Breadboarding at RF

[jgalak]

I am working on an RF project and have a 40MHz signal (it'll eventually be in the 140MHz range, I'm keeping it low for now since my scope is only 100MHz).  The signal currently goes from PCB, via coax, to the scope, and all is relatively well.

But now I want to insert a filter into the signal.  I don't have a coax inline filter, so want to breadboard something.  I have a couple of BNC-to-test lead adaptors, and tried to put the filter together on a breadboard and thus insert into the signal path.  The result is a disaster - both the input to the filter, and the output of the filter are horribly distorted.  A square-ish wave now looks like a jagged triangular wave.

My question is general - is there a way to use a breadboard at these frequencies?  Or should I just give it up as a bad idea and make my prototypes more with more permanent techniques (dead bug, varo board, etc.)? 

[TheUnnamedNewbie]

One of the biggest causes of problems on breadboards are the inductance of long leads. However, if you keep your leads short that can be, at least in part, reduced. In addition, remember that loop area is a key factor for inductance - keep the return path as close to the signal path as possible!

I think you should be able to do something on a breadboard to supprisingly high frequencies if you keep that in mind.

However, the stray capacitance will mess with your filter, so it will shift in performance (insertion loss, passband, etc).

Another thing that helps is using SMA connectors to right-angle board plugs such as these: https://www.amphenolrf.com/132136rp.html

I've found on some breadboards, with some tweaking of the connector (filing down the edges of the pins so the are a bit thinner) and some strategic use of sticky tape I can use these to connect to breadboards. That way you don't need long test lead adapters and you stay coaxial right up to the breadboard. Of course, this does need SMA Cables as well, but in my lab I've pretty much switched to SMA completely.

[tggzzz]

I am working on an RF project and have a 40MHz signal
...
My question is general - is there a way to use a breadboard at these frequencies?  Or should I just give it up as a bad idea and make my prototypes more with more permanent techniques (dead bug, varo board, etc.)?

You can use breadboards - as a doorstop.

You should do some Spice simulations including plausible parasitic capacitance and inductance. For the latter, assume 1nH/mm of wire. That will show you why breadboards don't work.

To find out how you should do it, look at ham (amateur) radio techniques; they are well documented. Fundamental principles: keep all leads small, have a groundplane, match impedances.

[Kalvin]

You may find this w2aew's video quite useful on how to breadboard things in RF frequencies:

https://youtu.be/kH110yjYZ2g?t=385

[hagster]

Just google 'die cast enclosure' there are hundreds of varients.

Depending on what you are doing you may not need a metal enclosure anyway.

If its for shielding you can get injection modled boxes with sheilding sprayed on the inside or you can buy rf shielding paint such as mg chemicals super shield if you want to do ot yourself.

[jgalak]

Most of what I'm doing on a bread boarding is experimenting with what will eventually go on a PCB - that's the nice thing about solderless breadboards.  Oh, hey, this cap didn't work?  Let me swap in another.  Fast and cheap (no wasted parts) compared to soldering things.  But, yeah, may just not work for these frequencies.

[cdev]

"Manhattan" style prototyping.

Vero board- you may be able to use that more effectively than manhattan for filters. Also ebay user "rfextra" sells a small filter PCB that you can use to make small filters very easily. It depends what frequency you need this for and how cheap you are..

Ugly style can work too the key to success being make everything incredibly short- especially the path the ground in critical places. This is why home made (but vias) or sent out PCBs using SMT is the way to go, ultimately, it allows you to have ground planes and predictability that is difficult to impossible to achieve in any other manner.

But FR4 is also lossy at higher freqs.

I am working on an RF project and have a 40MHz signal (it'll eventually be in the 140MHz range, I'm keeping it low for now since my scope is only 100MHz).  The signal currently goes from PCB, via coax, to the scope, and all is relatively well.

But now I want to insert a filter into the signal.  I don't have a coax inline filter, so want to breadboard something.  I have a couple of BNC-to-test lead adaptors, and tried to put the filter together on a breadboard and thus insert into the signal path.  The result is a disaster - both the input to the filter, and the output of the filter are horribly distorted.  A square-ish wave now looks like a jagged triangular wave.

My question is general - is there a way to use a breadboard at these frequencies?  Or should I just give it up as a bad idea and make my prototypes more with more permanent techniques (dead bug, varo board, etc.)?

Don't use a breadboard above the very bottom of HF, except at the very beginning. too much stray capacitance and inductance.

You could use the island technique for HF. Put SMA edge mount connectors for your connections but instead of clipping it to the edge as is done normally, have it higher, sitting on its edge mount connector like a little stand.  That will work fine for VHF too.

Doing this you can make some small filter prototyping boards.

Its easy, it takes longer for me to write it than do it, almost. The kind of SMA connector I get the most of is the edge connecting kind. Just soldertwo down facing each other a few cm apart from one another.. not grabbing the end of the PCB, sitting on it. Then build your filter in that space. If you anticipate maybe needing more room just use a larger piece of PCB so you have some room to expand. If its not big enough desolder it and move them a bit farther apart.

You can use the nibs chopped out by your nibbling tool as anchors for both through hole or SMT parts by using super glue, but ventilate well - with a fan blowing the smoke out your window  when you solder to its top..

[Yansi]

Even though FR4 is lossy, it is good enough up to 10 GHz homebrewing.  The increased loss tanget at very high frequencies even may be welcome when working with some crazy fast stuff, such as 25GHz ft transistors.

[vk6zgo]

I am working on an RF project and have a 40MHz signal (it'll eventually be in the 140MHz range, I'm keeping it low for now since my scope is only 100MHz).  The signal currently goes from PCB, via coax, to the scope, and all is relatively well.

But now I want to insert a filter into the signal.  I don't have a coax inline filter, so want to breadboard something.  I have a couple of BNC-to-test lead adaptors, and tried to put the filter together on a breadboard and thus insert into the signal path.  The result is a disaster - both the input to the filter, and the output of the filter are horribly distorted.  A square-ish wave now looks like a jagged triangular wave.

