Filters for Hermes-Lite

[ZL2APV]

Thought I would start a new thread for filters. Joe, your comment about a 30 MHz LP filter needed before any amplification I strongly agree with. The really nice 1206 inductors that John posted are running a Q in the 30's at the inductances required for 30 MHz filters. A bit on the low side but then losses are not important straight out of the HL as power is adjustable and subsequent amplification will sort it. Further down the amplification chain the losses are important and either air wound coils or larger toroids are necessary. At 5 watts T30's or better T37's will do and for 30 MHz I would be looking at 10 mix i.e.T37-10. A very good document on expected toroid Q's is here and you can see that expected Q's for 10 mix at inductances required for 30 MHz filters is around 150 which is pretty good.

Attached is a 5 pole Cauer filter plot via Elsie calculated with standard caps (78pF = 68//10 and 30pF = 15//15) T37-10 or T30-10 toroids (both have the same AL) with 10 turns and 11 turns respectively.

73 Graeme ZL2APV

[RHQQ2YXRKT]

 

John, what does the plot look like if you reduce the inductor Q ?

( what sim software do you use )

 

John G3UGY 

 

[ZL2APV]

Hello John,
It changes very little. Through loss (S21) is approx .25 dB at 10 mHz and 0.4 dB at 28 MHz so 1206 inductors are fine at low power. The out of band numbers hardly change.

The software used is "Elsie" by Jim Tonne and is free for student use and supports filters up to 7 poles which is enough for our needs.

I have attached the plot and schematic of the same filter adjusted to use the 1206 inductors and standard capacitors and as you can compare, this filter will work well for low power Tx and Rx use. Its about as good as you can get with a 5 pole filter.

73 Graeme ZL2APV

[John Williams]

Graeme, et.al.

For test purposes, I will place this filter on the frontend that I am modifying for use with 1.22 boards. I will use your T37-10 based filter for the 30MHz lowpass on the post-amp side. I still need a filter design for the BC High Pass.

I have 7 segment LPFs based on QRP-Labs values. Should I keep them or do you think  we can move to 5 pole filters like these and gain adequate performance?

See attached schematic...

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 Dick]

See the broadcast hi pass filter on this page: http://www.pan-tex.net/usr/r/receivers/elrtvfmbcftr.htm which gives some reference to its origin. Otherwise Elsie can easily be used to design the BCB hi pass

 

“Digital Steve”, K1RF

---

This email has been checked for viruses by Avast antivirus software. www.avast.com

[John Williams]

I found this BC filter that Graeme sent out a while ago. My main issue is how to build the inductors so they do not take up an inordinate amount of board space.

John

[John Williams]

Updated schematic with both of Graeme's filters...

[ZL2APV]

Hello List,

I made a careless error in my 1206 inductor version of the 30 MHz LP filter and used a Q 0f 50 instead of 30 which is the Q of the actual inductors. It made the loss at 28 MHz in the order of .65 dB but the rest changed little.

The B'cast Hi pass filter you have posted John, is the one I am currently using and have it in a little PCB box with a BNC on each end. I don't think it really needs to be as good as that one and I have designed with Elsie a 5 pole version which will do the job with one less inductor. Iron powder toroids needed too many turns for a lazy winder so I used ferrite ones from Phillips TN 9/6/3 4C65 which are approximately the same size as T37's. It may be a good idea to get a substitute core from an available supplier like "Kits and Parts" say an FT37-61 and tweak the design which I am happy to do. To give an idea of the approximate performance I have attached a shot of the Elsie plot. I was very surprised when I ran these filters through a VNA at how close the two matched and have a lot of confidence in Elsie predictions. As Steve pointed out it is easy to use Elsie for this type of filter. You can see from the plot that I used the optimizer to get the standard values and as a final step I run a Monte Carlo to see what the worst case could be for component tolerances and if OK I freeze the design.

John a couple of points on your circuit.

Is it necessary to include a relay on the Tx filter? I cannot think of a reason to take it out of circuit and if in the future HL is extended in frequency to say 50 MHz we would be re-designing and replacing the filter anyway. A good real estate saving possible here and one less relay to signal through.

It may be a good idea to fit a link so the antenna changeover relay can be bypassed as many users will want to drive an outboard amplifier and do the aerial switching at that point. Their requirement would be separate Rx and Tx lines but of course they have the option to not fit the relay and link across the contacts.

Perhaps "Band 0" could be the through connection and Band 1 to Band n, would cover the ham bands 160 to 10M etc. If that was the case the 74HC30 would not be needed. Both Quisk and ghpsdr3-alex would support this band switching configuration as you can enter your band settings in their config files.

73 Graeme ZL2APV

[John Williams]

Comments inline...

On 8/16/2015 3:58 PM, ZL2APV wrote:

Hello List,

I made a careless error in my 1206 inductor version of the 30 MHz LP filter and used a Q 0f 50 instead of 30 which is the Q of the actual inductors. It made the loss at 28 MHz in the order of .65 dB but the rest changed little.

The B'cast Hi pass filter you have posted John, is the one I am currently using and have it in a little PCB box with a BNC on each end. I don't think it really needs to be as good as that one and I have designed with Elsie a 5 pole version which will do the job with one less inductor. Iron powder toroids needed too many turns for a lazy winder so I used ferrite ones from Phillips TN 9/6/3 4C65 which are approximately the same size as T37's. It may be a good idea to get a substitute core from an available supplier like "Kits and Parts" say an FT37-61 and tweak the design which I am happy to do. To give an idea of the approximate performance I have attached a shot of the Elsie plot. I was very surprised when I ran these filters through a VNA at how close the two matched and have a lot of confidence in Elsie predictions. As Steve pointed out it is easy to use Elsie for this type of filter. You can see from the plot that I used the optimizer to get the standard values and as a final step I run a Monte Carlo to see what the worst case could be for component tolerances and if OK I freeze the design.

Will include it... I can convert the toroid...

John a couple of points on your circuit.

Is it necessary to include a relay on the Tx filter? I cannot think of a reason to take it out of circuit and if in the future HL is extended in frequency to say 50 MHz we would be re-designing and replacing the filter anyway. A good real estate saving possible here and one less relay to signal through.

What is the relay number that you are referring to. I do not understand your comment.

It may be a good idea to fit a link so the antenna changeover relay can be bypassed as many users will want to drive an outboard amplifier and do the aerial switching at that point. Their requirement would be separate Rx and Tx lines but of course they have the option to not fit the relay and link across the contacts.

Will see how to incorporate...

Perhaps "Band 0" could be the through connection and Band 1 to Band n, would cover the ham bands 160 to 10M etc. If that was the case the 74HC30 would not be needed. Both Quisk and ghpsdr3-alex would support this band switching configuration as you can enter your band settings in their config files.

I wanted a solution where if for whatever reason any band controls are not asserted, we would not transmit into no load condition. Is this overkill?

[ZL2APV]

Hi John,

I just knew you would see my post before I had corrected the comment. Forget all about the 30 MHz LP filter comments, I was having a brain fade and of course it is in circuit all the time.

The antenna changeover will be a little tricky as the LP filter is common. I would be inclined to do the linking so it stays in the TX line and the responsibility for Rx LP filtering would be with the outboard amp connection. Maybe as I have done the filter is inserted into the coax to the outboard circuitry.

Band 0 can be the default condition so if no relays are specified then band 0 is operated. I had a quick look at Quisk to see how this can be set up but was not obvious from a cursory look so will need to look harder and report back. I know ghpsdr3-alex definitely does this because I modified a little of Andrea Montefusco's code for auto filter switching and got familiar with the setup.

Sorry about the LP filter red herring will be more careful in future.

73 Graeme ZL2APV

[ZL2APV]

John, I found the Quisk config, it was in the Hermes directory and the pertinent section is ...

# Control the J16 connector according to the band.  If the band is not here, the default is 0x00.
Hermes_BandDict = {'160':0b0000001, '80':0b0000010, '60':0b0000011, '40':0b0000100, '30':0b0000100, '20':0b0000101, '17':0b0000101, '15':0b0000110, '12':0b0000111, '10':0b0000111}

As you can see I use a 3 of 8 decoder for the filters and also some of the filters are used across more than one band. If no filter is specified or there is no matching band button to the filter spec, the value of 0b0000000 will be used so you will always have a termination.

73, Graeme

[Steve Haynal]

Hello,

You've guys got some good discussion going on here. Some thoughts:

• Unlike v1.2x, v2.0 will have an onboard reconstruction filter that is used for TX. See https://github.com/softerhardware/Hermes-Lite/wiki/Hermes-Lite-2.0#lpf  I'm not convinced an additional LPF filter at 30 MHz is required on the companion PA, especially if the RX always uses on of the TX filter paths.
• There is quite a bit of overhead to include a BC HPF. Those who do WSPR at 0.136 and 0.4742 will not want it. What about just requiring a user to use one externally as Graeme already does if BC interference is a problem for them?
• The through path if no band is selected is a nice touch, but if you are going to add an IC, what about a full decoder for more bands?
• The tap with PE4259 looks good. Some of the not connected pins on the PE4259 will be 3.3V power.

Since there is interest and expertise on filters in this thread, perhaps I can persuade someone to help with the low pass reconstruction filter design for the main v2.0 board. I'm not sure that one has to be elliptical as a sharp cutoff is not required. I was thinking just a 5 pole Chebyshev. What I need help with is:

• Pick values and types for all filter components, both wind your own and surface mount
• Provide me with BOM for the filter items and build instructions
• Provide me with at least a sketch of what dual footprints (both SMT and wind your own) should look like with dimensions. Actual KiCad footprints would be better
• The above for the two baluns would be much appreciated extra help

73,

Steve

KF7O

[John Williams]

comments in this pen

On 8/17/2015 1:43 AM, Steve Haynal wrote:

Hello,

You've guys got some good discussion going on here. Some thoughts:

• Unlike v1.2x, v2.0 will have an onboard reconstruction filter that is used for TX. See https://github.com/softerhardware/Hermes-Lite/wiki/Hermes-Lite-2.0#lpf  I'm not convinced an additional LPF filter at 30 MHz is required on the companion PA, especially if the RX always uses on of the TX filter paths. - agreed
  
• There is quite a bit of overhead to include a BC HPF. Those who do WSPR at 0.136 and 0.4742 will not want it. What about just requiring a user to use one externally as Graeme already does if BC interference is a problem for them? - worried about board real estate so will leave off for now.
  
