Author Topic: Open Source HW RF Signal Generator  (Read 44946 times)

0 Members and 1 Guest are viewing this topic.

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3789
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: Open Source HW RF Signal Generator
« Reply #75 on: February 02, 2020, 06:33:48 pm »
Was any part of this tested and verified independently? I am not a close follower of this thread, but I still have a bit of bad taste of it being just a quick slaptogether. Nor do I see a single point to use any duino crap within designs that even uses FPGAs. (simply this is not a piece of a children toy, so why even bother making it child friendly?)

How will you manage to make any RF shielding lids? A custom milled aluminium blocks?

Is there really none ESD protection on the USB?

Also for example, from another quick look at the PCB: Why not flipping the 1.5Ghz LPF vertically? Would allow for much shorter interconnects.

I think the layout is far from what I'd call done. (hence my bad taste of a quick slaptogether). But I stay being corrected.
« Last Edit: February 02, 2020, 06:38:14 pm by Yansi »
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #76 on: February 02, 2020, 07:53:15 pm »
This board is a combination of all the individual boards I previously made and which I verified to be working. The only thing I didn't test so far is the FPGA-Board and its control of the AD9910 I have here. Control of the chip via SPI-Bus with the main CPU is working reliably.
The changes I've made compared to the version I have here with the individual boards I've detailed in this post: https://www.eevblog.com/forum/oshw/open-source-hw-rf-signal-generator/msg2819470/#msg2819470

Regarding your other questions:
I don't have a CNC-Machine, but I've got a 3D-Printer and plan to use conductive filament (and/or copper-foil or conductive paint) to print RF-Shielding enclosures. The holes in the PCB have a diameter of 3.4mm, which allows using M3 screws to mount the shields.

USB ESD-Protection... Oops. :)
I'll add the USBLC6 to the USB-Lines. Used that chip in other projects before and have some in my parts-bin.

The location of the filters could be optimized, yes.
If they were a bit cheaper, I'd be using tiny SMD-Filters by Minicircuits. That would allow more more filters with better attenuation of the harmonics.

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3789
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: Open Source HW RF Signal Generator
« Reply #77 on: February 02, 2020, 08:30:46 pm »
I wouldn't trust the 3D printed bullshit (yes, I'm a skepticist, especially with these types of filaments). Get a proper alu case milled (it is not that much expensive, but sure not cheap), or I'd opt to optimize the layout to use off-the shelf tinned steel sheet shielding cases.

Yup, I also go through a lot of USBLC6, both -2SC6 and -4SC6 (2 or 4 way).

There are other manufacturers of monolithic filters. Have you tried looking thoroughly enough?  Johanson technology, Taiyo Yuden, Murata, Epcos/TDK, etc etc etc. Even Mouser stocks heaps of them, for a fraction of MiniCircuits cost.

Performance of these filters suck (stopband attenuation 30-40dB), but can be cascaded with acceptable insertion loss. Very cheap. Also look at aliexpress and other far east sources. Lots of these available.



 

Offline thm_w

  • Super Contributor
  • ***
  • Posts: 2826
  • Country: ca
Re: Open Source HW RF Signal Generator
« Reply #78 on: February 03, 2020, 06:54:16 am »
3D printing makes sense regardless of what your end result is, as it lets you test fit the dimensions. CNC setup time is not cheap compared to filament.

Not sure if anyone has tested 30% copper PLA, would be interesting to see. Its not electrically conductive across the material.
The actual surface conductive stuff, afaik, is all carbon based, which is not great for RF absorption compared to say, Cu or Al.

These might be better options over carbon, just a question of if its good enough and won't melt PLA:

https://www.mgchemicals.com/downloads/tds/tds-842wb-l.pdf
"Surface Resistance @ 50 μm 0.0020 Ω/sq"

https://www.caswellplating.com/copper-conductive-paint-4oz.html
"Good surface resistivity of <1 Ohm/sq at 1 mil dry film thickness. Attenuation: more than 75 dB from 30 MHz to 1.5 GHz."

https://www.mgchemicals.com/downloads/tds/tds-841ar-a.pdf
"Surface Resistance @ 50 μm 0.60Ω/sq"
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1234
  • Country: cn
  • RF Engineer. Discord: https://discord.gg/SYZ4WwH9Z
Re: Open Source HW RF Signal Generator
« Reply #79 on: February 03, 2020, 08:54:46 am »
Why not use shield cans and clips?
Something like this: https://www.aliexpress.com/item/33003768263.html
I'm sure you can find other dimensions or fold these from copper foil for custom dimensions. On taobao some sellers can do custom shield cans.