My question is general - is there a way to use a breadboard at these frequencies?  Or should I just give it up as a bad idea and make my prototypes more with more permanent techniques (dead bug, varo board, etc.)?

Even if you build it properly, a filter will inevitably distort your "squarish" wave.
By their nature, they are narrowband devices, so will not pass all the harmonics making up your waveform.
The input Impedance of the filter will also change radically outside the designed bandwidth.

[jgalak]

I _want_ it to distort my squarish wave   It's meant to be a low pass filter to kill the odd harmonics and get something reasonably sine-like.

But that's not what it's doing.  Just the act of going from coax to breadboard is causing the square wave to be horribly distorted, and what comes out of the filter looks like a jagged triangle wave.  It's a mess.

[radar_macgyver]

Just the act of going from coax to breadboard is causing the square wave to be horribly distorted, and what comes out of the filter looks like a jagged triangle wave.

Might also be due to your probing methods. High impedance probes with ground leads can add enough inductance to cause distortions like what you're seeing. Jim Williams had an app-note or two about this.

Edit: It's AN-47. http://cds.linear.com/docs/en/application-note/an47fa.pdf

Found a good companion too: http://readingjimwilliams.blogspot.com/2011/10/app-note-47-part-1.html

[jgalak]

Might also be due to your probing methods. High impedance probes with ground leads can add enough inductance to cause distortions like what you're seeing. Jim Williams had an app-note or two about this.

Edit: It's AN-47. http://cds.linear.com/docs/en/application-note/an47fa.pdf

Found a good companion too: http://readingjimwilliams.blogspot.com/2011/10/app-note-47-part-1.html

In this case, I wasn't using a probe, but going back into coax to the scope.  But I get your point.

[Kalvin]

Might also be due to your probing methods. High impedance probes with ground leads can add enough inductance to cause distortions like what you're seeing. Jim Williams had an app-note or two about this.

Edit: It's AN-47. http://cds.linear.com/docs/en/application-note/an47fa.pdf

Found a good companion too: http://readingjimwilliams.blogspot.com/2011/10/app-note-47-part-1.html

In this case, I wasn't using a probe, but going back into coax to the scope.  But I get your point.

My guess is that you are seeing reflections due to mismatch with the filter, coax and the oscilloscope's input impedance. The impedance of your filter should match with the coax, and the input impedance on the oscilloscope should match with the coax.

[cdev]

the free program RFSIM99 is useful

[jgalak]

My guess is that you are seeing reflections due to mismatch with the filter, coax and the oscilloscope's input impedance. The impedance of your filter should match with the coax, and the input impedance on the oscilloscope should match with the coax.

The signal source is 50 Ohm, as are  the cables.  There's a 50 Ohm through-terminator on the scope input.  Obviously when it went to the breadboard, that wasn't any particular impedance - it's just bits of random wire and the breadboard itself.

I'm not sure what you mean by impedance of the filter - it's just a simple RC lowpass filter - 22k series resistor and a .1uF parallel capacitor.  How would one match impedance on that?

[Kalvin]

My guess is that you are seeing reflections due to mismatch with the filter, coax and the oscilloscope's input impedance. The impedance of your filter should match with the coax, and the input impedance on the oscilloscope should match with the coax.

The signal source is 50 Ohm, as are  the cables.  There's a 50 Ohm through-terminator on the scope input.  Obviously when it went to the breadboard, that wasn't any particular impedance - it's just bits of random wire and the breadboard itself.

I'm not sure what you mean by impedance of the filter - it's just a simple RC lowpass filter - 22k series resistor and a .1uF parallel capacitor.  How would one match impedance on that?

Typically the RF filters are designed for specific impedances for the best performance. If the impedance of the source feeding the filter input and/or the impedance of the filter output do not match with the load, the filter response may not be as intended.

It looks that you have terminated the coax properly at the oscilloscope's end, so there should not be any reflections. So, my guess was wrong

[ogden]

I'm not sure what you mean by impedance of the filter - it's just a simple RC lowpass filter - 22k series resistor and a .1uF parallel capacitor.  How would one match impedance on that?

Your filter cutoff frequency is 72 Hz. No wonder it does not pass 40MHz signal 

For frequencies as high ar 40MHz or 140MHz, you would want to design LC filter:

http://www.wa4dsy.net/filter/filterdesign.html

Ugly layout and soldering, but still illustration pictures of island method:

http://www.rowetel.com/?p=4804

[fattowave]

There used to be a nice product for rapid RF prototyping and debugging called "Mini-Mount" in Europe and "Solder-Mount" in the US sold by Wainright Instruments. The idea was self-adhesive transmission lines (50 ohm, 75 ohm) and package footprints that could be cut to size and stuck down onto a large sheet of Cu-clad which acted as the groundplane. I believe it's discontinued now, but you could implement a DIY version by fabricating some cheap, single-layer PCBs with all the footprints, microstrip lines, bends, junctions etc. that you might need, and gluing them to a copper plate or Cu-clad board with double-sided tape. I built complete radio system prototypes using this stuff in the early 90s and it works reasonably well up to 2 GHz with care.

Some relevant info here:
http://www.lpkfusa.com/articles/prototyping/edn_2_96.pdf

[jgalak]

I'm not sure what you mean by impedance of the filter - it's just a simple RC lowpass filter - 22k series resistor and a .1uF parallel capacitor.  How would one match impedance on that?

Your filter cutoff frequency is 72 Hz. No wonder it does not pass 40MHz signal 

The possibility that I'm a bloody idiot can never be ruled out...

No idea how I came up with those values.  I know I was using an on-line filter calculator, but no clue what I did....  Changing that to 270 Ohm and 10 pF made thing work much better.  I am now, in fact, getting something resembling a sine wave.  It's not clean by any means (since this is still on the breadboard), but at least the filter is now making things better rather than worse.....