• The through path if no band is selected is a nice touch, but if you are going to add an IC, what about a full decoder for more bands? - please amplify. I am covering all bands now. Do you want a lpf for each band? The only shared bands are 17/15 and 12/10 which place harmonics well above the filter cutoff.
  

• The tap with PE4259 looks good. Some of the not connected pins on the PE4259 will be 3.3V power.

Since there is interest and expertise on filters in this thread, perhaps I can persuade someone to help with the low pass reconstruction filter design for the main v2.0 board. I'm not sure that one has to be elliptical as a sharp cutoff is not required. I was thinking just a 5 pole Chebyshev. What I need help with is:

• Pick values and types for all filter components, both wind your own and surface mount - We have a 5 pole filter using SMT and toroids.
  

• Provide me with BOM for the filter items and build instructions
• Provide me with at least a sketch of what dual footprints (both SMT and wind your own) should look like with dimensions. Actual KiCad footprints would be better
• The above for the two baluns would be much appreciated extra help

73,

Steve

KF7O

[James Ahlstrom]

Hello,

I worked a bit on this problem and found it tricky.  I will work some more, but I am unfortunately buried in other problems with Quisk.  So the best I can do now is throw out some thoughts.  If I can free some time, I will try to be more useful.

I like Graeme's use of Cauer filters.  With a little tweaking, the zero can be placed on the most troublesome image at 46 Mhz.  I wonder if just a 3 pole design would be enough.  That depends on measuring the images on all bands, and I will do this when I get time.  There is no reason to wind toroids for the filter.  Small chip or molded inductors are suitable for this low power level.

The output of the AD9866 IAMP is a current source, and so has an infinite impedance.  This means that designing a filter assuming a 50 ohm source is wrong, but more importantly it means that the reactance of the filter and its load can drive the AD9866 into non-linearity.  The AD9866 data sheet (page 31) shows a two transistor buffer to fix this.  I am worried about someone connecting a reactive load to Hermes-Lite and clipping the AD9866 output.

We are trying to get maximum power from the AD9866, but the data sheet (page 32) warns that this can overheat the chip.

There is a balun at the AD9866 output to perform the balanced to unbalanced conversion.  People can get the phasing wrong.  An alternative is using IOUTP followed by an opamp.  I admit that the extra cost is undesirable, but it eases construction and removes heat from the AD9866.

Jim

N2ADR

[Steve Haynal]

Hi John,

No, I am not really interested in covering more than 6 or 7 bands with a small companion PA. I actually prefer no logic on the companion PA and would give up one of the TX band paths as a through.

Fine business on the SMT components. 

73,

Steve

KF7O

[Steve Haynal]

Hi Jim,

Thanks for taking some time on this. Your very specific input is appreciated.

When do you see and how strong is your image at 46 MHz? Is this on 10M with a 73.728 MHz or 61.440 MHz oscillator?

So far, I've stayed away from the voltage-mode operation described on page 31 in favor of better linearity and lower power. We don't use IOUTG. I am interested in learning more. Version 1.2x was designed with a detachable frontend card with essentially direct connection to the AD9866 so that experiments with various frontends could be easy. There is even a PCB template to facilitate this. Experimental results an opamp to remove the balun or two transistors in voltage mode would be very helpful.

I am a wary and conservative about too much change in v2.0. Ideas like switching to SMT inductors or small variations in the frontend can be done as tweaks to the v2.0 layout for v2.1. Right now, I am shooting for a frontend very much like what has been used and proven in v1.22 plus reconstruction filter with the possibility of using dual SMT/wind-your-own footprints. I'll need hard convincing experimental evidence to do anything much different in v2.0.

73,

Steve

KF7O

[va7...@gmail.com]

Hi Graeme,

My updated test report, reflecting transmitter re-tests performed after the master oscillator replacement, is on my website:

http://www.ab4oj.com/sdr/hermes_lite/hl_notes.pdf

I tested the transmitter without harmonic filters, and then with the appropriate Mini-Circuits BLP-series LPF for each of the bands I tested. These filters bring harmonic content and spurs well below the limits specified in radio regulations.

73, Adam VA7OJ/AB4OJ

[ZL2APV]

Hi Adam,

Took a while to read and absorb that but not as long as you would have taken to create it. Brilliant document and puts a few things into perspective. It will be a huge help to all working on firming up a design and it would be great to think we could prevail on you for a repeat of the tests on the final version.

Many thanks,

Graeme ZL2APV

[ZL2APV]

Hi all,

Attached is some Quisk VNA plots of a sample of filters to give an idea of the performance to be expected. To have some idea of the pro's and con's of a practical realization may be a good aid to the decision making.

The first plot is of a 5 pole Chebyshev filter with T37-2 toroids in a 2 compartment enclosure. The out of band feed-through is better than 50 dB and the in band attenuation typically 0.25 dB with 2nd harmonic attenuation of 27 dB. Perfectly adequate for a Tx filter especially if a push pull amplifier is used. It is the top filter in the photo.

The next filter is a 5 pole Cauer built from ordinary Neosid 6mm slug tuned formers and again in compartments. The second harmonic is around 50 dB down and the 3rd about 42 dB. In band loss is high at around 1 dB due to low Q coils so this design would need to be made with toroids. I expect this filter would be very similar to filters made from smd inductors. Sorry I didn't have any to try.

The next filter is from QRP Labs and the filter itself is very good but in the board with the relay switches has a fair bit of blowby which gets worse with increasing frequency as you would expect and is around -37dB at 30 mHz. Still not too bad but could be improved with layout and screening. The filter was built for a particular radio and to get it to plug in, the input and output of the filter get very close together. I don't need the plugin facility so will cut the surplus tracks plus try bypassing the relay coils and switch lines to see if that helps. It raises the issue of tight board packing and lack of screening will impact on performance.

The last filter is a 7 pole Chebyshev 30 MHz low pass filter suitable for insertion into an aerial lead as a roofing filter. As you can see with enough input/output separation the signal does not couple through and the filter drops off with increasing frequency. I posted a 5 pole Cauer design previously which I have not built yet but it should be better than this one.

My VNA has a floor of -60 dB largely due to my using a changeover wafer switch to switch between the radio circuitry and the VNA's BNC sockets on the front panel. There is a little bit of spill over on the wafer switch but -60 dB is enough for the level of experimentation that I am doing. Quisk VNA is one of the greatest tools I have ever had. Every time I use it I say "thanks Jim".

The photos go top to bottom in the order I described the filters.

73 Graeme ZL2APV

[va7...@gmail.com]

Hi Graeme,

Glad you found my report helpful. Of course I will be more than happy to re-run the test suite on the final version. I plan to co-ordinate with Dave VE7PKE on this; I believe Dave will be able to loan me the final hardware platform.

73, Adam VA7OJ/AB4OJ

[James Ahlstrom]

Hello Group,

Here are some measurements on the Hermes-Lite version 1.22.  The clock is 73.728 MHz.  Measurements were made using Quisk with the "Spot" feature.  I used my Anritsu MS2670A spectrum analyzer to measure the output.  This is an old unit with about a 70 dB dynamic range, so it is not possible to see spurs below about -70 dBc.  If you want to repeat these measurements with Quisk, be sure your tx_level = {None:255, '60':255} in your config file is set to maximum output.  Right-click the Spot button and set to maximum output, and be sure to select SSB, because the digital modes are set to reduce power.

Shown below is the output in dBm for each band at the test frequency.  I also searched for spurs near the output and below the second harmonic.  The output is dBm, but the spurs are dBc; that is, below the output level.

Band      Freq       Output dBm

-------      ------       ----------------

160           1.90                18.9

  80           3.80                19.4

  60           5.50                19.7

  40           7.20                19.6

  30          10.11               19.6

  20          14.20               19.8

  17          18.13               19.6    Spurs: 20.56 MHz, -62 dB

  15          21.20               19.5    Spurs: 41.50 MHz, -49 dB

  12          24.90               19.6    Spurs: 22.97 MHz, -53 dB;   26.86 MHz, -62 dB;   48.84Mhz, -54 dB

  10          29.00               19.4    Spurs: 44.74 MHz, -49 dB;   55.55 MHz, -55 dB

I found the lack of DAC images surprising until I remembered that the DAC is set to interpolate by a factor of two.  The image (calculated without 2X interpolation) on 10 meters is at 44.74 MHz, and is 49 dB down.  This is only 44.74/29.00 or 1.5 times the fundamental, so it will be hard to filter.  Graeme's Cauer filters may do the job, especially if a zero can be placed at 45 MHz.  But the image is weak anyway, so maybe another 20 dB from a Chebyshev is good enough.

My homemade temperature indicator (my finger) did not sense any alarming temperature rise for the AD9866 during extended testing at full output.  I am still concerned about the problem of someone attaching a reactive load and clipping the AD9866 output.

Jim

N2ADR

[Steve Haynal]

Hi Adam,

Thanks for the updated report. It is full of information what will take me time to adsorb and understand. One thing that stands out and will need further explanation in a QEX article is:

"1. The receiver’s DR2/IP2 (2nd -order IMD) performance is inadequate for use without a good preselector in Region 1 or in other areas where high-power HFBC transmitters are present."

What is considered adequate and by how much are we missing? We have a bulk of our users in Region 1. What do they think of this statement given their experiences?

73,

Steve

KF7O

 

On Monday, August 17, 2015 at 6:14:43 PM UTC-7, 

[Steve Haynal]

Hi Jim,

Thanks for the measurements. I agree that the DAC interpolation helps cleanup the signal and simplifies any reconstruction filter. This is why I was hoping originally to get by with just the TX filters. Given the interpolation, we do see anticipated DAC spurs at 2*Fs - Ftx. Did you see anything at ~118 MHz when transmitting on 29 MHz? I made similar measurements here. My understanding of spurious emissions when transmitting below 30 MHz on amateur frequencies is that they must be 43 dB less than the fundamental in the US and 43 dB + 10log(P) or 50 dB less than the fundamental in ITU regions. It looks like we can meet this with a standalone Hermes-Lite and simple reconstruction LPF in the low 30 MHz range. The 44.74 spur you see on 10M (Fs/2 + (Fs/2 - Ftx)) was much worse with the old 61.440 MHz oscillator to the point that 10M was pretty much unusable.