There are clips that solder to the PCB:
[attach=1]
Discord: https://discord.gg/SYZ4WwH9Zu
Email: OwOwOwOwO123@outlook.com
GitHub: gabriel-tenma-white
 
The following users thanked this post: thm_w

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #80 on: February 03, 2020, 06:05:50 pm »
Soldering these clips to the PCB should be possible. The surfaces free of soldermask are 5mm wide.

I was inspired to use 3D-Printed pieces by the video of TheSignalPath in which he 3D-Printed some horn antennas and Waveguide-Filters with pretty good results.

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #81 on: February 19, 2020, 07:10:35 pm »
I've done some research and the method to go (even used by at least one company I found that has invested some research in this area) seems to be coating the 3D-Printed objects with conductive paint or metallic foil.

Metallic foil is probably the best choice here price wise. Or cans out of copper sheet metal. :)


Regarding the filters: There's a ton of them available on Digikey and Mouser, but they all seem to be targeting the 2.4 GHz (WLAN)-Area. So they're outside the area of interest and when I have to order parts directly from the manufacturer, Minicircuits seems to be the better supplier.

Offline eb4fbz

  • Regular Contributor
  • *
  • Posts: 143
  • Country: es
Re: Open Source HW RF Signal Generator
« Reply #82 on: February 20, 2020, 09:41:47 am »
Those microstrip lines are pretty thin! What is the dielectric thickness?

You will have big mismatches due to discontinuities created by the components pads. Those lines look very thing even compared with 0402 passives!

I would never ever try again to make distributed elements filters on FR4. Not even for <1GHz. It's not about loss, but also uncontrollable variations between boards and ultimate rejection. I would use SMD low pass filters from Murata or Johanson. They are cheap, and performance will be better and painless compared to those filters on FR4.
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #83 on: February 20, 2020, 08:10:42 pm »
The thickness is 0.1mm (Controlled Impedance 4-Layer Board made by JLCPCB). Dielectric constant is 4.05.
Yes, I noticed the thickness discontinuities of the pads. Unfortunately, that's one of the compromises I had to make to keep cost low.

Regarding the filters: Okay, you guys convinced me to switch to some ready made SMD-Filters :)
It also seems like Mouser has recently added Minicircuits Filters to their inventory. They do not ship to Germany though. Hopefully that'll change in the near future. So I'll redesign the filter-section with the filters by Minicircuits and update the layout of the filter-section. Should save a lot of space as well.

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3789
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: Open Source HW RF Signal Generator
« Reply #84 on: February 20, 2020, 11:39:23 pm »
These thin transmission lines are nightmare to work with. Use different stackup or forget 50ohms. On a generic FR4 board, even controlled impedance, it will be from floor to bottom in its value.

What is also commonly done, is to use layer 1 for signal (as you have), then omit layer 2  (no copper under RF parts), layer 3 becomes RF GND, layer 4 other signals (supply, control signals).

That way you can get much thicker traces for 50R, which is welcome, as the etching tolerances won't bother that much. Also will enable you to obtain much higher impedances for distributed filter elements.
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1234
  • Country: cn
  • RF Engineer. Discord: https://discord.gg/SYZ4WwH9Z
Re: Open Source HW RF Signal Generator
« Reply #85 on: February 21, 2020, 04:07:33 am »
There is also the 0.2mm stackup which JLC offers and is de-facto standard for boards with RF.
Discord: https://discord.gg/SYZ4WwH9Zu
Email: OwOwOwOwO123@outlook.com
GitHub: gabriel-tenma-white
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #86 on: April 15, 2020, 12:14:35 am »
Quick update:

I've redone the Filter-Banks (added one Stage btw.) with SMD-Filters by Johanson since Mini Circuits-Components aren't sold to Germany by Mouser for whatever reason. 
Tiny Problem I ran into is the fact that these filters are only available for certain frequencies and I need a filter with a cutoff-frequency around 2.2 to 2.4 GHz. So that Filter needs to be made with a stripline-filter.
The preceding 750 MHz, as well as 1.2 GHz Filter and the succeeding 3GHz-Filters are made out of 3 Filters in series.

I'm also redesigning the RF-Path after the Filtering since the RF-Switches by Analog Devices required negative control voltages and are a bit on the expensive side. As is the ADF5355: 95€ + Taxes! The almost pin-compatible ADF4355 (One pin is an output instead of GND) is only 37€. I've put a 0603 resistor right next to the pin in question, which can be populated by 50 Ohms or 0 Ohms, depending on the chip that's being used.

In addition to that, I've switched to the 0.2mm Stackup, which makes the traces and pads look much more uniform :)

Thx for all the suggestions. Keep 'em coming. I have set the deadline to order boards to the end of may, so there's still time to get them implemented :)

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #87 on: April 18, 2020, 12:35:25 am »
I managed to find a suitable filter from TDK that is in stock at Mouser and started layout of the new filter-banks.

I also added RC-Filters to each control-input of the RF-Switches to isolate them from digital noise and isolate the CPU from RF-Noise (the RF-Switches are being driven directly by the CPU).

Offline eb4fbz

  • Regular Contributor
  • *
  • Posts: 143
  • Country: es
Re: Open Source HW RF Signal Generator
« Reply #88 on: April 18, 2020, 11:48:30 am »
Have you taken the RF switch isolation into account? That will limit the rejection, so maybe cascading 3 filters does not improve anything. These switches usually have 20-30dB isolation at most for >2GHz. Layout will have some leackage too, and it gets worse when transmission lines are not properly terminated, as happens at frequencies where your filters are rejecting (reflecting) the unwanted components.

I also recommend you to simulate the cascaded filters response including interconnecting transmission lines length (phase), as most filters rejection doesnt simply add up, and even its passband could be distorted due to adjacent filters impedance.
« Last Edit: April 18, 2020, 11:51:39 am by eb4fbz »
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #89 on: April 18, 2020, 12:05:15 pm »
No, I haven't done that.

The SP4T-Switches have an isolation of 18 dB at 4 to 6 GHz. So if I understand this correctly, the outputs of the RF-Switch that are turned off will have a -18 dBm signal present that will then be added up at the inputs of the last switch and be visible at the output of the filter-bank. So as an example, the 4 GHz and 6 Ghz-Harmonics will appear at the output with an attenuation of roughly 36dB + the insertion loss of the 6 dB-Filters, correct?
Cascading filters in this case won't make much sense then, if the signal can just travel along a different path :)

How can I simulate the filters? I have access to Sonnet Lite, but that's about it when it comes to RF simulation-software. Is there a way to put the filters in there as black boxes with pre-defined S-Parameters?

Offline eb4fbz

  • Regular Contributor
  • *
  • Posts: 143
  • Country: es
Re: Open Source HW RF Signal Generator
« Reply #90 on: April 18, 2020, 12:59:33 pm »
You are right. But in a SP4T there are 3 leackage paths! These will add at the output, and depending on each filter bank loss and phases, you could be limited at 30dB or even less maximum rejection at some frequencies.

To simulate cascaded filters, download the s2p files from manufacturer and you can use them on a free tool as QUCS or RFSim.
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #91 on: April 18, 2020, 04:23:37 pm »
Thanks for pointing that out. I've switched to the ADRF5250 SP5T-Switches. They are a lot more expensive at 7€ per piece, but also offer 46dB isolation at 4 to 6 GHz.

Simulation in RFSim looks good. The Filters are behaving like advertised and the transmission-lines don't distort their behavior too much. There's just a little bit of ripple near the cutoff-frequency.
I've set the variance for the impedance of the transmission-lines to +/- 10% btw.

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3167
  • Country: us
Re: Open Source HW RF Signal Generator
« Reply #92 on: May 03, 2020, 09:19:29 pm »
Please forgive my being lazy.   Where can I find a block diagram of the current design concept?

It seems to me that the obvious implementation is a pair of ADF4351s, low pass filters and a mixer plus ancillary bits.