[ogden]

No idea how I came up with those values.  I know I was using an on-line filter calculator, but no clue what I did....  Changing that to 270 Ohm and 10 pF made thing work much better.  I am now, in fact, getting something resembling a sine wave.  It's not clean by any means (since this is still on the breadboard), but at least the filter is now making things better rather than worse.....

Now it is time for you to watch this video to understand why breadboard for RF is "no go":

[medical-nerd]

the free program RFSIM99 is useful

I have been unable to get it installed in Windows 10, any suggestions since this is the only windows version I have now.

Cheers

[Yansi]

Use a Virtual machine for it.

RFSIM99 runs fine on Win7.

[cdev]

RFSIM99 also works great on Wine

[james_s]

Dead bug it on blank copper clad PCB. It's not pretty but it's easy to do and can perform very well. It's actually challenging to make an etched PCB that works as well.

[Yansi]

I disagree on the etched PCB.

[james_s]

You disagree that it's challenging to make a good RF PCB? I guess maybe if you're a skilled RF engineer, but I certainly find RF boards to be a challenge. Not impossible by any means but more difficult than dead bugging a prototype.

[ogden]

I guess maybe if you're a skilled RF engineer, but I certainly find RF boards to be a challenge. Not impossible by any means but more difficult than dead bugging a prototype.

BTW it depends on what you consider as RF frequencies. For 900MHz I would disagree that deadbug gives better results than PCB

When you dead-bug, then anyway you shall know RF properties of circuit & components, same knowledge apply to PCB as well. Don't be scared of PCB, especially at <= 100MHz frequencies. Make everything preferably in straight line (at least for starter prototypes), short signal traces as possible. To be on safe side - always use 2-sided PCB but etch only one side, leave bottom as solid ground plane. Obviously don't be shy of some extra vias for ground traces on top layer, put them close to "load" - filter capacitors or RF components. Also preferably you shall learn how to make 50Ohm microstrip trace. There's various online calculators on internet for that. Basically that's it. Up-to 100MHz - just try PCB to see it's not that scary and usually it happens to work.

[Neomys Sapiens]

RFSIM99 also works great on Wine

Could you be a bit more specific? 
I use the software myself but I have not noticed a correlation between results and my state of inebriation.
So: white or red? vintage? HOW MUCH OF IT?

[KJDS]

RFSIM99 also works great on Wine

Could you be a bit more specific? 
I use the software myself but I have not noticed a correlation between results and my state of inebriation.
So: white or red? vintage? HOW MUCH OF IT?

Rather like dancing, there's an optimum amount, sufficient to lose your inhibitions but not enough to leave your legs wobbly.

[3db]

RFSIM99 also works great on Wine

Could you be a bit more specific? 
I use the software myself but I have not noticed a correlation between results and my state of inebriation.
So: white or red? vintage? HOW MUCH OF IT?

Nice one 

[whalphen]

I found that I frequently have a need to prototype RF filters.  I keep a good assortment of SMD capacitors and inductors.  And I had a supply of generic filter PCBs made.  They have SMD pads to allow a variety of filter topologies.  I can just leave unneeded positions empty or put jumpers.  The PCBs have an SMA footprint at each end and are narrow enough to slide into a short piece of 1/2 inch diameter copper pipe.  For the ends, I add copper pipe caps drilled out for the SMA connectors.  Bulkhead nuts on the SMA connectors can hold the caps in place.  I usually don't use the copper pipe unless I need the shielding or if I want the filter for long term use.  I've used this technique for a long time and use it for other things, too.  I keep some PI pad and T pad PCBs, too, for attenuators, DC blocks, and impedance matches.  I've even used it for RF amps -- usually with larger copper pipe, and feedthru caps to feed the power thru the end cap of the pipe.  I recently had a board made for use as a directional coupler to fit in a 1 inch pipe.

By the way, Tonne Software's free Elsie program is the filter designer I typically use.  I pick a topology, specify the filter, add in some limits for the various parameters, and let it optimize a filter design for me.  I can then quickly prototype a filter and check the results on a spectrum analyzer.  It almost always matches the expected results.  If anyone is interested in the PCB layouts, let me know and I can make them available.

[jgalak]

Oooh, that's clever.  I was going to do something similair on my next board, leave in space for a few passives in a filter configuration, but I really like this idea...

[cdev]

I'd be interested in your filter design PCB templates.

Thank you.

Ive use copper pipe a bit but need to revisit it because I've only recently gotten a decent drill, so its time to buy some copper caps and go to town, so to speak.

I've found with small LNAs the case makes all the difference.

What (brands?) caps and inductors do you prefer? I would be interested in your observations on this.

 I have enjoyed homebrew of small RF passive filters and receiver gain element blocks a lot.

I found that I frequently have a need to prototype RF filters.  I keep a good assortment of SMD capacitors and inductors.  And I had a supply of generic filter PCBs made.  They have SMD pads to allow a variety of filter topologies.  I can just leave unneeded positions empty or put jumpers.  The PCBs have an SMA footprint at each end and are narrow enough to slide into a short piece of 1/2 inch diameter copper pipe.  For the ends, I add copper pipe caps drilled out for the SMA connectors.  Bulkhead nuts on the SMA connectors can hold the caps in place.  I usually don't use the copper pipe unless I need the shielding or if I want the filter for long term use.  I've used this technique for a long time and use it for other things, too.  I keep some PI pad and T pad PCBs, too, for attenuators, DC blocks, and impedance matches.  I've even used it for RF amps -- usually with larger copper pipe, and feedthru caps to feed the power thru the end cap of the pipe.  I recently had a board made for use as a directional coupler to fit in a 1 inch pipe.

By the way, Tonne Software's free Elsie program is the filter designer I typically use.  I pick a topology, specify the filter, add in some limits for the various parameters, and let it optimize a filter design for me.  I can then quickly prototype a filter and check the results on a spectrum analyzer.  It almost always matches the expected results.  If anyone is interested in the PCB layouts, let me know and I can make them available.

[whalphen]

Attached are some PCB layouts (Kicad files and Gerbers.) Schematic PDFs are there, too.  I've shared them all on OSH Park in case you just want to get the boards. (Links are in the attached file.)  Because they're so small, they are very cheap on OSH Park.  (The filters are $2.55 for three PCBs, delivered.)  The filter modules use 0805 parts, but 0806 and 0603 parts can fit, too.  You may notice a couple of the boards are laid out linearly.  One is laid out with adjacent inductors orthogonal to minimize interaction.

I typically purchase the capacitors and inductors from DigiKey.  I use low cost 50V ceramic caps C0G/NP0 or X7R.  For the HF and VHF stuff I do I haven't found any noticeable difference with the higher cost RF caps.  For inductors I prefer wirewound, low series resistance.  I also keep a couple of full sets of cheap capacitors from AliExpress to use in case I run out or need an uncommon value.  I've had good experience with the ones from China, except that I sometimes find them mis-lableled.  For example 15pF caps labeled as 13pF.  I always measure the values before using them.

[jgalak]

A bit off topic, but since we are talking about this:  is there a good source (book, website, ebook) folks can recommend on filter design that covers very VHF/UHF?  Experimental Methods in RF design only goes to HF, but something at about that level for 2m and 70cm would be welcome reading.

[jgalak]

I'm getting errors from Kicad for a "wbhLibrary".  Is that perhaps your personal library?

[whalphen]

Yes.  That's a library with my custom symbols.  I think what's happening is that my custom library is still in the list of Component Library Files (under Preferences) that you downloaded.  I think you have three options to resolve that: 1) download the files again because I've corrected and replaced them, 2) go into Preferences and delete the reference to my library, or 3) just ignore the message.
 
Also, attached is another picture I shared on another thread.  Maybe it will give you some ideas.  The larger diameter one is a preamp I use for satellite comms.  The power is supplied thru the two feedthru caps.  As you can see, I sometimes put a male SMA connector on the module.  This is actually soldered on the outside of the cap with holes drilled for the center pin and the four PCB mounting pins. 

In almost all cases, I solder the SMA connector on one end to the cap and then solder it to the PCB.  Thus the PCB has an SMA connector and a cap attached to one end and only an SMA connector on the other end.  The cap on the opposite side is soldered to the pipe, but not the SMA connector. The pipe and its attached cap are slid over the PCB and held in place by the other cap on the far end and by a bulkhead nut on the SMA connector at the near end.  This way it can still be disassembled if needed.

Sometimes I coat the outside of the copper shell with clear spray paint to protect the copper.  And, as you can see in the photos, I sometimes use heatshrink tubing over the entire shell.  The pipe caps are called 'test caps' and are available from plumbing suppliers.  I usually get them from Zoro.com.  It's a bit tricky to hold the small cap while drilling.  I keep a 1" long piece of copper pipe for that.  I put the cap on the pipe and clamp both the cap and the pipe in a drill vise or a lathe chuck.  It helps a lot to use lubricant while drilling.

[Neomys Sapiens]

What (brands?) caps and inductors do you prefer? I would be interested in your observations on this.

I suppose that this question was not exclusively directed at whalphen, so:

C= Eurofarad(Excelia), CalRamic, American Technical Ceramics, Presidio Components, Novacap, Syfer, SRT Microceramique, DLI, Johansson, Cornell Dubilier
L= Microspire, Fastron.de, Renco, Piconics, AEM, Coilcraft, API Delevan, Venkel, Coiltronics, Johansson, NIC Components,
Other: Yantel (Temperature Compensation Attenuator, tunable SMD inductor), Anaren, State of the Art Inc

(also some Ind/Cap lines by AVX, Kemet, some branches of Vishay, but thats probably common knowledge)

Those elements shown by whalphen look like fine work. I wish I had a service for that!

[jgalak]

Hey whalphen, do you know the specific brand/model of the SMA edge connectors you used on the boards?  I can't find one that matches your pad dimensions, and I don't really know what's "close enough"...

[whalphen]

I usually buy them by the hundred from AliExpress. Here's one vendor I've used recently:
https://www.aliexpress.com/item/100pcs-lot-Gold-SMA-female-jack-Plug-Straight-Receptacle-solder-PCB-clip-edge-mount-RF-connector/32285453963.html?spm=a2g0s.9042311.0.0.mxrBlK

Also recently, this vendor sent me a hundred that got destroyed by the US post office equipment.  All I received was a mangled package with an apology form letter from the USPS.  (Must have been interesting to see 100 SMA connectors pinging around in a high speed mail sorting machine!)  The vendor sent me replacements at no additional cost.  https://www.aliexpress.com/item/Hot-Sale-100-Pcs-Gold-SMA-Female-Nut-Bulkhead-Solder-Deck-PCB-Clip-Edge-Mount-RF/32512250123.html?spm=a2g0s.issue_5ptha.0.0.p8rRER

For special sizes or lengths, I usually go to http://www.rfsupplier.com/, another Chinese company.

The pads on the PCBs are sized for the standard 0.25" wide connector.  Some of my PCB designs can also accommodate a wider one that I no longer use -- but even these work with the 0.25" connector.  You have to be careful to order the ones for 1.6 mm thick PCBs if that's what you want.  The ones for 1.2 mm PCBs are also common on AliExpress.  If you're not careful, you can end up with those.  I keep both.  When I order PCBs from China (usually PCBWay) I often order 1.2 mm boards to cut down the shipping weight.  (All my 1.2mm PCBs are masked in Blue so I don't get them mixed up with my 1.6mm green PCBs.)

The SMA connectors from China are a bit lower quality than US sourced connectors.  Dimensional tolerance is not quite as good.  But, for non-commercial use at HF and VHF, I find them to be completely adequate.  Of course you can get all these from major manufacturers via major distributors at higher quality, but the cost is several times higher.

I know some readers will be critical of Chinese sources.  But, over the past 3 years, I've placed nearly 500 orders for tens of thousands of parts from AliExpress vendors and have no complaints.  On rare occasions I've had a problem with a purchase, but the vendors nearly always were eager to resolve it.  In the few cases where I haven't gotten an adequate response from the vendor, AliExpress has always quickly refunded my money.  My experience with US vendors on eBay has been about the same rate of success as with Chinese vendors on AliExpress.  When I want high quality parts right away, I go to DigiKey (but then I pay a lot more.)

[jgalak]

Do you buy the bulkhead nuts separately? 

[Yansi]

Rfsupplier.com ... never came across that one. Very nice find. Already seen "Superbat" appearing a lot on Aliexpress.

Y.

[whalphen]

Do you buy the bulkhead nuts separately?

They usually come with the bulkhead connectors.  And since I don't always use the bulkhead connectors in bulkheads, I end up with extra nuts.  For the copper pipe caps, you can use the bulkhead connector with its supplied nut.  Usually, I install a regular connector on the end where the cap is soldered to the connector and I install a bulkhead connector on the end where the cap is held on by the nut.

[whalphen]

Rfsupplier.com ...

They are also a good source for enclosures.  They have a huge variety.

[jgalak]

Ok, so stupid question for whalphen.  I had some of your filter 1 boards made up and used elsie to design a chebyshev low pass filter.  It doesn't use every pad on the board, but fits nicely into the topology.  Problem is, it requres three of the pads to be shorted - L3, L5, L7. 

No matter how much I struggle, no matter how much flux and solder I use, I can't get a solder bridge to form across the two sides of these pads.  The solder mask is doing its job way to well...

So what do you use to short these out (if anything)?  I've thought of using a bit of wire, or a 0 Ohm resistor, but I'm not sure if they are going to introduce stray inductances or capacitances that'll mess up the circuit...

(for everyone else, these are 0805 SMD footprints that I'm trying to short out)

[KJDS]

Just use a short length of wire. That wire will have about 1nH of inductance. Add it to the circuit analysis and see what happens

[Yansi]

Iti is pointless to model a 1nH inductor as a wire jumper accross SMD 0603/0805 pads, unless you model the whole PCB with its all parasitics.

[KJDS]

One step at a time

[jgalak]

The software I'm using, Elsie, doesn't really let me model the extra inductance (afaik) - this jumper would not be in a place where the designed circuit has an inductor, so I can't just add an extra nH.  Elsie is more of a design tool than a modeling tool.

I suppose I could model the whole thing in LT Spice and see what happens.

What about using a 0 Ohm resistor, instead?  I have some of the right footprint as part of a resistor kit.  Would that introduce less stray inductance than a wire?

[Yansi]

O ohm resistor might have a bit less inductance than a plain piece of wire, due tonhe fact it is a flat thin film wire. But as i say, the difference ma be in %.

To simulate any kind of paasive rlc structure, get a copy of RFSIM99. It is a great good simple software for this task. Otherwise get any kind of suitable simulator, preferably one that does S-parameters as the rfsim. I have just stumbled upon the QUCS STUDIO. Also feee and very easy to learn and work with. Has plenty of extensive step by step tutorials.

Dont forget to add loss components when simulating the filter structure.  The real inductors definitely do not have infinite Q factor.

//sorr mistyped from my mobile phone.

[whalphen]

I just use 0 ohm resistors.  At HF and VHF, I don't think it adds enough inductance to be significant.  You can add these to an Elsie model.  After you have the circuit design done using the topology in Elsie, just edit the circuit and add a stage using Insert Part Prior to This Stage.  Put an inductor there.  Anyway, the parts you buy and install will likely have tolerances greater than the reactance from a nH or two that may result from the 0 ohm resistors.  Elsie is a great tool for modeling this.  Add the inductor as I described.  Then set the stepsize values and run the Monte Carlo analysis to see just how much effect the variation has.  If you find it to be significant, just add a bit of capacitance in series to offset the reactance.  Just calculate the amount of capacitance needed to produce a similar magnitude of reactance as the inductance at the frequency of interest. 

I seldom stick to the design Elsie comes up with initially.  I always set limits and run the optimizations several times, improving the filter to meet my specific needs and to minimize sensitivity to the higher tolerance parts.  A couple of Elsie hints: check the transmission angle to make sure you don't get a 180 degree phase shift at the frequencies you are using.  For some modes, this will distort your signal.  You can put limits and run the optimization to correct this.  Check the return or VSWR and set limits if you want to control reflections during optimization.  I also sometimes use the 'swap end - end' feature and check return or impedance from the tail end of the filter, too, if I'm concerned about that -- for example, with filters used in diplexers.  And, as Yansi suggests, set reasonable Q factors in the Elsie analysis tab.

By the way, another very useful tool is the free Iowa Hills Smith Chart software.  I use it often.  It gives the benefits of a Smith Chart without having to spend too much time learning how to use it.  It's great for coming up with impedance matching circuits.

[jgalak]

Thanks. I'm only understanding maybe 1/4 of the things in Elsie, so lots of learning to do

I had to tweak the design extensively, since it generates oddball L and C values that don't exist in real components.  "Nearest 5%" helps, but not enough.  So took quite a few iterations.  The filter is for an APRS tracker, so my frequency of interest is pretty limited, which makes it easier.

Here's what I ended up with:

[jgalak]

Note that this is a "first approximation" filter.  It's not going to be used on the air.  For now, it's just there so I can kill extra harmonics when feeding the signal into a SpecAn/receiver.

For on-air use I'll either get a commercial one or, hopefully, will have a real SpecAn by then so I can actually test the filter "as built".

[whalphen]

Looks good.  And it meets the FCC requirements, so you could use it on the air.  Attached is the one I used on one of  my tracker designs.  It's very similar to your's.  I had Elsie design the filter.  Then I set the limits -- shown in red -- and had Elsie optimize it.  Then selected nearby inductor values from my inventory, set their step sizes to 0, and had Elsie optimize the capacitors to work with my inductors.
When I built that tracker I didn't have a spectrum analyzer.  I just fed the output of the tracker through an attenuator and then to an SDR dongle to look at the harmonics. It's not as accurate as a spectrum analyzer, but it was good enough to confirm that harmonics were well below the limits.

[jgalak]

Yeah, SDR dongle is the best I have right now. 

[jgalak]

So I finally got myself a Spec An, and was able to characterize the filter I made.  Looks pretty good to my eye:

[ogden]

So I finally got myself a Spec An, and was able to characterize the filter I made.  Looks pretty good to my eye:

Looks nice indeed. Always do "reality check" as well - full freq range of your SA. Depending on layout/enclosure sometimes RF can find leakage/reflection path around your filter at high frequencies.

[jgalak]

So I finally got myself a Spec An, and was able to characterize the filter I made.  Looks pretty good to my eye:

Looks nice indeed. Always do "reality check" as well - full freq range of your SA. Depending on layout/enclosure sometimes RF can find leakage/reflection path around your filter at high frequencies.

So that was an interesting point and I decided to do just that.  Below two images were taken with the exact same setup (slightly different cable and connectors than previous one).

First shows 0-500MHz.  The second shows full scale, 0-3.2 GHz.  Is it really not filtering at all at those frequencies?  At 2.9GHz, it appears to actually be 2dB _above_ reference!  I'm assuming its because at those frequencies you get weird coupling effects on the PCB, and the trances and components are no longer working as expected.

Now, in practice, this probably isn't a huge deal, as there shouldn't be any energy that high up going into the filter in the first place, but it's definitely good to know...

[Howardlong]

I disagree on the etched PCB.

Same here.

My rule of thumb is HF and low VHF is fine for Manhattan as long as you take care (e.g. short leads), mid-VHF and above you need to consider fabricating your own boards.

At the risk of flogging a dead horse, the Busboards with solid copper backplanes are my preferred alternative to Manhattan up to about 100MHz or so for RF and mixed signal prototyping.

[ogden]

Now, in practice, this probably isn't a huge deal, as there shouldn't be any energy that high up going into the filter in the first place, but it's definitely good to know...

Strange coupling at 2.9 GHz. Filter could have built-in "antennas" I wonder how your filter layout looks. Did you "fold" your filter so in/out is close to each other? Use non-SMD capacitors? I would be worried about poor filter performance around 2.4GHz. Nearby WiFi transmissions can saturate LNA or impact AGC operation, if any.

Note that by adding two shunt capacitors at input/output, thus converting filter from T to PI topology, you will get 9th order filter using same number of inductors in place of current 7th order filter.

[jgalak]

The layout is whalphen 's, above.  All components are smd.  Filter is inside a piece of 1/2" copper pipe with end caps (though caps are not soldered on), as per his design.

The goal of this is to be an APRS tracker output filter.  The particular "transmitter" I'm using is a bit weird, and outputs a square wave, not sine.  Running the unmodulated carrier through this filter gave me just one unwanted spike, at the third harmonic, but down about 50dB from the primary.  Unfiltered, it gave a beautiful rake at the odd harmonics.

[xaxaxa]

just a tip: SMA connectors from aliexpress/ebay are notorious bad because of their fake gold plating; the connector looks and functions fine for a month until they corrode.

It's easy to test though; just prepare a salt & vinegar solution (adding salt until saturated is fine), put one drop onto the connector and wait for it to dry. If there is visible discoloration then there is no gold plating; a good connector will go almost untouched by the solution.

You can also tell by price; if it's below $0.4 usd per connector it's guaranteed fake; good quality connectors with real plating average about $0.5 usd in china, and about $2 usd from mouser/digikey. I would look for a listing that specifically say gold plating and do the vinegar test; if it fails the test I'd have a message with the seller.

[daslolo]

You may find this w2aew's video quite useful on how to breadboard things in RF frequencies:

https://youtu.be/kH110yjYZ2g?t=385

What are the spiral wires for?

[Howardlong]

You may find this w2aew's video quite useful on how to breadboard things in RF frequencies:

https://youtu.be/kH110yjYZ2g?t=385

What are the spiral wires for?

They're hand wound inductors, there's a ferrite toroid that they're wound around.

[eb4fbz]

So I finally got myself a Spec An, and was able to characterize the filter I made.  Looks pretty good to my eye:

Looks nice indeed. Always do "reality check" as well - full freq range of your SA. Depending on layout/enclosure sometimes RF can find leakage/reflection path around your filter at high frequencies.

So that was an interesting point and I decided to do just that.  Below two images were taken with the exact same setup (slightly different cable and connectors than previous one).

First shows 0-500MHz.  The second shows full scale, 0-3.2 GHz.  Is it really not filtering at all at those frequencies?  At 2.9GHz, it appears to actually be 2dB _above_ reference!  I'm assuming its because at those frequencies you get weird coupling effects on the PCB, and the trances and components are no longer working as expected.

Now, in practice, this probably isn't a huge deal, as there shouldn't be any energy that high up going into the filter in the first place, but it's definitely good to know...

You get re-entry response at high frequency due to inductors parasitic capacitance (interwinding capacitance). When designing high frequency filters you should have all parasitic effects into account, and even use RLC models for every component. Check the self resonance frequency (SRF) of the inductors you have used, the inductors will act as a perfect short at that frequency.

[jgalak]

This filter is designed for APRS, so the frequency of interest is at 144MHz.  The source, feeding into the filter, puts out a square wave, so there's also the odd harmonics of 144MHz.  Shouldn't be anything anywhere near the re-entry response, I'd think.

Looking at the signal with the filter, I get a single extra spike at the 3rd harmonic, about 40dB below the fundamental, and nothing else is visible within the dynamic range of the SA. 

So I think I can safely call it "good"

[Dwaine]

No idea how I came up with those values.  I know I was using an on-line filter calculator, but no clue what I did....  Changing that to 270 Ohm and 10 pF made thing work much better.  I am now, in fact, getting something resembling a sine wave.  It's not clean by any means (since this is still on the breadboard), but at least the filter is now making things better rather than worse.....

Now it is time for you to watch this video to understand why breadboard for RF is "no go":

I learned something new today.   

[tggzzz]

No idea how I came up with those values.  I know I was using an on-line filter calculator, but no clue what I did....  Changing that to 270 Ohm and 10 pF made thing work much better.  I am now, in fact, getting something resembling a sine wave.  It's not clean by any means (since this is still on the breadboard), but at least the filter is now making things better rather than worse.....

Now it is time for you to watch this video to understand why breadboard for RF is "no go":

That's half the story; the other half is the inductance of the wires - assume 1nH/mm.

Then calculate the resonant frequency of a wire connected to a breadboard.

[Wolfgang]

IMHO, solderless breadboards and RF are not a good combination.
What does work, is "Bronx" style using components above a groundplane.

https://electronicprojectsforfun.wordpress.com/prototyping/

I used techniques like this (with 50Ohm striplines made from FR4 PCB strips) successfully up to 23cm.
RF feels OK in a defined impedance environment. A chaotic rats nest of wires is a no-no (except if you care for all kinds of surprise effects).
Deblock everything with caps, use a groundplane, keep wires extra short or put signals on 50Ohm lines, apply shielding, use SMA edge connectors,
avoid probes, ... and your life will be easy

[ogden]

IMHO, solderless breadboards and RF are not a good combination.

Yes, Mr.Obvious.

This is exactly what we are talking here. You are advised to read thread before posting.

[Wolfgang]

If you have a problem that I agree with something, I probably cannot help you.
BTW: I have read the other entries.

[ogden]

If you have a problem that I agree with something, I probably cannot help you.
BTW: I have read the other entries.

There's huge difference between statement and agreement.

Could you point out something like "I agree to what's said multiple times before" in following sentence?

IMHO, solderless breadboards and RF are not a good combination.

[Wolfgang]

The simple answer is : forget a breadboard at 40MHz.

If you need to try something out at 40MHz, it should look similar to this one:

https://electronicprojectsforfun.wordpress.com/an-noninteger-frequency-divider-with-sinewave-output/

If you scroll down, you can see a lot of prototypes built on copperclad PCBs. There is also a bandpass filter ("BPF"). All these work fine,
and the technique will be OK up to 150MHz if you keep all leads very short and all components close to the copper ground.

The picture also shows a finished amp module in a shielded box. This is a safe way to prevent unwanted coupling or radiation.

[james_s]

For a long time that deadbug construction technique looked so sloppy I just couldn't bring myself to do it. In more recent years I've played around with it some and found it performs extremely well, I've really come to appreciate it.

[Wolfgang]

I had the same learning curve. One has to differentiate between optical and electrical uglyness 

A lot of "neat" circuits suffer from extra lead lengths, unwanted couplings due to "all in a row" aligned components, ...

Ugly construction (I call it "Bronx" style to separate it from the very nice "Manhattan" style by Jim Kortge) can be used will into the GHz region if ou know hat you are doing.

[CopperCone]

If you think that bronx is ugly chances are your handtools are not up to snuff.

I recommend you have a good look through say the swanstrom catalog to find nice hand tools to work with.

Hand tools are important, remember that time travel episode of startek TOS? You also want a fine vise (I have a mini machinists vice and several watch makers vise) to hold things in place. The watchmakers vise is the size of a silver dollar.

I think the problem is that you can work extremely rapidly to throw some shit together. Like any other skilled craft, you can make things alot nicer if you want to. You can solder resistors to each other with 45 degree cut leads (like a needle) so that they match nicely and have what appears to be a single wire connection if you want. Much better then the twist&solder or parallel solder that most people do. If you are willing to go under the microscope you can make structures that appear nice for mating three or more components together at what looks like a solid connection... remember you can set everything up on mini clamps, over solder, then use a wick to very carefully remove solder.

A microscope/heavy magnitifer and jewelers files are useful too (like diamond mini files).

[Wolfgang]

same discussion as the one about Stalinistic shielding. Cliches and Metaphors are determined by your social background.  Here in Europe the Bronx is deemed less nice than Manhattan
What is meant is that maximum care should be spent on "electrical" beauty, with niceness only where it does not hurt

[CopperCone]

Well I thought manhattan being rich ass socialites basically used the equivalent of table coasters, so their fine french finish on the 5000$ coffee table is not stained (manhattan style)

while bronx style is basically some poor guy having a hot dog on the street and they don't care if they stain the sidewalk ... or at least they are not Bourgeoisie enough to use coasters.

Manhattan style from what I under stand is more like the style used for high impedance connections and bronx is classical 'deadbugging'

I thought for manhattan style they sell little pads that you are supposed to glue down to the copper clad board to act as islands. Bronx you typically would give it a twist and solder or just bring them kinda close together and join with a solder bead...

whereas I am saying to do fine work like miter your component lead cuts, clamp them in a vise, carefully join with the minimum of solder and then remove or even file away excess solder to make yourself a seamless looking connections (I have yet to see anyone publish this method). I mean you can even polish your solder joints if you want.. I am just saying you can make something typically ugly look really nice....


And if soldering is anything like brazing, you theoretically SHOULD have some kind of jigs to hold things together.. if you look at braze strenght curves, you will see that for good brazing you want to really precisely align up your parts, use shims to determine if the spacing is correct (might be somewhere like 5 thousandths of an inch, it depends on all sorts of factors) then apply braze for maximum strength.... but again it does not matter at all with soldering.. maybe the joints are statistically less reliable.

When I think typical bronx I imagine some guy holding parts together and touching them with a soldering iron held in his mouth!

[Wolfgang]

... and I was naive and thought the "Manhattan" came from mounting all components upright like Manhatten skyscrapers 

[CopperCone]

its a specific style where you use PCB board punch outs to act as connector islands. I think the rational is that it looks like manhattan from a view of how the 'foundations' are planted.

The next step would be 'D.C.' style construction where you make sketches and do pre-design work on the layout so its like a planned community layout, like washington DC.