This paper is a nice overview of ways to interface to a DAC. On page 4, they say:

"The transformer not only serves to convert the differential output into a single-ended signal, but it also isolates the output of the DAC from the reactive load presented by the LC filter, thereby improving overall distortion performance."

Do you think that adding a termination resistor to the TX output at the AD9866 pins (this will reduce power but make VNA users happy), the partial isolation by the transformer, and the partial isolation by the LPF filter (especially if lower Q components are used...) will be enough to mitigate you concerns regarding highly reactive loads and clipping the DAC output? Can you quantify and describe the worst case scenario you imagine?

73,

Steve

KF7O

[ZL2APV]

Hi Jim,

Thanks for the output levels and spurs data. Great help for designing filters as you can put the zeros in the right places. I have whipped up a filter for 30 MHz and put the notch at your test point of 44.74 MHz. The Elsie design showed a pretty good response so a filter was built with junk box caps (mixture of silver mica and polystyrene) and toroids wound on T37-6 cores. The Q's worked out at around 55 but the turns required were nothing like Mini Ring Core Calculator suggested e.g the 363 nH with the calculator was 11 turns but in practice came out at 8 turns fairly well spread around the core. The turn spread was sensitive and a bit tricky to adjust but the end result was a close match to Elsie. I think T37-10 cores would be a big help here. I have got a few add on's for the VNA which I will describe when the VNA part of the project is being tackled properly. Like Steve I am trying to avoid "feature creep".

[Steve Haynal]

Hi Graeme,

That is great that you can put the zero on the spur. It makes me wonder if we could have done that with the 61.440 MHz oscillator. Thanks for all the filter information you've been posting.

73,

Steve

KF7O

[Steve Haynal]

Hi Graeme,

Don't you think we can squeeze in an option for some sort of screen between the filters in the layout? One could use strips of copper clad boards provided there are some ground pins between filters.

I was looking at SMT transformers on www.coilcraft.com and noticed they have some free software that might aid understanding of what their chip inductors can do.

73,

Steve

KF7O

[ZL2APV]

Hi Steve,

Thanks for the link to Coilcraft. I downloaded their designer software and installed it with "wine" as I am Linux only here. Happy to say it worked perfectly. They make some lovely 1812 inductors with Q's around 80 which would make a relatively high Quality filter and probably good for 100 Watts. Their series resonant frequency is way over 100 MHz so no problems there either. The 0805 inductors predict an insertion loss of 1.8 dB at 29 MHz and 0.74 dB at 15 MHz which is getting up there. Unfortunately I can't see any way to use the 1812 inductors in the design but I will contact Coilcraft and see if they can help.

The layout that John is using is superior to that of the QRP Labs board and crosstalk or spillover will be less of a problem. I anticipate that the 0805 inductors will be the limiting factor and no shielding will be required but if it is PCB strips make good shields as you suggest.

73, Graeme ZL2APV

[John Williams]

Graeme,

These 1812 inductors would be smaller space consumers than the toroids for the higher freq filters, right? My statement on use of 805 was mostly towards caps and resistors.

In the layout, I am contemplating organizing the the filters first, and perhaps making a test board with 2 or 3 together to place under actual test conditions prior to committing to a entire board. Perhaps we could agree on a set of gerbers, and have elecrow or oshpark ship them to you for testing?

Thoughts?

John

[ZL2APV]

That would be my thoughts John. I too would be reluctant to commit to the whole PA board without testing the filters first. I would be anle use the prototype boards in another project. I think they make boards in batches of 5 but not sure as I roll my own. Of course 4 layer is not possible for me. I ask because if it is 5 boards and you do 3 filters on them I would like 2 of the boards. Have you designed the filters? Are you going with the QRP Labs design? Do you want me to make a copy of your pcb (double sided) and run a filter before you commit to buying any parts. I can find enough components except for the inductors to do a couple of filters without the relays. BTW running latest KiCad here. I will start straight in and do say a 20 metre filter.

73, Graeme

[John Williams]

Graeme,

I count 9 turns on the outside inductors and 7 turns on the middle one?  1 Turn = 1 pass thru the center, right?

John

--

[John Williams]

Updated schematic...since the turns are critical, would this be a good filter to test with the coilcraft 1812 inductors?

On 8/20/2015 12:31 AM, ZL2APV wrote:

--

[John Williams]

Graeme,

I hope to try a design this weekend, but at this point I do not have a pcb layout. I was planning to use the qrp-lab's layout for internal routing, but separate the inputs and outputs as much as possible. I ordered 20 relays by china-post on ebay. Were quite reasonable at 10 for $5 USD. You may want to get a china package coming your way. I think it is absolutely important to include the relays in the testing and have 3 filters together so we can judge interaction effects when we ground 2 and activate one.

 Thus, I want the complete TX and RX path thru the PTT relay, the always active 30MHz filter, and then 3 switchable filters to test. I will forgo placing any other components for now. That will make a good test board...

John

[James Ahlstrom]

Hi Steve,

I am continuing to work on this; that is, on a reconstruction filter to be used on-board at the output of the AD9866.

I see now that you are biasing the IAMP at 5.0 volts, and squeezing out 4.28 VPP at each IOUTN pin.  This results in 6.28 VPP at the output.  Very nice.  Substituting an opamp powered from 5 V would reduce output power.  The opamp has the advantage that it would eliminate the transformer and extend response to almost DC.  I am thinking here of the low frequency < 1MHz experimenters.  But on balance I can't justify a change to an opamp.  Bringing out the IOUTP pins in case someone want to work at VLF is good enough.  

I did not look for output higher than the second harmonic in my tests.  I assumed that the Tx filters would remove higher frequencies.  But that is incorrect.  A reconstruction filter needs to clean up the output so junk does not mix in subsequent amp stages prior to the final Tx filters.  I will look for output in the range 30-150 MHz when I get time.

I am not convinced that the transformer isolates the AD9866 as Analog claims, because of the high power levels used.

I want to look at a three pole Cauer filter with low Q made from standard 5% parts.  Maybe it is sufficient.

As far as clipping the output, I need to model everything with the filter in place, because it will be highly reactive above the cutoff.  Maybe a low Q filter and a small pad will be sufficient.  Back terminating the AD9866 as you suggest will lose a quick 6 dB of output (as you know).

Jim

N2ADR

[ZL2APV]

Hi Jim,

A 3 dB pad on the output would at least stop wild impedance excursions although a far from perfect termination. It could also be helped by doing a diplexer filter thus extending the termination up to the 100's of MHz.

Graeme, zl2apv

[John Williams]

We have a 3db pad on the input to the amp we are designing.

--

[ZL2APV]

Hi John,

Still in bed, will get back on the turns on the toroids.

Yes the 30 Mhz filter is the most challenging and now I believe it should be a diplexing filter. I have contacted Coilcraft and they have agreed to support me with their samples program and will allow me enough samples to send a set to you and Jim and Steve will need a set too so I will arrange enough capacitors for this as well. I will leave the boards and schematic in your hands.

I have not responded to an earlier question so yes Yaego capacitors have worked well in filters for me. I have always used COG caps at around 200 vw

Agreed about the relays. Could you send me the address of where to buy them from and I will get them under way today. I think the filter test board should contain all the filters as the effect of all the relay contacts on the common in and out lines need to be examined. This line should be a printed transmission line and Kicad has a tool for this. The highest frequency filter and the bypass relay should be at the far end of the line as the open line becomes a stub to the near end relays but it is insignificant if they are the low frequency filters.

Given Roger Rehr's post, to avoid tears in the future we need to look at supporting up to 31 MHz although it may be better for him to come down to 23 MHz. I'll query this on his thread.

I am really pleased to hear that you are prototyping a push-pull amplifier. I am doing the same here although sidetracked by filters etc. I am experimenting with pd85004's which work fine right through 6 metres and are grounded source (big advantage) but like you heat sinking is the issue. Maybe a tmp100 and a chip fan may be a good idea although passive heatsinking is my first choice.

Cheers, Graeme

[Graeme Jury]

Sorry Steve the spur is too close in with the 61.440 oscillator. You have to trade off too much of the rest of the filter performance to get a zero there. I'm going to give up on the 61.44 oscillator and get a replacement on order today.

73, Graeme zl2apv

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/CQAs7LGt4kM/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.

[John Williams]

Be glad to purchase one here and send it to you...

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.

[John Williams]

Graeme,

Before I get too far, I need to solidify the footprint for the coilcraft filters. I have been using the 5mmx10mm rectangle to allow a T37 toroid inserted vertically. For the lower freqs we may need to use T50 but hope that will still be adequate space. If we go to a mix, need to know by component/filter what you plan. Use my latest schematic and tell me which inductors will be impacted.

I will revise my plan and include all of the filters. I just started layout and a 4 layer board will simplify power and control line routing immensely.  See attachment... Only the first filter is complete except for the grounds. Pretty tight...  Feel free to suggest changes. Should I leave space to allow a PCB shield, for instance?

I will use Yageo COG caps but will switch to 200V instead of the 50V I have been using.

Get well soon...

John

[Duncan Clark]

Hi Folks,

Before one starts making filters using SMD inductors just have a look at
PA3AKE's site. As the filters will invariably be in line on Rx you run
the risk of generating IMD in those filters on Rx, especially on 40m
here in Europe.

http://martein.home.xs4all.nl/pa3ake/hmode/bpf_intro.html

I thought Martein did some early tests on various SMD, Toko, Lodestone
coils but I can't now find his data - suffice to say they were nowhere
good enough for his mixer IP3. I know we are not after his performance
but still worth thinking about.

Duncan
--
Duncan Clark
G4ELJ

[ZL2APV]

Hi Duncan,

Thats a really good summary of the compromises and pros and cons of the design. We have physical constraints with vertical height and board area which precludes air wound and forces toroids or smd inductors. We are going to do just the filters as an evaluation before committing to a full PA layout.

I have commented previously on how many smd caps are unsuitable for filters and brands which produce RF capacitors should be used. These capacitors can be quite expensive.

In my own personal case my 30 MHz filter uses air wound coils and polystyrene capacitors with each of the 3 sections in a compartment equal to 2 coil diameters by 2 coil lengths and BNC plugs on each end. Perfect in the aerial line but impractical in the proposed case.

We have to accept a certain level of non ideal construction in order to miniaturize the project but it will be kept compliant with regulations.

Thanks for that very useful url. We need all the info we can get at this stage.

73, Graeme zl2apv

[ZL2APV]

Hi John,

You really are packing it in, there isn't much empty space on that board.

Ideally at least the centre inductor should be at right angles to the ends. Preferably all inductors at right angles to each other.

When toroids are wound to occupy 75% of the ring, the leads would be on each side of the toroid with 1 side 25% from the top and the other side 25% from the bottom rather than directly opposite.

Running the common transmission lines feeding the relay inputs and outputs down the outsides of the relays will give very good separation.

I am not sure whether it is better to put a very short tap from each relay to the transmission line and keep it straight or to kink it to divert to each relay contact with zero length connector. Maybe someone on the list knows.

Before I know for sure which inductor to use I really need to try them. I am mindful of Duncan's warnings and don't want to blindly go ahead and use them just because they look good with a design tool and are a convenient fit on the board. Sorry more delays, I know you are keen to drive this along.

I would still go for a full set of filters so we can properly test for crosstalk and feedthrough plus input and output capacitors may need tweaking due to residual reactances when in circuit. I know I am introducing delays here but there is quite a bit to consider.

Can you confirm if you are using the 12 volt or the 5 volt versions of the relays? You mentioned relay currents in the twenties as I recall which I would expect from the 12 volt version but the schematic shows 5 volts.

I want to do a test to destruction by hot switching at 25 watts for (hopefully) thousands of operations. I'll program an Arduino to provide the switching and monitor the swr via the relay.

I am not sure if there is a price break on board size which has lead you to do 3 filters.


73, Graeme zl2apv

[John Williams]

Price is not as much a goal as a good test bed. I will do the needful...adding all of them now...

Thanks on the orientation advice, will build a revised footprint from the standard ones, and will orient at right angles. Glad I did a screenshot!

Am using 5V relays to start. That is what I purchased.

Send your address. Not taking no for an answer, will be sending a oscillator...

John - W9JSW

[ZL2APV]

On Friday, August 21, 2015 at 10:18:57 AM UTC+12, John Williams wrote:

Price is not as much a goal as a good test bed. I will do the needful...adding all of them now...

 

Great, this will be more realistic and give a better result.

Thanks on the orientation advice, will build a revised footprint from the standard ones, and will orient at right angles. Glad I did a screenshot!

If 1808's are the best choice I will easily fit them into the toroid space. They may be the best choice simply because there is no fiddling to get the right value which could be pretty difficult for many builders. There is more than performance to consider for a semi kit.
 

Am using 5V relays to start. That is what I purchased.

Ok I will get a set on the way now.

Send your address. Not taking no for an answer, will be sending a oscillator...

You are a good man John. Thank you, I will PM you shortly.


73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

Even with the 73.728 MHz oscillator, it may be hard to get a zero in the best place as the spur will move in frequency.

73,

Steve

KF7O

[James Ahlstrom]

Hi Steve,

Here are some more measurements for frequencies above 30 MHz.  Since the AD9866 is set for 2X interpolation, I looked for the spurs at 147.456 MHz +/- the Tx frequency.  Here are the amplitudes in dBm.  I set my spectrum analyzer to 146.456 center frequency, and used the Quisk Spot feature.  Clicking up the bands showed the spurs spreading out from the center.  Note that these are dBm, not dBc.

Band      Lower   Upper

160          -28        -29

  80          -22        -23

  60          -18        -20

  40          -15        -18

  30          -12        -16

  20            -8        -14

  17            -5        -12

  15            -3        -11

  12             0        -10

  10        +1.7        -10

I also looked around 73.728, but there is no consistent spur around this frequency, other than 10 meters.

Next I looked for spurs above 30 MHz that we should filter out.  I focused on frequencies from 30 MHz to 180 MHz.  This span showed many spurs and harmonics that were highly dependent on output level.  I tested with the Quisk Spot feature by changing the level between 1.000 and 0.707.  That is a 3 dB change.   Just looking at the screen, it was obvious that most extra "grass" disappeared with a 3 dB reduction in output level.  But here are some specific measurements.

Here is the level of harmonics for 40 meters tested at 7.2 MHz.

Harmonic                           1        2           3          4           5

dBm at 1.000 output     20.06     -37      -22.4      -47     -28.5

dBm at 0.707 output     17.03     -43      -36.1    <-50      <-50

Change                          -3.0        -6       -13.7         x    >21.5

We note the relative absence of even order harmonics, and the 10~20  dB reduction in the level of odd harmonics for a 3 dB reduction in drive level.  Although some may think the existence of harmonics in unsurprising, a DAC should not produce any harmonics.  The output should be just the fundamental and the expected spurs about the clock frequency.  These harmonics may be innocent for amateur Tx, but may be troublesome for a VNA application.

Unfortunately there are non-harmonic spurs too.  Here is an example from the 17 meter band tested at 18.13 MHz:

Frequency MHz         dBm at 1.000 output      dBm at 0.707 output     Change

           54.44                           -15                              -28                           -13

           55.65                           -35                              -38                             -3

           56.87                           -27                              -42                           -15

And another from the 10 meter band tested at 29 MHz:

           60.46                        -18.9                           -34.5                        -15.6

My feeling is that we are overdriving the AD9866 output.  Can anyone else confirm/deny these measurements, or provide additional information?  We could use more data.

If we want to reduce the AD9866 output level, a possibility is to change from a 1.4 turns ratio transformer to a 1:1 transformer.  That reduces the output by 3 dB.  A suitable transformer may be the 1465-1317-1-ND from DigiKey, $2.44 quantity one.  It has response down to 300 kHz.  In case you are shopping for RF transformers at DigiKey, they are not under "transformer", they are under "RF/IF and RFID / Balun".  Minicircuits has suitable parts too, but requires a special order.

Jim

N2ADR

[Steve Haynal]

Hi Jim,

Thanks for your measurements, suggestions and understanding of the issues. I agree that a 3 dB reduction in power is worth it to clean up "grass", especially for VNA use. Unfortunately, I don't have a spectrum analyzer and have to go to OIT to use their lab to do such measurements so can't add much more data.

There are a few more knobs available to us. Rset described on page 29 of the data sheet is set to 1.6K and can be changed to effect TX output. The basic frontend card has a 0 Ohm resistor as a place holder for a resistor to limit Ibias as described on page 31. 

The first versions of firmware had the gain settings hard coded in the RTL for a fixed 20 dbm out. When I added the ability to adjust the gain, I had to do some unexpected resetting of some registers to achieve the advertised 20 dbm as I was only seeing a maximum of 17 dbm. It would be interesting to see if the previous way of setting a fixed 20 dbm out was any cleaner and if I messed up with gain adjustment. Currently, to adjust the gain settings I adjust register 0x0a as seen on page 20 and also (unexpectedly) register 0x10 seen on page 21. There are additional settings for TX current in register 0x12 that I have not played with. One easy solution would be to just limit the max power in the firmware via how these registers are set.

Since you are a Python fan, you may also be interested in the RTL code to interface to the ADC. It is all in myhdl as I wanted to experiment with mhydl. The source is in Hermes-Lite/rtl/myhdl/ad9866.py.

Thanks also for the link to the digikey transformer. I've been looking at minicircuits, coil craft and some on digikey and mouser to have a SMT option. Do you have any opinions on which manufacturers are best?

73,

Steve

KF7O

[James Ahlstrom]

Hi Steve,

On Friday, August 21, 2015 at 11:59:53 AM UTC-4, Steve Haynal wrote:

Thanks also for the link to the digikey transformer. I've been looking at minicircuits, coil craft and some on digikey and mouser to have a SMT option. Do you have any opinions on which manufacturers are best?

I don't have an opinion on quality.  But covering 1.8 to 30 MHz is not challenging.   Going to lower frequencies becomes difficult, but will be requested by the VLF crowd.  The Pulse CX2041NLT goes to 50 kHz, but DigiKey does not stock it Q1.  The real issue is availability, especially overseas.  Adding a part to a DigiKey order beats a special order to Minicircuits, and also beats a special order for the binocular core.  The DigiKey part is the MABAES0060 by M/A-Com, and is their version of the ETC1-1T, a very common transformer indeed.  The PCB footprint could include through holes for a homebrew substitute, as you no doubt have planned.  Maybe someone from overseas could see what is available.  I am still thinking of that $100 order for a few forgotten parts.

Jim

N2ADR

[va7...@gmail.com]

Hi Steve,

From my experience with other SDR's,I have seen DR2 figures in the range 80 to 110 dB. I would say that 80 dB or higher DR2 is very good. For example: If our receiver noise floor is -118 dBm and DR2 = 80 dB, two interferers at 6.1 and 8.1 MHz and at a measured level of -38 dBm at the receiver input will drop a 14.2 MHz IMD product at the noise floor.  -38 dBm is S9 + 35 dB. Broadcast signals on the lower HF bands can easily attain or exceed such levels at night in Region 1.

With the measured 43 dB DR2 of the Hermes Lite, the IMD2 product will be at the noise floor with interferers at only -75 dB, i.e. one-half S-unit below S9. Most broadcast signals on the lower HF bands far exceed this level.

It is noteworthy that measured DR2 was 50 dB on the ELAD FDM-S2, which has a 16-bit ADC but no preselector, By contrast, the Perseus DR2 tested at 110 dB with the preselector out.

I believe that this area bears further investigation.

73, Adam VA7OJ/AB4OJ



On Wednesday, August 19, 2015 at 9:15:39 PM UTC-7, Steve Haynal wrote:

Hi Adam,

Thanks for the updated report. It is full of information what will take me time to adsorb and understand. One thing that stands out and will need further explanation in a QEX article is:

"1. The receiver’s DR2/IP2 (2nd -order IMD) performance is inadequate for use without a good preselector in Region 1 or in other areas where high-power HFBC transmitters are present."

What is considered adequate and by how much are we missing? We have a bulk of our users in Region 1. What do they think of this statement given their experiences?

73,

Steve

KF7O

 

On Monday, August 17, 2015 at 6:14:43 PM UTC-7, 

Hi Graeme,

My updated test report, reflecting transmitter re-tests performed after the master oscillator replacement, is on my website:

http://www.ab4oj.com/sdr/hermes_lite/hl_notes.pdf

I tested the transmitter without harmonic filters, and then with the appropriate Mini-Circuits BLP-series LPF for each of the bands I tested. These filters bring harmonic content and spurs well below the limits specified in radio regulations.

73, Adam VA7OJ/AB4OJ

[in3otd]

Hello Adam,
thanks for the interesting report. I would like to know if, in your experience, the IM2 products levels for direct-sampling RXs usually follow the expected power law or if they behave more similar to the IM3 products, being not-so-clearly dependent on the input levels. It will be interesting to see a graph of the IM2 products amplitude vs. input power, as done for the IM3 products. This might explain some of the huge differences between the different direct-sampling RX tested.

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Jim,
the datsheet for the MABAES0060 transformer from Digikey says "Absolute Maximum RF Power : 250 mW" so it might be a little too close to the 100 mW+ from the Hermes Lite, might need to check its distortion at that power level. I guess this limit its due to the small core size.
Digikey and Mouser have in stock a core from EPCOS/TDK, B62152A7X1, which seems similar to the BN61-2402. It's not a finished balun, but might be useful to avoid to have to buy just a couple of cores from other sources, for those who do not have something suitable already in their junk box, hi.

On another subject, in a previous email you expressed some concerns about the DAC output filters being driven by a current source; if the filters were designed to be terminated by 50 ohm at the input and output, this will surely change their response and from a quick simulation this should cause the actual filter bandwidth to become somewhat larger and more important to have a peak just before the filter corner frequency. In theory filters can be designed to have the proper response when driven by a current source (singly-terminated filters), this will give the correct response and avoid the peaking. A filter designed in this way will also limit the excursions of the impedance presented to the current source.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Claudio,

On Sunday, August 23, 2015 at 9:13:52 AM UTC-4, in3otd wrote:

the datsheet for the MABAES0060 transformer from Digikey says "Absolute Maximum RF Power : 250 mW"

Thank you for pointing this out.

On another subject, in a previous email you expressed some concerns about the DAC output filters being driven by a current source; if the filters were designed to be terminated by 50 ohm at the input and output,

Yes, you are right.  The filter can be designed for a current source, but the input capacitor becomes very small.  I just designed a filter that assumes 150 ohms in parallel with the current source.  That stabilizes the impedance, and doesn't waste much power.

What do you think about using 50 ohms in parallel with the current source?  That loss is 3 dB but the filter is stable, and it has the advantage that the output impedance becomes 50 ohms.  That would be an advantage if the Hermes-Lite is used as a VNA, a signal source or a transverter front end.

Jim

N2ADR

[John Williams]

Graeme, et.al.

Here is the first pass of the board. I still have to take an overall look at how the components are laid out around the filters. Very tight. I tried to keep RX separate from TX as much as possible. It helps that the ground plane is internal, so that bottom components are isolated from the top filters.

Please be very critical on comments. We need to get this right...

I uploaded to Oshpark to get the purple views. The 3 board cost is $124. Would like to find a less expensive board house if that is possible. It may not be for a 80x100 4 layer board.

Could not find a transmission line tool in Kicad. Previously I calculated a 50ohm line on a 2 layer board as 2.4mm wide. So that is what I used. Do not know if it is correct on a 4 layer board. Will try to recalculate it.

John - W9JSW

[John Williams]

Folks,

Some minor updates on labeling...

As to intent, this is a test board. No effort was made to try to make it
up with the to-be-designed v2 board. Thus, interconnect is ribbon cables
and molex connectors to Steve's test frontend board. The RX and TX can
connect with molex 2 pin plugs on rg-174 coax. I use these all the time
and they work very well. Easy to make. The rest can be soldered ribbon,
or molex cables.

If you are looking for a gerber viewer, I use GerberLogix on Win8. It
has a free version that works very well.

It should fit Steve's suggested 80x100 box. Not planning to put it in a
box yet.

That box is 10.99 at MCM - http://www.mcmelectronics.com/product/21-14472

John

[in3otd]

Hello Jim,
yes, the singly-terminated filter will have smaller caps, but still with reasonable values, IMHO; see enclosed schematics and responses compared for the proposed 30 MHz LPF and a more-or-less equivalent singly-terminated version.

I agree that terminating the current DAC will be a cleaner solution, but I think there were some doubts whether there was still enough power to fully drive the proposed PA in this case.

73 de Claudio, IN3OTD / DK1CG

[ZL2APV]

Hi John,

The transmission line tool is available from the Kicad main project window. Second to last tool button is "PCB Calculator" from there choose "TransLine" tab and you will be presented with several choices, the simplest being the Microstrip line but you may be able to fit the coplaner wave guide with ground plane with the 4 layer board if you are using ground as the layer next to the top layout substrate. The diagrams on the tool page are self explanatory and the board manufacturer will be able to supply the substrate parameters.

I'll spend quality time today looking over the board design.

73, Graeme zl2apv

[ZL2APV]

Hi Claudio,

You may recall I got an amplifier board from you for your "Push-pull power amplifier with PD85004" project which I have been experimenting with for an output amplifier for the Hermes-Lite. This amplifier would be sensitive enough for the doubly terminated input as you of all people in the world would know. I have hardly progressed with it so far as my time seems to be spread too thinly but terminating the DAC issue looks like it will force this forward. The board that John Williams is doing already has facilities for a 3 dB pad on its input so if the termination was on the DAC it would simply need a zero ohm resistor on his pad.

It looks like the Hermes-Lite board may contain the DAC terminated and LP filtered producing around +13 dBm but even 100 mW should be enough as at this stage we can design for what ever it takes with the exception of the amplifier board John is doing and he will need to comment on the power input requirements.

Glad you have joined our list, your input will be very valuable.

73, Graeme zl2apv

[ZL2APV]

Hi John,

What are you doing to heatsink the RD16HHF1? I have built a Mobo project which used one and at 5 watts out the bias needed to be around 600 mA. The supplied heatsink was totally inadequate and I had to Fan cool it. With your packing constraints I foresee problems here. Maybe the FET could be fixed to a heat spreader which slides against the case as the board slides in and then a couple of CS screws could pull the spreader tight to the case so it can become the heatsink. This in itself could produce issues of oscillator stability although I doubt it and would not expect big temperature cycles of the case between Tx and Rx.

Thank you for acknowledging me with my call sign on your copyright notice but my input was really just ham chatter about your project which is what we do, it is our hobby. I did not lay a single pad on the board or draw a line on the schematic so feel that my call should be removed. You are the one who put in the hours and thought and effort.

When you look at the microstrip line see if you can get some info on whether it is best to run the line as a straight line with little 2 or 3 mm spurs to each relay contact which at that frequency should hardly show a discontinuity or whether it is best to lay the line as already on the board with a non-straight line to clear the adjacent relay pins. I don't know the answer to this.

The cost of the boards made me gulp. Claudio, IN3OTD has used http://www.elecrow.com/blog/elecrow-pcba-service/ which could be worth a look. He may be able to provide further comment on this manufacturer.

OK now for some breakfast and back to looking over the board.

73, Graeme zl2apv

[John Greusel]

John,

The board looks great. I think the ribbon cable is fine.

Take a look with this 3D viewer:

John

KC9OJV

3D Gerber Viewer | Mayhew Labs

				3D Gerber Viewer | Mayhew Labs This tool is free for anyone to use. If you find it useful please consider donating to help us continue work on it. Your support will go directly to funding
				View on mayhewlabs.com	Preview by Yahoo

---

From: John Williams <jswi...@gmail.com>
To: herme...@googlegroups.com
Sent: Sunday, August 23, 2015 9:54 AM
Subject: Re: Filters for Hermes-Lite - Some practical filters

--
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+unsub...@googlegroups.com.

[John Greusel]

seeedstudio.com is very competitive on their board prices although I've never ordered a 4 layer board.

John

KC9OJV

Fusion PCB, customize PCB prototype | Seeedstudio

		Fusion PCB, customize PCB prototype | Seeedstudio $9.9 for 5 pieces. Seeedstudio offer 5 pieces - 100 pieces PCB fabrication service. Lead time are 5-7 days.
		View on www.seeedstudio.com	Preview by Yahoo

---

From: ZL2APV <gvj...@gmail.com>
To: Hermes-Lite <herme...@googlegroups.com>
Sent: Sunday, August 23, 2015 11:47 AM

Subject: Re: Filters for Hermes-Lite - Some practical filters

[ZL2APV]

That's a huge price improvement at around $50 for 5 boards going down to $6 each in lots of 20. I guess you are happy with the quality of the 2 layer boards.

73, Graeme zl2apv

...

[John Greusel]

----- Forwarded Message -----
From: John Greusel <gre...@sbcglobal.net>
To: ZL2APV <gvj...@gmail.com>
Sent: Sunday, August 23, 2015 1:46 PM
Subject: Re: Filters for Hermes-Lite - Some practical filters

Yes, I would say their board quality is premium Chinese PCB quality.

Not quite to US domestic quality but very good- at least that's their 2 layer board output.

Their service with troubleshooting is good too. If you have a size that deviates "slightly" they'll let you do a custom charge and not not bump you to the next larger category.

John

KC9OJV

---

From: ZL2APV <gvj...@gmail.com>
To: Hermes-Lite <herme...@googlegroups.com>

Cc: gre...@sbcglobal.net
Sent: Sunday, August 23, 2015 1:40 PM

[Steve Haynal]

I did a price survey ~4 months ago and www.elecrow.com came out the best: http://www.elecrow.com/10pcs-4-layer-pcb-p-1178.html

Currently, $49.90 for 10 10x10 4 layer boards from elecrow compared to $59.90 from seeedstudio.

I used elecrow for the 100+ 4 layer Hermes-Lite boards I have sold and have yet to see a PCB complaint. If you use expedited shipping (DHL from Shenzhen), your order will arrive in 9-10 days. I think all of these outlets (seeedstudio, iteadstudio, elecrow) are in Shenzhen and may be middle men for the same PCB house.

73,

Steve

KF7O

On Sunday, August 23, 2015 at 11:50:58 AM UTC-7, John Greusel wrote:

Sent: Sunday, August 23, 2015 1:46 PM
Subject: Re: Filters for Hermes-Lite - Some practical filters

Yes, I would say their board quality is premium Chinese PCB quality.