Reg
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #93 on: May 05, 2020, 04:22:31 pm »
I don't know if I have posted an updated block diagram :)


But quick description is as follows:
- CPU communicates through an FPGA with AD9957 and ADF4355 (ADF4356, ADF5355 and ADF5356 are pin compatible btw.) to generate the signals.
- AD9957 is used up to 425 MHz and also supplies modulated signals
- ADF4355 is used from 425.01 MHz to 6 GHz (Max Amplitude and flatness might suffer above 3 GHz - It's all built on cheap FR4-Board to keep cost down)
- Signal from ADF4355 is LowPass-Filtered via a bank of filters that are switched in and out to filter harmonics
- Output-Amplitude is leveled by ALC-Circuits
- Digitally controlled Step-Attenuator controls output-power from -90 to +10 dBm (Higher output-power possible, but requires an amplifier-module connected to the PA Output and Input)

I did think about getting lower frequencies by mixing two ADF4355 Signals together, but using the AD9957 for lower frequencies saves components and PCB-Space to the tune of almost 100€ for one device. It saves not only a ADF-Chip, but also all the supporting circuitry, additional filters, Mixer and the PCB can stay within the 200x150mm frame (anything larger becomes expensive really fast).

The Step Attenuator isn't on the main board, since there are so many attenuator-modules available on ebay that fit the bill. Most are less than 30€ and having the attenuator as an external module helps with keeping the board-size within the mentioned 200x150mm.

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3789
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: Open Source HW RF Signal Generator
« Reply #94 on: May 05, 2020, 04:43:52 pm »
So you have a LF path to the DDS directly and then HF path, where you mix the DDS with just single ADF435x?

I don't think that way you can get all the unwanted mixing products away easily. You would nee lot of band pass filtering to do so.
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3167
  • Country: us
Re: Open Source HW RF Signal Generator
« Reply #95 on: May 05, 2020, 06:47:01 pm »
What are the ADF4355 filter specs?  That is pretty critical to overall performance.

Looking at the AD9957 spec, it seems to me you'd get better SPFD if you ran it in the 10-25 MHz range or at 75 MHz and mixed that with the ADF435x to get the desired output frequencies.

I can't see buying a 6 GHz ADF4355 and putting it on FR4.  Either spend the money on better PCB material or drop back to a cheaper chip.

A block diagram and a more detailed description of the design logic would be a big help.  This thread has meandered along for a long time and it's hard to tell where you are.

A OSHW/OSSW replacement for an HP 8648C would be where I would start.  100 kHz to 2.9 GHz with an emphasis on very clean output and tight power level control.  That covers the 2.4 GHz WiFi band and is still reasonable to do on FR4.  I paid ~$1000 for mine in the US.  I'd love to see something comparable under $150.  And at $100 it would be OK if it only went to 2.2 GHz.

Have Fun!
Reg
 

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #96 on: May 05, 2020, 09:16:40 pm »
I've drawn a block diagram (see attached picture and yEd-File).

The Filters for the ADF4355, which I have chosen over the 4351 because of its much finer frequency control (16bit Integer + 24 bit Fractional instead of 16bit INT + 12bit Fractional) - the 6 GHz is basically a bonus that comes with it, but can be left out by chosing the 10€ cheaper ADF4355-2BCPZ that can only go up to 4.4GHz, are:
- 750 MHz: Johanson 0500LP15A500E
- 1200 MHz: Johanson 0915LP15B026
- 2 GHz: TDK DEA202025LT-5052C1
- 3 GHz: Johanson 2450LP14A100T
- 6 GHz: Johanson 5515LP15A730

The Filter for the AD9957 is a Johanson 0400LP15A0122.

The current setup (AD9910 Module from ebay with lumped element LC-Filter) has basically no harmonics and just a few fixed spurs around 200 to 300 Mhz, which are most likely caused by some deficiencies on the module.
Approaching 400 MHz, a tiny amount of image frequency can be seen approaching 600 MHz, but its almost down in the noise. With the Johanson-Filter I expect this image frequency to be absent.

The signal of the ADF4351 has a harmonic at twice the fundamental frequency that is pretty consistent 33 dB weaker than the fundamental. Not quite the level of THD the HP 8648 is capable of, but keep in mind, this is the result of Microstrip-Filters on 0.1mm 2113 Prepreg FR4-Board with no shields over the filters. So adding shields and using the SMD-Filters, I'm guessing the result will be better, maybe even reaching -40dB.
The 3rd harmonic is mostly gone, except for frequencies around the switch-threshold of the filters (800 MHz, 1.5 GHz) - The finer granularity of the filters on the new Mainboard should suppress this third harmonic better.
Output-Power was about 0dBm each time I measured, which equates to an output-power of about +1 dBm at the output of the Filter-Board.

Once I'm done taking the measurements (and I found my USB Floppy-Drive) I'll post the measurements, comparing ADF4351 "RAW"-Output and the filtered one.

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #97 on: May 05, 2020, 10:43:27 pm »
Okay, so here are the measurements I just did with the Spectrum Analyzer first connected directly to the ADF4351 and then filtered, as well as the AD9910. I adjusted the ALC-Circuit to a Level-Control Voltage of 1.25V and 2V Bias-Voltage to avoid driving the amplifier that sits at the output of the filters into saturation.

Please note, the 410 MHz-Measurement is completely off, because the signal path is selected wrong (bug in the selection-routine that I can't fix without an extensive rewrite of the program at this point), but starting at 700 MHz, the correct filter is selected. You can see the effect of the 3 GHz Microstrip-Filter, however. :)
As you can see, second harmonic is down 38dB, while any higher harmonics are completely gone at 700 MHz - At this frequency the 750 MHz LPF by MiniCircuits is selected.
At 1.2 GHz the second harmonic is down 30 dB - Filtering is done by the 1.5 GHz Microstrip-Filter.

The filters on the new board will have more overlap, so especially at lower frequencies, the suppression of harmonics will be better.

The measurements of the AD9910 show a rather clean signal, except for the constant spurs that are appearing in the spectrum.
Since I don't see those in spectra I've found on the internet by other people, I'm guessing those come from external circuitry on the module I've got here (Cheap module from ebay with 1117 regulators and 40 MHz Oscillator), or the way I've programmed the internal PLL of the chip to generate the 1GHz Sample-Clock - There's quite a lot that can be configured and I have stopped fiddling with those values once I got a signal.
The image signal at Sample-Clock / 2 - Output-Frequency that appears around 350 MHz could be a problem, but the Johanson-Filters claim to have -30 dB attenuation at that frequency and I'm using 3 in series, so that should provide sufficient suppression of those signals.

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #98 on: May 06, 2020, 07:35:34 pm »
I've done some additional measurements after I've fixed the bug in the filter-selection.
To set the amplitude correctly, I've used the Automatic Level Control by setting the Level-Voltage to the required value and the Bias-Voltage to about 2 Volts via the DACs on the filter-board.
Starting at around 1.3 GHz the output-power is limited by the amplification on the analog-board and can't exceed -1 dBm anymore.

To look at the signal at frequency-ranges beyond 3.1 GHz, I've used the software SATSAGEN to turn my ADALM PLUTO into a Spectrum Analyzer.
Note: The signals around 3.2 GHz are always there and come from the ADALM PLUTO.

According to the manual of the HP 8648C, the spectral purity should have harmonics weaker than -30 dBc.
My current setup doesn't achieve these figures in all ranges. Especially the higher range, where just the 3 GHz Microstrip-Filter is active, shows significant harmonics.
Also of note is the fact that at 0 dBm output-power the amplifier is being overdriven and distorting the signal.
With the new design, the signal is routed either to the Attenuator or a power-amp/directly to the output, and the amplifier has its P1dB-Point at about 20 dBm instead of 16 dBm.

In the block-diagram included in the large picture, you can see the signal-chain I have currently set up.

Offline SaabFAN

  • Frequent Contributor
  • **
  • Posts: 733
  • Country: de
Re: Open Source HW RF Signal Generator
« Reply #99 on: May 07, 2020, 06:29:33 pm »
Does anyone here have RF generators by HP, Philips, Tektronix, Rigol, etc. and a Spectrum Analyzer to do similar measurements to compare the spectra of these generators to my prototype-setup?



Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf