the issue. The most difficult filters to design for will be the 60/40M and 30/20M ones which both will just sneak in with > 22 dB or so suppression on the second harmonic if you are using an elliptic function filter of 5 poles and > 35 dB as you go up in frequency. It seems easy to place a zero on the second harmonic of the higher frequency of the dual band sets but not so easy with the third harmonic. All the other bands would be fine with Chebyshev or even half wave LP filters and they do have the advantage of increasing attenuation with frequency and matching inductor values. I know from your previous posts that you have already been down this trail and will already have arranged a suitable filter set so it is really a matter of knowing for certain what the harmonic output of the final version is. For my own filter I have finished up with 5 pole Cauer filters both HP and LP except for the 1.8 MHz HP filter which needed to be a 7 pole one to get a sharp enough stopband transition while keeping a low (better thn 30 dB) return loss in the passband. My intention is to have the filter as a universal filter but designed primarily for Hermes-Lite i.e. 100x100 board with either I2C or 4 digital control lines capable of levels up to 10 watts and providing ptt out and programmable  delayed PA bias and facilities to pass back pwr and swr from an external amp via I2C. Over the top for most HL applications but as I said it is to be a general purpose one.

73, Graeme zl2apv

[in3otd]

Hello Jim,
there were some data on a standalone test PA in this thread and here enclosed are some measurements done on the current H-L v2b2; the PA supply was 9 V, with lower voltage the harmonics may be slightly worse (there were also some data on this somewhere).
There will surely be some variations between units and maybe w.r.t. the H-L v3, difficult to say how much.

73 de Claudio, IN3OTD /DK1CG

[James Ahlstrom]

Hello Graeme and Claudio,

Thank you for the valuable information.  I am still trying to make a simple filter for HL2 on a 5x10 cm board.  But I had to move to all SMD parts because of the SOICs, available space and parts availability.  But with SMD, the board could be at least partially assembled.  I plan to use all SMD capacitors and inductors for the filters.  All LPF will be optimized Chebyshev types.  There will be a single high pass filter with a cut off at 3 MHz which will be in-line at all times.  The T/R relay on the HL2 board will remain and will not be moved to the filter board.  The connection between HL2 and the filter will be RF2 and a ribbon cable at CN7.  So far this is just a paper napkin design, and it may change.

Jim

N2ADR

[Steve Haynal]

Hi Claudio,

Thanks for the simulation. Is this with QUCS? Can you share the scripts? How do you think Q will affect the passband at frequencies relatively far from the stopband? I am asking to get a better sense of filter losses when cascading 2 in series. The plot of the 10/12M filter is not encouraging for use with it always inline as it is showing significant loss at 20M even without Q being modeled.

73,

Steve

KF7O