Not quite to US domestic quality but very good- at least that's their 2 layer board output.

Their service with troubleshooting is good too. If you have a size that deviates "slightly" they'll let you do a custom charge and not not bump you to the next larger category.

John

KC9OJV

[John Williams]

I will investigate. Sounds much better.

[John Williams]

Graeme,

Your work on filters will get you comfortable with the inclusion of your call. This is not my strength.

I am going to log issues in github for each item brought forward. Will start tomorrow. Watching golf today.

My plan is to work obvious issues for a while, then go for boards when we agree to move forward. Will then order ten, so more can participate.

John

[John Williams]

On the heatsink, was planning to use the case. Also worry about adequate heat transfer. On my amp, I cut a access hole and mounted a black anodised heat sink. Worked well.

John

[John Williams]

My board makes 4 to 4.5 watts on the current 1.21/22 Hermes lite with a 3db pad in place, depending on band.

John - W9JSW

--

[Steve Haynal]

Hi Adam,

Thanks for the explanation. Although I am sure your measurements are accurate and properly done, I feel that there is something missing regarding how a Hermes-Lite would actually be used. For example, in an environment with so many strong HF signals, one would lower the LNA gain to avoid clipping. With good antennas and at certain times of day, the ADC will continuously clip at +19 dB gain even in region 2. The default LNA gain for Quisk is lower at +10 dB. I realize this is a trade off between MSD and DR2, but the point is that it is possible and what would probably happen. I don't think MSD will go down by 10 dB if the LNA is reduced by 10 dB. Perhaps a graph of DR2 versus LNA gain would be helpful.

Second, most users will connect the radio to an antenna or ATU which is tuned to the frequencies of interest. This will provide some attenuation of the signals at for example 6.1 and 8.1 MHz. I often use the ZM-2 Z-Match ATU with my Hermes-Lite and find that I can turn the LNA gain well above 20 dB without seeing ADC clipping. This effectively makes a preselector and is why I think the Hermes-Lite receiver outperformed 16-bit SDRs as described here.

Third, a low DR2 may not amount to much real interference. For someone in region 1, does this mean 2-3 spots, a dozen, or hundreds of frequencies with interference on 20M on an active night? I'd appreciate reports from those in region 1. If it is just 2-3 spots, then I have more interference problems from my LCD monitors.

Fourth, this result should still prompt us to improve the Hermes-Lite. Our frontend balun/filter may be contributing to the problem as described here. Although I need to think about this more, perhaps the firmware is contributing somehow to the problem. We should consider band pass filters on the PA/filter companion card. This could be 4 HPFs  (15, 8, 4, 2MHz, etc) switched by Peregrine devices that are active only on the RX path. Or this could be as simple as providing plans for a user to build their own external HPF(s) as ZL2APV has described on this list for AM broadcast stations.

At the end of the day, the Hermes-Lite has 3 fewer effective number of bits when compared to most 16-bit SDRs. It isn't hard to show in a lab situation that it is more limited, and I don't think that should be the focus of a QEX article. The more interesting question to me is how much does a user give up when using a Hermes-Lite receiver? Is his or her experience only 50% of what it would be with a 16-bit SDR? Above 90%? My belief is that you realize more than 95% of what you would with a 16-bit SDR, and in some cases the -12 to +48 integrated LNA provides advantages over out-of-the box 16-bit SDRs. This is good value for something that costs less than 20% of most 16-bit SDRs.

73,

Steve

KF7O

[John Williams]

I would note that the receive path is separate from the transmit path
between the planned v2 Hermes Lite and the proposed PA. It is my
suggestion that if a HPF/BPF is desired for RX, it can be created as a
separate third board. It can be stacked in the enclosure, with
appropriate cabling for RX path insertion, as well as sharing control
lines for filter selection. It seems to me that there is no need nor
desire to pack any more filters onto what is effectively a transmit board.

Just my 2 cents... I would be happy to design such a board.

John - W9JSW

[John Williams]

Graeme,

I ran the simple approach using the existing stacking of signal on top,
power, ground and then signal. The total board is 1.6mm thick. Thus the
gap between each layer is around .5mm and the gap between top and the
ground is approx. 1mm. This gives a microstrip width of 1.85 for a Zo of
50 ohms. Length does not vary it much, and there are no interfering
power lines crossing the signal line. If I swap power and ground, the
width becomes .91mm. Should I stay with the wider trace or swap to get
the narrower trace?

I do not see a way to model straight with stubs vs daisy chaining as it
is now...

Latest gerbers and schematic attached.

John

[John Williams]

Now revised with help from Steve Dick - K1RF. Have signals on top, then
ground, then power and lastly control signals. Revised microstrip traces
and all interconnects for filters to .91mm which equates to Zo of 50
ohms for the board parameters from elecrow. Here are the gerbers...

I also opened up 3 issues on github to track amp issues.

If anyone has enclosure suggestions with heatsink options, pls share. I
found one company that has some nice enclosures but the sourcing is not
very good (attached). I would like to find a 100mmx160mm with a
heatsink on the top. This would allow the V2 and the amp board to ride
in one slot and have an optional upper slot for a HPF/BPF board. And of
course the cost has to be reasonable ... hihi

73's

John - W9JSW

[James Ahlstrom]

Hello Group,

Here is the schematic and an Elsie data file of the filter I have been working with.  It is a Chebyshev 5 pole design with an added zero.  Here are my thoughts on the design.

A 5 pole Chebyshev has limited attenuation at 45 MHz.  To get rid of the image at 45 MHz requires a Cauer filter or some other filter with a notch at 45 MHz.  The trouble with a 5th order Cauer is that attenuation is limited above 100 MHz, and it can be fiddly to get the two resonators to the right frequency.  This filter has a single notch at 45 MHz, and the final three sections have a monotonic increase in attenuation above 60 MHz.  Values at the moment are 150 ohms, 68p, 330n || 37.9p, 150p, 390n and 150p.  The inductor Q is 40.  I propose building it next to the AD9866 with small chip parts.  I ordered some leaded parts to test it.

The complete design has a termination at the AD9866 of either 150 or 50 ohms, a 1:1 transformer, and a change in the center tap voltage from 5.0 to 3.3 volts.  The parts values are subject to change depending on whether 150 or 50 is chosen as the termination, and whether the ripple in the passband is too large.  The change to 3.3 volts means the main Hermes-Lite board only needs a single 3.3 volt supply.  The expected power loss is roughly -3dB for a 50 ohm termination, -3 dB for the 1:1 transformer, a -2 dB for the filter loss.

The transformer is a suitable SMD, and the board has through holes for a homebrew unit.  Two capacitors are shown for the trap, and I hope the standard board layout will mean no tuning is required.

Jim

N2ADR

[ZL2APV]

Hi John,

You have been pretty busy! Yes you have the layers right with the ground under the signal tracks and the power under the control tracks and your microstrip line design looks good. When I think about it the relay contacts are a spur off the transmission line in themselves so it probably doesn't matter much.

I guess you have been reading Jim's post on his post DAC filter research and I think it has some implications for your design in as much as you will be seeing a very clean signal which is a luxury not usually enjoyed by amplifiers at your level. What it means is the constraints on your filters are eased a bit and probably 5 pole filters will do the trick. It would need to be verified but I would expect the raw output of your amplifier biased at around 500 mA to have harmonics not exceeding -20 dBc assuming a clean input from the DAC/Filter on Hermes-Lite board. To keep all spurious outputs below -55 dBc you only need 35 dB of out of band attenuation from the filters.

I did a quick experiment this morning and built a 17/15 Metre band filter with 2 x T37-6 toroids and junk box capacitors in a little strip of matrix board and plugged it into the QRP Labs filter board. Had to peel a turn off each toroid and away it went with screen shot attached you can see that 36 MHz is 33 dB down and 42 Mhz is 40 dB down.

I think The 5 pole Cauer's are good enough and are easy to design to keep all high out of band products to -40 dB. Note that Jim's design constraints are different to yours as he is trying to notch out an unwanted product and make everything else get lesser as it moves out from the cut off frequency whereas you need to keep everything above cutoff at least 30 dB down assuming none of your IMD's are greater than -20 dBc.

The heat sink problem is going to be a bit of an issue. I think those cases you presented look expensive and I tend to favor bolting some heatsink to the case. A heat spreader made from a cube of aluminium equal to the distance from PCB to underside of the case would allow it to be screwed hard against the case top after the PCB has been slid in and the screws could go through an external heatsink as well. Having the RX board on the bottom and the TX board on top might be best.
a) Heat dissipation would be best from the top.
b) The power and ground layers of the PA board would provide some screening of the PA from the HL board.

Oh yes The Q's of the T37-6 toroids worked out to be much better than 80 which I first measured. Not sure what went wrong there but will investigate.

73 Graeme zl2apv

[John Williams]

Graeme,

I would like to move forward on purchasing boards as appropriate. I plan
to test the relay logic in a day or so. The China relays arrived today
(as well as your oscillator). Want to make sure the filter controls and
bias controls, even simple, are correct. And I want to insure the
footprints I built for the relays are correct. Then a go ahead from you
on a starting filter layout is all I need to press the button.

John - W9JSW

[ZL2APV]

Hello John,

I have done another filter for 30/20 Metre bands and overlayed the Elsie design so you can clearly see the tradeoffs..

The Pros and Cons of Using the Cauer vs the Chebyshev (QRP Labs) filters.
Pros
1. Insertion loss is better in Cauer i.e. ).3 dB vs 0.8 dB. In real terms with your amplifier producing 5 Watts you will get into your dummy load 4.65 watts with the Cauer and 4 watts with the Chebyshev.
2. The second harmonic attenuation of 10.1 MHz signal is better with -29 for Cauer and -20 for Chebyshev. -20 is marginal.
3. Only 2 inductors needed for Cauer vs 3 for Chebyshev.
4. It is obvious to see if the zero frequencies are correct.

Cons
1. The second harmonic attenuation of 14.0 MHz signal is worse with -34 for Cauer and -47 for Chebyshev. Both acceptable.
2. Third and greater Harmonics cannot be guaranteed to be better than - 40 dBc
3. The designs are not proven on a fully working PA board.
4. Best results for the filters need the zero frequencies adjusted by spreading the turns on the toroid and not all constructors will have equipment to do this

Some factors to consider
1. The input from the DAC and filter is not yet finished and plotted so the outputs from your amplifier is to some extent unknown.
2. Both Chebyshev and Cauer benefit from tuning the inductors accurately and we should work on a simple method to do this.
3. If the Hermes-Lite VNA was working it would be a big help for lower band filters
4. The aade inductance meter gives surprisingly good results even though it is operating at around 10 kHz.

I guess it is crunch time and I'll be chicken and suggest that you try to lay out the filters so that you can install either. The advantage is that whichever is the best can be accommodated but the disadvantage is that you cannot realize the gains from the physically smaller Cauer if you don't totally commit to it. either way I will spend whatever time it takes to get them going the best they can.

Great on the oscillator :-) it sure has been holding me back. Your relays have arrived very quickly, and mine are not due for another 2 weeks so I will be behind there.

73 Graeme zl2apv

[Steve Haynal]

Hi Jim,

Nice work! I like the idea of eliminating 5V. What are your thoughts on the worse OIP3 when Vcm is 3.3V as seen in Figure 42, page 18, of the AD9866 datasheet?

I like that a 150 Ohm termination will result in the same ballpark power output from the AD9866 as before, the more common 1:1 balun, and your matching LPF.

I sent you the extra frontend board this morning and look forward to hearing about how this works.

73,

Steve

KF7O

[Paul Phillips]

Hi John,

You probably have come across these boxes.  I don't know if they are high enough (53mm) but they have slots for 100 x 160 boards and have one side detatchable so that a heatsink could be mounted.  Available from Amazon and others:

http://www.amazon.com/gp/product/B004RIKZ6A/ref=pd_lpo_sbs_dp_ss_3?pf_rd_p=1944687742&pf_rd_s=lpo-top-stripe-1&pf_rd_t=201&pf_rd_i=B00TJ49K4U&pf_rd_m=ATVPDKIKX0DER&pf_rd_r=12F9G0YHRXFST7NCNJWH

I've used similar ones and they are nice looking boxes.

73,

Paul

G4KZY

[John Williams]

Very nice! Will check out further. Thanks so much!

John

[John Williams]

Graeme,

Is it ok to just plan to jumper one of the inductors in the filter and
not populate the corresponding 2 caps in order to build the 5 pole cauer
vs the 7 pole filter? That way we can test both and then decide which
way to go. On the next revision we can then remove the extra places if
necessary?

Comments please?

John

[John Williams]

I have a few extra superband amplifiers laying around. I can send one to
you in this oscillator package so you can view the raw output of the
amplifier on each band and see how the harmonics occur. I have one that
is wired passthru on one leg so you can test all bands without filters.
That would give you a worst case feel for the raw amp output.

More data is good...

John

On 8/24/2015 11:32 PM, ZL2APV wrote:

[James Ahlstrom]

Hi Steve,

On Tuesday, August 25, 2015 at 12:49:23 AM UTC-4, Steve Haynal wrote:

Hi Jim,

Nice work! I like the idea of eliminating 5V. What are your thoughts on the worse OIP3 when Vcm is 3.3V as seen in Figure 42, page 18, of the AD9866 datasheet?

My thought is that 3.3 volts might not work, and we will have to test it.  I have found the AD9866 data sheet somewhat unhelpful compare to other data sheets from AD.

Jim

N2ADR

[Alan Hopper]

Just a thought,

are we sure that any issues on output are all being caused by the dac and output circuit? Could the up conversion be causing any of the spurs.  I have no reason to believe it is but also don't know that it isn't.

73 

Alan M6NNB

[ZL2APV]

Hi John,

I think that would be a good idea. I am OK for components - balun, rd16hhf1's etc. so just bare board will do. I was going to make a 2 layer board for testing but this will be much better. I have a 100 mm wide case with slides and will try a few heat sink ideas with it.

Your proposal for a MKI board where we jumper out any unused components in the filters sounds good. Are you doing through hole or surface mount or maybe surface mount with a through hole on the pads?

73, Graeme

[John Williams]

The amp is already built, ready to plug into the pci-e slot, with heatsink attached. Bias set to around 350ma. I chose to jumper 1 band and fill the other with a 17/15 bpf. It is on a 1.1 board. I have a new 1.3 board so this older board is surplus to my needs. Might as well make use of it.

I need you to tell me which filters need a smt pad and what size it will be (assume 1812?). Guessing that it will only be an option for higher frequencies. For now, all inductors are 5x10mm rectangles with thru holes on each opposing corner. Take a look at the latest gerbers.

John

[John Williams]

Package on the way....

On 8/25/2015 11:53 AM, ZL2APV wrote:

[John Williams]

Graeme,

I tried to add the SMD 1812 pad to the existing layout and it significantly complicates things.

We can go a few ways here -

1. Change the filter to SMD and add thru holes
2. Leave the thru hole outline and place pigtails on the smd inductors to try them out.
   

A complicating factor is I have the optional caps in the center of the toroid e.g. C39 thru C44. If we change to SMD, I have to move them elsewhere. There is not much room elsewhere. If we use 5 pole and SMD then space will work out. Staying at 7 pole and using SMD is going to be tough.

I am leaning towards option 2 and not changing the layout until we know where we are going here.

Thoughts?

John

On 8/25/2015 11:53 AM, ZL2APV wrote:

[John Williams]

Steve,

One thought when you design the frontend board, can you provide the capability for a balanced output? If we go to a push pull amplifier, it would alleviate having to add a transformer on the PA board. After we get some test boards running, I plan to focus on testing the bread-boarded push-pull PA..

John

[Steve Haynal]

Hi John,

I'll have to think about this. V1.22 has the balanced output available... As a build option, it may be possible to leave off the balun and LPF on v2.0 and jumper the balanced output to the connector. I'd have to keep the ground separate. You'd have to put the reconstruction filter on your board, or someone could design a balanced filter for reconstruction for use before the balun or push/pull PA feed.

73,

Steve

KF7O  

[James Ahlstrom]

Hello Group,

I built the reconstruction filter on a front end board that Steve sent me.  Take a look at the new picture in the gallery.  The filter is made from leaded parts on a ground plane.  The final filter will be built SMD.  Here are some early results:

I measured the filter response on my spectrum analyzer with a tracking generator.  It is fairly flat to a -3 dB point at 33 MHz, has a null of -52dB at 47 MHz, then attenuation decreases to -38 dB at 56 MHz, then the attenuation increases until 130 MHz.  At 145 MHz there is an abrupt loss of attenuation.  The response is about as expected, except for the anomalous behavior at 145 MHz which is due to open wiring and leaded parts.

The output of Hermes-Lite with the filter in place and a 150 ohm back termination is:

160    14.6 dBm

  80    14.5

  60    14.4

  40    14.2

  30    13.9

  20    14.0

  17    14.4

  15    14.6

  10    13.3

The 10m spur at 60.46 MHz that was -18.9 dBm is now -52 dBm.  All but one of the 17m spurs are gone, except for 54.4 MHz which is -52 dBm.  The spectrum looks fairly clean at lower frequencies, but there is still some grass that disappears when the drive level is reduced by another 1.6 dB.  This grass is harmonics, so I am not concerned.  At higher frequencies the spectrum is clean up to 100 MHz.

The images at 147.456 MHz +/- the output frequency are still visible.  At 29 MHz, the spectrum analyzer shows a peak at 119.5 MHz (really 118.5) of -22 dBm.  It was measured without the filter at +1.7 dBm.  But the predicted level is 1.7 dBm less predicted filter attenuation of 53 dB equals -51 dBm.

I think the filter is working as designed and should be suitable, provided it is built using VHF/UHF practices.  I think the problems at VHF are due to the open wiring and the 6 cm distance between the AD9866 and the filter.

Even if we can get great isolation, the predicted image at 118.5 MHz is only -64 dBc after the reconstruction filter, and this is marginal.  There is no problem adding another two poles to the reconstruction filter.  But the power amp will generate VHF harmonics too, and these will not be attenuated much by the low pass transmit filters because of the open wiring and relay switching.  There should be a VHF/UHF filter permanently in line.  I think I remember that the Hermes had this.  With a VHF/UHF transmit filter I think we are OK.

There is still the issue of whether the output is sufficient to drive the final amp, or whether another stage is needed.

Jim

N2ADR

[Glenn P]

I am concerned about the LPF using single ended PA.

I am not a RF design engineer (or any sort of Eng.) but single ended amps generally have high levels of 2nd harmonic.  To get a feel for this I found a reference
https://books.google.com.au/books?id=cjRv6SSqSDoC&pg=PA132&lpg=PA132&dq=single+ended+pa+harmonics&source=bl&ots=ZcFkJbL-AK&sig=xJXkatw2Ni3cV4fNq1IIgYslIr8&hl=en&sa=X&ved=0CCsQ6AEwAjgKahUKEwjQrb24s8_HAhXIsJQKHbcXCPM#v=onepage&q=single%20ended%20pa%20harmonics&f=false

which states that singled ended 2nd harmonic is only 6dB down typically.

Some of the LPF are dual banded.   I used RFsim99 to simulate the 15/17M LPF.  Worst case here will be the 17M band.  If I checked the 2nd harmonic of 18.0MHz then the filter is only down about -23dB at 2nd harmonic (36MHz).   (This does not take into account spread of values etc which could mean a slightly better or worse result.) In this case 2nd harmonic at the output is now only -23dB.  The results of the 30MHz LPF following does little to help the situation, rolling off <1dB or so at 36MHz.

It would appear to me that a strong case could be mounted to use the push pull final amp.

glenn
vk3pe



On Sunday, August 16, 2015 at 10:01:16 AM UTC+10, ZL2APV wrote:

Thought I would start a new thread for filters. Joe, your comment about a 30 MHz LP filter needed before any amplification I strongly agree with. The really nice 1206 inductors that John posted are running a Q in the 30's at the inductances required for 30 MHz filters. A bit on the low side but then losses are not important straight out of the HL as power is adjustable and subsequent amplification will sort it. Further down the amplification chain the losses are important and either air wound coils or larger toroids are necessary. At 5 watts T30's or better T37's will do and for 30 MHz I would be looking at 10 mix i.e.T37-10. A very good document on expected toroid Q's is here and you can see that expected Q's for 10 mix at inductances required for 30 MHz filters is around 150 which is pretty good.