[HB9EVI]

I was quite disgusted by the look of a lot of manhatten style projects; and it was obvious to not be much RF compatible. I adopted the 'drilling island' system mentioned by om Hans DJ1UGA using a Dremel cutter like this https://www.dremeleurope.com/ch/de/fraeser-hss-6-4mm-135-ocs-p/ . they sell in different diameters, drilling appropriate islands to solder SOT23 transistors and 0806 Rs and Cs. that's really very neat to build good RF circuits; I used the technique for 5 band HF-transceiver together with basic through hole components like crystals and some dip8 as dead bugs

[Wolfgang]

Maybe the proper name for the many island styles would be "polynesian" style 

I think all styles are OK as long as "electrical beauty" (.i.e. minimal parasitics and unwanted coplings) is honoured. The rest is a matter of taste.
Manhatten is definitely not the best for high frequencies due to redundant wire lengths caused by the all-upright mounting method.

The ideal mounting style is the one where all components are connected just by soldering joints at their ends (no longer leads), or, if not possible,
by defined impedance transmission line segments.

The technique to drill islands into a copper plane is OK, but not as flexible regarding circuit modifications as the pure Bronx.

[HB9EVI]

agreed! it requires a bit of planning, since afterward mods are tricky - but still not impossible. In every case it's in the meantime my favorite way of construction.

I like calling it the 'polynesian style'

[bd139]

Interesting thread this.

For TH, I’ve been using a variation of Manhattan style which relies on the components being mounted almost legless and flat between the islands which are carefully spaced to make sure there is just enough room. This removes all those neat but electrically horrible excess inductors everyone leaves on one end of a component and decreases the weight of the final assembly. Similar to what you found in old point to point wired equipment but without the rat’s best.

I call this Endor style after the Ewok’s tree houses in Star Wars.

BUT honestly there’s not much point in using crappy old TH components now for ham stuff at least unless it’s high power now. And almost definitely not for anything VHF and above. And manhattan is ugly as sin and extremely difficult to modify once done because of all the glue stuck all over the ground plane.

Just draw the sub unit schematic out (say an LPF), cut some square islands in FR4 with an X-Acto and just bridge the gaps. No messy glue, cheaper and smaller. SMD stuff costs bugger all as well, even the brand name stuff. When you have a sub assembly working, glue that to another bit of FR4. It also acts as a very convenient low impedance star topology ground.

Typical module (10Mhz VCTCXO + power supply):

[Wolfgang]

SMD is great for people with eagles eyes and very steady hands (not me).
For this reason, a lot of prototypers still prefer thru-hole components if frequency range permits.
In case I really needed SMD stuff for performance reasons, I made a PCB, too.

For UHF and above, the PCB becomes part of your circuit anyway and must be taken into account if you want a repeatable result. 

[tggzzz]

For UHF and above, the PCB becomes part of your circuit anyway and must be taken into account if you want a repeatable result. 

Very true.

Unfortunately too many people with modern logic operating at 1Hz don't realise that is also operating at UHF!

[dmills]

Clock frequency does not matter, edge rates do!

[bd139]

Fortunately most logic families are relatively forgiving of misuse

[Wolfgang]

The "fastest" logic family I was able to make a Bronx style circuit with was an 80MHz LFSR using 5 ACMOS chips.
When I tried to do the same with ECL100K and 200MHz it never reliably worked, despite using 50Ohm striplines.
After making a PCB with controlled impedance tracks and terminators it finally worked.

On the other hand, I was able to make a working 23cm two-stage amplifier using MMICs and striplines on a Bronx-Style backplane.

https://electronicprojectsforfun.wordpress.com/prototyping/

So, my experience is that fast logic is even *less* tolerant to construction style than plain analog !

[HB9EVI]

Fast logic (74AC) I still got running at 120MHz as dead bugs to create the phase shift for an I/Q mixer without issues. Analog worked out so far in polynesian style up to the 23cm band on FR4.
I'm normally simply too lazy to layout a pcb and etch it

[Wolfgang]

The simpler the circuit, the better the chance to get away with suboptimal wiring. 120MHz for 74AC in a very simple circuit with 1-2 ICs might work. An LFSR with 5 ICs all in the signal path is on the edge regarding gate delays and setup times. If reflections and ringing gets on top of that, it is likely to fail at higher frequencies. I tried 100MHz and it failed, even on a nice PCB.

Just to see how far you could go you could check out the multi-GHz ugly construction cicuits from LY3LP:
http://laboratory1475.rssing.com/chan-34446832/latest.php

[james_s]

If you need a LFSR running at 10s of MHz it starts to make a lot more sense to use a CPLD. In a sense that's like a very small breadboard with very short connections.

[Wolfgang]

Yeah, agreed. I planned to do that, but I was held up by other tasks. What CPLD type would you recommned for a 32Bit register, and what would be a realistic clock frqeuency ?

[james_s]

You'd have to look at the datasheets, I'm not up to date with what's currently on the market nor have I tried pushing the limit with the clock frequency. I've used the Xilinx XC9536XL and a similar Altera part (Max II?), the dev boards are cheap so you could pick one up to try it out, something like this https://www.ebay.com/itm/Altera-FPGA-programmatore-CPLD-USB-Blaster-compatibile-LC-MAXII-EPM240/172477210445?

If you need a hand with the code I could whip up some VHDL for you if you have a schematic of the LFSR you want.

[tggzzz]

The simpler the circuit, the better the chance to get away with suboptimal wiring. 120MHz for 74AC in a very simple circuit with 1-2 ICs might work. An LFSR with 5 ICs all in the signal path is on the edge regarding gate delays and setup times. If reflections and ringing gets on top of that, it is likely to fail at higher frequencies. I tried 100MHz and it failed, even on a nice PCB.

Even a simple circuit will fail, if there is a heavy load and "long" ground leads: ground bounce can be a serious problem.

[cdev]

Its easier to use a short length of mini coax than a PC trace sometimes, while keeping the PCB attached at both ends for the extra groundplane. You get the best of both worlds that way. Its quieter.