Author Topic: RF design and layout [questions/feedback]  (Read 1719 times)

0 Members and 1 Guest are viewing this topic.

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
RF design and layout [questions/feedback]
« on: June 25, 2019, 09:03:34 am »
Hello! I'm designing a board with an LNA and filter and am looking for some feedback. I also have a question about ground filling the top layer; what distance should I observe from rf traces to ensure the same 50ohm impedance? Also, would a metal can help improve the performance of the board in a significant way?

Layout:
http://prntscr.com/o6c1q5
http://prntscr.com/o6c1wr
impedance calculations:
http://prntscr.com/o6c251
Thanks again!
« Last Edit: June 25, 2019, 09:13:30 am by aaronhance »
 

Offline Ysjoelfir

  • Frequent Contributor
  • **
  • Posts: 543
  • Country: de
Re: RF design and layout [questions/feedback]
« Reply #1 on: June 25, 2019, 09:07:15 am »
403 Forbidden.
Greetings, Kai \ Ysjoelfir
 
The following users thanked this post: aaronhance

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: RF design and layout [questions/feedback]
« Reply #2 on: June 25, 2019, 09:08:14 am »
Upload the images to the forum directly!  :scared:
 
The following users thanked this post: aaronhance

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
Re: RF design and layout [questions/feedback]
« Reply #3 on: June 25, 2019, 09:14:25 am »
Uploaded images,  |O
 

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: RF design and layout [questions/feedback]
« Reply #4 on: June 25, 2019, 09:17:56 am »
That does not look right at all.

What are you trying to achieve? What is your goal?
 

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
Re: RF design and layout [questions/feedback]
« Reply #5 on: June 25, 2019, 09:18:57 am »
Amplify the signal and FM band stop.
 

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: RF design and layout [questions/feedback]
« Reply #6 on: June 25, 2019, 09:26:41 am »
Still it doesn't look right.  Your filter assumes 50ohms on both sides. What is the signal source? Will it be always 50ohms? What is the load side, after the filter? Will that always be 50 ohms?

Your schematic is a complete mess, please redraw it with correct schematic symbols, so it has any value.

Also not sure what is your thought was behind using a USB-A host connector on a DEVICE side.

And please be more specific and thorough in your description! We can't help much, based on your angry spits of a couple words length.

 

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
Re: RF design and layout [questions/feedback]
« Reply #7 on: June 25, 2019, 09:27:59 am »
Input is an antenna, the output is RTL SDR, both around 50 ohms. I'm not angry. :)
 

Online hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
Re: RF design and layout [questions/feedback]
« Reply #8 on: June 25, 2019, 09:34:50 am »
Use the 'Coplanar Wave' option in the saturn calculator to work out the effect of encroaching ground plane.

Use plenty of vias to connect the two ground planes, 1/20 of a wavelength apart, and use somewhat irregular spacing instead of a uniform grid of vias.

Edit :- 1/20 is excessive actually, 1/8th is probably OK but with 1/20 along the coplanar traces. If you search around you will find out what is OK from people who do this all the time.
« Last Edit: June 25, 2019, 09:38:19 am by hendorog »
 
The following users thanked this post: aaronhance

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
Re: RF design and layout [questions/feedback]
« Reply #9 on: June 25, 2019, 09:42:32 am »
How far apart should they be for a wavelength of 2.18m, 137Mhz?
« Last Edit: June 25, 2019, 09:44:47 am by aaronhance »
 

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: RF design and layout [questions/feedback]
« Reply #10 on: June 25, 2019, 09:44:56 am »
Antenna is rarely close to 50ohms, but that is not as problematic, as it is connected to the amplifier input.

What is more problem, is that RTL SDR is really anything  but 50ohms. Mind you, the tuner was designed for 75ohm environment and the available shit-documentation for R820T2 tuner (if it even is the case with your dongle!) is everything but specific about input impedance of that tuner.

Output impedance of that integrated amplifier rarely is close to 50ohm, especially at these low frequencies.  So plug in the correct model of the amplifier in the RFSIM, and check again. Not only for port impedance variance, but also do a component value spread simulation. 

Your supplied values of +-2% capacitors and +-5% inductors is not very realistic I think.  Also, including some real Q values of inductors helps.

//EDIT: You should think also about swapping the filter first, amp second. Otherwise strong local signals will overload the amplifier, distort, mix, and the filter then not  helps much, as out of band signals will now become in-band due to inherent non-linearities of the overloaded amplifier.

« Last Edit: June 25, 2019, 09:47:10 am by Yansi »
 
The following users thanked this post: aaronhance

Online hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
Re: RF design and layout [questions/feedback]
« Reply #11 on: June 25, 2019, 09:47:58 am »
Is that the only wavelength which can hit the LNA -  What is preventing higher frequencies getting in?

This is outside my experience, but something else to bear in mind is your filter will be highly reflective for some frequencies and that might cause issues with the amp.
 

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: RF design and layout [questions/feedback]
« Reply #12 on: June 25, 2019, 09:53:30 am »
Yes, checking the amplifier stability is also important. ADL5544 does not seem to be one of the most stable ones, especially, when poking in the low frequency band. The datasheet even suggests adding external network to make it stable.

The filter shall be probably replaced with a diplexer, to become wideband 50ohm termination for the amplifier, and the other side also.

Input of the amplifier shall at least be low pass filtered to get the highest frequency crap out.
 