Attached is a 5 pole Cauer filter plot via Elsie calculated with standard caps (78pF = 68//10 and 30pF = 15//15) T37-10 or T30-10 toroids (both have the same AL) with 10 turns and 11 turns respectively.

73 Graeme ZL2APV

[John Williams]

I do not disagree with your comments.  I will let Graeme comment further.

We know that a push-pull amp will be superior here and I hope to move in that direction. However, experiments are planned while we are testing to see how difficult it is to build a cauer filter with a null at the second harmonic. Further, as can be gleaned from Jim N2ADR there are other harmonics that will have to be removed. Filter performance with a Cauer filter after the null will have to still be pretty good to insure the amp does not misbehave into the VHF/UHF range.

It is my plan to stay with the current single mosfet amp while we examine the filters. It may very well prove that this amp can only support 7 bands instead of 10 if the single device amp remains.

I have a 2 mosfet amp on the bench to test with, using RD16HHF1 components. To translate that onto the 80x100mm board may prove very difficult, esp. if we have to add another stage of pre-amplification. Graeme is modeling an amp using a board mounted heatsink. That amp may prove to be a better choice to conserve space but my result in a set of heat related issues that can be hard to solve.

We shall see... Part of the fun of this hobby is the experimentation.

John - W9JSW

--

[James Ahlstrom]

Hello John,

The measurements of the AD9866 show that it has very low values of even harmonics.  Of course, subsequent amplification will reintroduce even harmonics.

The design of Cauer filters with a null at the second and third harmonic was described by James L. Tonne in an article "Harmonic Filters, Improved"  in QEX in September 1998, page 50.  He has published software to calculate these improved filters, and it is available at http://www.tonnesoftware.com/optlowpass.html.  I have not build these filters, but I think they could be just what we need at the output of the power amp.

No Cauer filter has good attenuation at VHF, and no filter at all has good attenuation at VHF unless it is build with VHF design practices.  Think of the old TVI filters we used to use to see how to build one.

Jim

N2ADR

[in3otd]

Hello Glenn,
only 6 dB down seems a bit pessimistic; I have updated my web page where I did some measurements on the single-ended PA initially proposed here some months ago, which is very similar to the current one (except maybe for the output network). I have added a couple of graphs comparing the harmonics levels of this PA and of the QRP-PA2008, which is again similar to the proposed push-pull (its driver stage does not contribute significantly to the overall distortion), measured at 10 MHz.
Scroll down to the last two graphs at http://www.qsl.net/in3otd/ham_radio/AC9HY_RD16HHF1_PA/AC9HY_RD16HHF1_PA.html ; you will see that the single-ended PA 2nd harmonic will go to -14 dBc max, if you push it hard enough.
The harmonics levels of the QRP-PA2008 are quite frequency dependent not sure if because of mismatched devices, layout, transformer or what. Did not have time to investigate and to measure the single-ended-PA over frequency.

73 de Claudio, IN3OTD

[Steve Haynal]

Hi Jim,

Thanks for prototyping this and making the measurements. Did you notice any difference with Vcm at 3.3 instead of the recommended 4.8 V? I'd like to understand if there is any practical compromise with going to 3.3V. 

If we are to add an additional drive stage, I'd prefer something that will work off of the 12V to 13.8V power supply. The DL2EWN 5W PA is very similar to what John is doing but includes a drive stage. Pascal uses this with his Hermes-Lite. I believe it only takes 7-10 dBm to drive.

Your earlier measurements in this thread showed that the worst spur between 30 MHz and 100 MHz was the 44.74 MHz at -49 dBc and above 100 MHz was 146.456-10M TX at 1.7 dBm or -17.7 dBc. My understanding of the regulations for this output power (20 dBm) and given we are transmitting below 30 MHz is that 43 dB suppression of harmonics is sufficient for all ITU regions. I measured the 146.46-10M TX harmonic with the 5W PA and found it to be -43dBc. See this. This leads me to believe that a filter providing at least 30 dB at 146.456-10M TX and something less than that below 100 MHz would be sufficient.  This doesn't seem too hard to do and maybe we shouldn't use a Caurer filter here for better VHF suppression. You appear to be more concerned about this filter than me, and I am wondering if I am missing something.

What do you think of using a TVS diode on the DAC output to protect from extremes as we do for the RX input? One could replace the termination resistor with this and gain ~3 dB. 

73,

Steve

KF7O

[James Ahlstrom]

Hi Steve,

I still have to study the 3.3 versus 5.0 volt Vcm.  I want to do IMD measurements to make sure distortion doesn't increase.  Stay tuned...

On Sunday, August 30, 2015 at 1:39:45 PM UTC-4, Steve Haynal wrote:

If we are to add an additional drive stage,

I we use the RD16HHF1 device, it has 16 Watts output and 16 dB gain.   The 16 Watts is 42 dBm, minus 16 dB means we need 26 dBm of drive.  We never had that, so a drive stage is required.  Even the smaller RD06HHF1 at 6 Watts and 16 dB gain needs 22 dBm.  I agree it should run from 12 volts.

Your earlier measurements in this thread showed that the worst spur between 30 MHz and 100 MHz was the 44.74 MHz at -49 dBc and above 100 MHz was 146.456-10M TX at 1.7 dBm or -17.7 dBc. My understanding of the regulations for this output power (20 dBm) and given we are transmitting below 30 MHz is that 43 dB suppression of harmonics is sufficient for all ITU regions.

 

There are two issues here, the reconstruction filter at the output of the AD9866, and the low pass transmit filters.  In both cases we are required to meet the spurious emission standard and also to follow good engineering practice.  FCC Part 97.307(c) says "All spurious emissions from a station transmitter must be reduced to the greatest extent practicable."  Also, in my humble opinion, non-harmonic spurs are more objectionable than harmonics.

There is a known spur at 44.74 Megahertz, and no possibility of removing it with a transmit LPF.  Thus the reconstruction filter with a notch at the spur.  It has more spur attenuation than even a 7-order Chebyshev.  And the parts are cheap and small.  As for the lack of attenuation at 146.456-10M TX (which is 117 MHz), I hope the final SMD version fixes this.  Or we can let the Tx LP filters deal with it.

With the 44.74 MHz spur gone, we have only a normal requirement for Tx LPF.  But experience shows that open wiring and relay switched filters have poor attenuation at VHF and above.  The solution is to make a VHF filter with a cutoff of 35 MHz or so, and leave it in-line at all times.  A VHF filter means a filter that looks like and is constructed like a two meter filter.  If we want a VHF filter anyway, there is no point in worrying about the spur at 117 MHz because the Tx filter will take care of it.  I am sure the HPSDR Hermes used this design, but I can't seem to find the reference.

What do you think of using a TVS diode on the DAC output to protect from extremes as we do for the RX input? One could replace the termination resistor with this and gain ~3 dB. 

The 150 ohm resister also stabilizes the input impedance of the filter.  It is not just for protection.  I don't think 3 dB is worth it.  We could use a higher value of resister, but I don't think the extra 1.5 dB is worth it either.  I really would prefer a 50 ohm back termination, so that the Hermes-Lite would have a proper 50 ohm output impedance when used alone as a test instrument.  Then we would lose a further 3.5 dB.  None of this loss matters if we add a drive stage.  I really think a 50 ohm back termination should be considered seriously.

Jim

N2ADR

[John Williams]

Jim,

As an aside, this RD16HHF1 circuit came from Jim Veatch WA2EUJ (Hardrock 50 designer from HobbyPCB), who altered some of the components for us to get 20db of gain. See attached pdf.

We are using this component to achieve 5W of power, not 16. Thus, we need +17db from the Hermes Lite to achieve this. In practice we have been getting +20dbm of gain. We added a 3db pad to insure a 50 ohm input to the PA. We are getting around 4-4.5W of power out now. If we remove the attenuator on the PA, we can drop to 50mW of power and still get 4-5W of power. It has never been our intent to get the full output power out of the mosfet, mainly because due to the inefficiency of a class A design, we would generate a lot of heat.

I agree that we may need to add a driver stage. I am very interested to see how your circuit tests out on Steve's frontend to see if we need to add it. Are you a proponent of the RD06HHF1 as the driver or is there another approach? Perhaps a PD85004 MOSFET?

John - W9JSW

[John Williams]

Attached is the pre-v2 circuit with a driver design using the PD85004 from Claudio IN3OTD...lots of good data in this link.

http://www.qsl.net/in3otd/ham_radio/PD85004_SE_PA/PD85004_SE_PA.html

Should be easy to breadboard a small circuit and add it in line for testing. I may even crank out a small test board to give it a try. Would try to use all SMT inductors except for the big one.

John

[ZL2APV]

At this stage everyone has been working with the expectation of at least +17 dBm from the Hermes-Lite board and have experimented accordingly. Everything you have done has produced worthwhile unwanted emissions improvements but will leave the other work being done short of RF input. As has been pointed out an amplifier will fix this so the question is should the amplifier go onto the Hermes-Lite board with pads to give good 50 ohm terminations at +17 dBm or should the output of the Hermes-Lite be revised down to say +10 or +13 dBm and the amplification provided on the RF boards.

I hate revising scope but in this case the improvements are significant enough for a reconsideration. Steve's work won't be affected, you are taking care of the HL filtering so its a case of do you or John or anyone else do the pre-amplification and where does it go?


73 Graeme zl2apv

[John Williams]

What is the best choice given the use as a vna? I was pretty much thinking the pre-amp would go on the PA board.

John

--

[ZL2APV]

Yes there is no argument that push pull will have lower second harmonics. It is more a question of is the single ended amplifier good enough? Here are some thoughts ...

An almost identical amplifier was used in the Softrock/Mobo project and extensively tested at 5 Watts out and around 600 mA of bias. 2nd harmonic emissions were typically -20 dBc and well documented and verified by independent testing.
John already has built successful boards based on the proposed amplifier so we already have a working model.
The amplifier is very robust and will withstand open and short circuit loads which is ideal for developing filters etc.
The cost and ease of construction is low.
It is important to try to stay within scope and look to a mark 1 then a mark 2 version rather than build the final version on try number 1. It can be done but usually needs access to labs and resources we don't have.

A push pull version is planned for the and it will connect to filters and peripherals which are already built and tested on the MKI board which will still work and be perfectly OK even with subsequent designs arriving on the scene.

73, Graeme zl2apv