Online hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
Re: RF design and layout [questions/feedback]
« Reply #13 on: June 25, 2019, 10:09:24 am »
Yep, I wonder if it would be good enough to just move the filter to the input side.

Edit: Good page about via spacing here:
https://www.edn.com/electronics-blogs/the-practicing-instrumentation-engineer/4406491/Via-spacing-on-high-performance-PCBs
« Last Edit: June 25, 2019, 11:31:09 am by hendorog »
 
The following users thanked this post: aaronhance

Offline Nitrousoxide

  • Regular Contributor
  • *
  • Posts: 156
  • Country: au
Re: RF design and layout [questions/feedback]
« Reply #14 on: June 26, 2019, 05:38:05 pm »
It should also be noted that the noise figure of the total system will increase if you place the filter before the amplifier (I just thought I would mention it). It is good practice to place the amplifier as close as possible to the antennae since even transmission lines diminish the noise factor.

The output of the amplifier can be matched for maximum gain as it has no bearing on the noise performance of the amplifier. However, the input matching does affect noise performance. I would recommend looking at the datasheet and through the use of a Smith chart, plot circles for the desired noise figure and constant gain. It would be then possible to find the maximum possible gain for the desired noise figure (by the intersection of circles), and thus the required matching impedance which can then be realized by lumped or distributed elements (for sub 500MHz you can use lumped).

And yes! As previously mentioned you must check stability. Check Rollet's conditions to give you a good understanding. But, be careful to compute these with the operational s parameters (correct bias currents/voltages). If the system is unconditionally stable, then there is nothing to worry about. If not, then certain regions of impedance matching should be avoided.
« Last Edit: June 26, 2019, 05:40:50 pm by Nitrousoxide »
 
The following users thanked this post: aaronhance

Offline aaronhanceTopic starter

  • Regular Contributor
  • *
  • Posts: 51
  • Country: gb
  • Damit
    • aaronhance.me
Re: RF design and layout [questions/feedback]
« Reply #15 on: June 28, 2019, 02:23:58 pm »
By using a 500Ohm resistor in place of the 1.5kOhm resistor on the datasheet I've managed to make it unconditionally stable between 10Mhz and 6Ghz. Do I need to worry about stability above this? I also have S21 15dB and s11 -13dB @ 137Mhz, I assume this is OK? If I were to increase the value of the resistor to get better performance it would be conditionally stable, how would I calculate if it would be stable, if it's simply a matter of mismatching the impedance at the conditionally stable frequencies, how much mismatch would be needed?
« Last Edit: June 30, 2019, 09:10:55 am by aaronhance »
 

Offline Nitrousoxide

  • Regular Contributor
  • *
  • Posts: 156
  • Country: au
Re: RF design and layout [questions/feedback]
« Reply #16 on: July 04, 2019, 06:01:36 pm »
1. Do I need to worry about stability above this?

2. I also have S21 15dB and s11 -13dB @ 137Mhz, I assume this is OK?

3. If I were to increase the value of the resistor to get better performance it would be conditionally stable, how would I calculate if it would be stable, if it's simply a matter of mismatching the impedance at the conditionally stable frequencies, how much mismatch would be needed?

1: Yes. Unconditional stability over a certain range is only guaranteed... in that range. You never know that the amplifier may be unstable at frequencies that are significantly larger than 6Ghz. Now is it likely that it is? Probably not. However, if you find the amplifier outputting a power that is significantly lower than what you'd expect per bias current, then there is some oscillation occurring.

Obviously, the change (and inclusion, i don't see it in the original layout?) of the feedback network will affect the noise figure as the input reflection coefficient is now not solely dependant upon the device, but also the output matching network.

2: S21 (forward gain) looks like about 17.5dB. Considering you've decreased the attenuation (of the FB network) or increased the negative feedback the gain has dropped. Which is to be expected. S11 of -13 looks about right from the datasheet. Again, look at what the datasheet says because it will most likely give a good indication of performance.

3: Apologies as I'm slightly confused by your wording. Are you asking "How do I calculate under what conditions I get a conditionally stable system?" and "What are the tools I can use to do this?". If so:

Conditional stability occurs when the transistor does not behave unilaterally, which is the case with most real transistors. Depending upon how unilateral they are can dictate their stability performance.

There are a few steps for calculating stability, in the order they generally are:
1. Is the transistor unilateral, meaning that S12 = 0? If yes, then it is unconditionally stable. If not, then move on to see if we can ASSUME that the device is unilateral.
2. Calculate the unilateral figure of merit from the S parameters. It is acceptable to assume the device is unilateral if this figure of merit is less than 1dB (some people say 0.5, some say 1.5).  :-//
3. If the previous two steps fail, then the device is conditionally stable. It is then important to calculate Rollet's stability criterion for multiple frequencies. If K > 1 and abs(delta) < 1 then the device is stable at that frequency. If either of these conditions are violated. You have found where the device is unstable.
4. Use the formulae to plot the stability circles. Now, if |S11| < 1 and |S22| < 1 then the area defined by the circle boundaries that include the origin of the Smith chart is stable. If |S11| > 1 and |S22| > 1, the area bound by circles NOT including the origin is stable.

Now, how can we vary all this parametrically? This is where mathematics computing software saves the day. You will have to mathematically represent the circuit as such that you can parametrically vary the feedback resistance, thus varying the S parameters and in turn affecting the stability criterion.

Note: There are other stability measures, such as the mu-factor that give slightly more detailed information as to where the instability is occurring but are slightly harder to compute.
 
The following users thanked this post: aaronhance


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf