20dB 'RF' attenuator - seeking feedback to improve

[enut11]

My first 50 ohm 'RF' attenuator built on a 25x65 double-sided pcb.
20dB attenuation using metal film resistors.
As far as I can measure, it is -1.5dB at 45MHz.
I would like to do better than this and get to about 300MHz.
Is this possible using conventional axial resistors.
Suggestions?
enut11

[awallin]

build the same thing with SMD and SMA-connectors and compare?
something like that:
https://www.banggood.com/DC-4_0GHz-RF-Fixed-Attenuator-Radio-Frequency-Fixed-p-1119478.html

a grounded metal box around everything (aluminium foil in the prototyping-stage?) might help.

[T3sl4co1l]

Well don't fly it in air, put those puppies down on the ground plane!

Also use connectors that aren't two teeny wires.  Those things are dumb, they shouldn't exist.   An edge-launch or 5-pin style will do alright.

You may still need to use wider or flatter (or more in parallel?) resistors to keep Zo right, or use SMT chip resistors which, obviously, don't stand up at all, they stay flat to the plane.  Carve a trace across the board so you get a microstrip or coplanar waveguide geometry.

Tim

[ebastler]

Here's a short video that demonstrates a more compact style of building these, with better HF properties. A bit of theory first; then he shows his homebrew build starting around 6:30:

[Leo Bodnar]

Free air impedance of 50mm of 0.5mm wire is +j*14.1Ohms at 45MHz https://chemandy.com/calculators/round-wire-impedance-calculator.htm
I think that's the main killer of this construction style.
Leo

[enut11]

Free air impedance of 50mm of 0.5mm wire is +j*14.1Ohms at 45MHz https://chemandy.com/calculators/round-wire-impedance-calculator.htm
I think that's the main killer of this construction style.
Leo

Thanks Leo Bodnar. That's what I needed to know.

I want to avoid SMA if possible as it involves too many adapters. Ditto wrt surface mount but may revisit this later.

I will start by lifting the ground plane up to the series resistors and see what improvement that offers.
enut11

[duak]

Others have covered rebuilding with the resistors closer to the ground plane.

Two resistors in parallel will give twice the parasitic parallel capacitance of a single resistor.  It'd be better to use two in series if you can't get the exact value needed.

A trimmer capacitor across the central resistor could compensate for the parasitic capacitance of the resistors.  If you model the circuit including the parasitic capacitances, I think you'll find a compensation cap is needed for VHF.  Passive scope probes have these.  The central resistor to ground will have an inherent series inductance  caused by the loop area.  If significant, two resistors in parallel but arranged in opposite directions like a '+' can help.

I built a filter in a similar case.  Adding a grounded metal divider with the signal passing through a hole helped ultimate rejection.  I'd consider an upside down U shaped one to shield the central node.  The series resistors would pass through the two holes in the divider.  However, if the ground plane can be brought up to the resistors, a grounded shield can be placed on top making a sandwich of it and no holes would be needed.

Cheers,

[enut11]

Thanks @duak. Lots to consider for my next upgrade.
Meanwhile, have a look at my latest mods, done before I read your posting.
Yet to be tested as I need a reliable HF generator (current one max is 50MHz).

The resistor network now sits on a grounded plane with shortened input/output leads poking through grounded shields.

Maybe a pi configuration would be better if I must use series resistors?
enut11

[David Hess]

Axial resistors like that in a cavity are good to 1 GHz so hug the ground plane.  Use a connector which terminates closer to the ground plane like edge mount.  Through-hole 1/8 watt parts are smaller and will perform a little better.  Mount the shunt resistors to make a cross along the ground plane instead of a tee.

[enut11]

Axial resistors like that in a cavity are good to 1 GHz so hug the ground plane.  Use a connector which terminates closer to the ground plane like edge mount.  Through-hole 1/8 watt parts are smaller and will perform a little better.  Mount the shunt resistors to make a cross along the ground plane instead of a tee.

Thanks David. The shunt resistor is 10 ohm for the T-attenuator.
If I use 18 ohm and 22 ohm in parallel I can form a cross arrangement of resistors. 18//22=9.9 ohm
The cross arrangement was also suggested by @duak.
enut11

[Miti]

Suggestions?

https://www.minicircuits.com/WebStore/dashboard.html?model=HAT-20%2B

[EEEnthusiast]

I have made tons of those attenuators. The layout is the key in all of these.
Few tips
1. Use the smallest possible resistor case size (0402 or 0201 if you can solder them)
2. Use micro-strip to route the signal in and signal out.
3. Use end launch SMA connector for the ends.
4. Match the width of the trace to the width of the component. For e.g. in case of 0402 , use a trace width of 20 mils
5. Adjust the ground plane height in order to get 50 Ohms at 20 mils width.
6. Use a PI or T network. The shunt element should have its pad resting on the transmission line so that there is no stub.
7. For the ground connection, use multiple vias to connect to the ground plane below.

If you do all these, and with enough luck you may be able to get an flat attenuator well into couple of GHz.

[cncjerry]

There are/were smd boards unpopulated on ebay for attenuators and filters.  They used sma edge connectors and the circuit board lines were 50ohm.  I bought some.  Then I found these little screw on attenuators that were really nice for such low cost I gave up making my own.  The nice thing about making your own though is if you burn them up, they are cheap to fix.

[enut11]

All
Thanks for all the suggestions. And yes, I could buy a 20dB attenuator for a few dollars but there is little to learn in doing that

One thing that I have noticed is that members prefer SMA to BNC. Can anyone explain this given that most test gear is fitted with BNC?
enut11

PS
At the moment the max I can test at is 50MHz and even that is probably a little dodgy. Any suggestions for an affordable 500MHz or 1GHz generator?

[Miti]

Any respectable test instrument that goes into GHz is fitted with N type connector. BNC was designed, according to Wikipedia, for applications up to 2GHz but I wouldn’t use it in a test system that goes close to that frequency. SMA can go up to 18GHz. I use SMA and SMA to N adapters in most of my test equipment.

[ogden]

At the moment the max I can test at is 50MHz and even that is probably a little dodgy. Any suggestions for an affordable 500MHz or 1GHz generator?

You may consider NanoVNA. Not quite generator, but next best thing to measure your attenuators. Far from "lab grade instrument", yet gives surprisingly good results up-to 300MHz and with (quite) some compromises up-to 900MHz.

[enut11]

I have made tons of those attenuators. The layout is the key in all of these.
Few tips
1. Use the smallest possible resistor case size (0402 or 0201 if you can solder them)
2. Use micro-strip to route the signal in and signal out.
3. Use end launch SMA connector for the ends.
4. Match the width of the trace to the width of the component. For e.g. in case of 0402 , use a trace width of 20 mils
5. Adjust the ground plane height in order to get 50 Ohms at 20 mils width.
6. Use a PI or T network. The shunt element should have its pad resting on the transmission line so that there is no stub.
7. For the ground connection, use multiple vias to connect to the ground plane below.

If you do all these, and with enough luck you may be able to get an flat attenuator well into couple of GHz.

Thanks EEenthusiast
I cut my (electronic) teeth on audio and analog power supplies. RF is new learning for me.

1) So one must sacrifice power handling for frequency performance by using smaller resistors?
2) Not familiar with 'microstrip' but will do a search.
3) Looks like I will have to succumb to SMA on the next design . However, If I use adapters to connect to my BNC equipment, it seems I will not be gaining much??
4), 5) and 6) : RF track layout is outside my current knowledge so will have to research.
7) Easy to impliment
enut11

[EEEnthusiast]

Microstrip is nothing but a simple trace with a complete ground plane below it. You can search for online calculators, which will calculate the width of the trace needed for your dielectric.
What kind of power are you planning to dissipate. For higher power dissipation, you could use multiple resistors in series parallel combination.
For a 20dB attenuator, you make it into multiple stages. For eg. 10 stages of 2dB each. So each stage will dissipate less power.

[EEEnthusiast]

Your final PCB will look like multiple PI or T sections cascaded in series.
Something like this...

[enut11]

Microstrip is nothing but a simple trace with a complete ground plane below it. You can search for online calculators, which will calculate the width of the trace needed for your dielectric.
What kind of power are you planning to dissipate. For higher power dissipation, you could use multiple resistors in series parallel combination.
For a 20dB attenuator, you make it into multiple stages. For eg. 10 stages of 2dB each. So each stage will dissipate less power.

At this point in time I do not have a particular 'power' application and am just learning 'RF' methodology.
I have been characterising my instruments (HF voltmeter and DSO) and need to reduce the signal for the lower ranges (power requirement is very low).

For the attenuator circuit board layout you say that a strip of a certain width is needed to suit the ground plane I am using (1.5mm double sided copper fiberglass PCB). Does this mean it can only be used over a small frequency range?
enut11

[Berni]

All
Thanks for all the suggestions. And yes, I could buy a 20dB attenuator for a few dollars but there is little to learn in doing that

One thing that I have noticed is that members prefer SMA to BNC. Can anyone explain this given that most test gear is fitted with BNC?
enut11

PS
At the moment the max I can test at is 50MHz and even that is probably a little dodgy. Any suggestions for an affordable 500MHz or 1GHz generator?

Because SMA is a better connector in a lot of ways and is very common everywhere.

The BNC connector while very common has some issues. While it is very easy to plug in (No need to screw it on all the way) it does suffer in terms of RF performance and is physically pretty large. Have a look at RF test equipment and try to find a BNC on there, you won't. What you find there is N connectors or sometimes APC7 connectors (Mostly only on network analyzers) because of there much superior RF performance. Any BNCs you find on RF test gear is for slow low bandwidth signals like modulation, leveling control, gating, trigger etc..

In general typical BNC starts having problems once you approach 1GHz, fancy good quality BNCs go higher into a few GHz but you also need the nice high quality connector on the cable too. While the N connector has no problems with GHz and good ones go to 20GHz and up. Only problem is that N is twice the size of BNC so soldering N connectors on your deadbug style prototype circuit is not convenient at all. However there is also the popular SMA connector, featuring a size significantly smaller than BNC while offering good RF performance to 4GHz for the cheap ones while higher performance variants like the 3.5mm or 2.5mm go beyond 20GHz in the identical form factor (Even SMA compatible, but please be nice to fancy RF connectors). You can buy a bunch of cheep board mount SMA connectors and just solder then into  your circuit whenever needed, treat them as disposable.

Have a look at one of the biggest manufacturers of RF modules: https://www.minicircuits.com/ Most of there modules are SMA because its such a convenient connector that provides great RF performance in a small form factor.

But... my <insert testgear here> only has BNC!
Well.. so? What does it matter? You do need a cable to hook it up to something anyway. So you just get a cable that is BNC on one end and SMA on the other. Any place that sells RF cables will have such cables. Or if you want to DIY it you can just cut up a BNC cable in half and put SMA connectors on the end yourself. So i just keep a bunch of BNC to SMA and N to SMA cables and a box full of cheep various kinds of board mount SMAs.

If you want cheap cables but want it a bit more legit than a random ebay seller this is a good Chinese company: https://www.rfsupplier.com They have every possible RF connector and will make any kind of custom RF cable for you even in small quantities.

As for a RF generator, used market like ebay is the best choice to pick up an old boatanchor that goes to a few GHz (Tho australia might be tricky)

Also if you are just getting into RF maybe you might find the NanoVNA very useful: http://nanovna.com/ Its a $50 network analyzer to 900 MHz (Tho best performance is only up to 300MHz). This lets you characterize your RF blocks properly across frequency and show you things like phase and return loss.

[joeqsmith]

 

All
Thanks for all the suggestions. And yes, I could buy a 20dB attenuator for a few dollars but there is little to learn in doing that

One thing that I have noticed is that members prefer SMA to BNC. Can anyone explain this given that most test gear is fitted with BNC?
enut11

PS
At the moment the max I can test at is 50MHz and even that is probably a little dodgy. Any suggestions for an affordable 500MHz or 1GHz generator?

Seeing that you are experimenting, you may want try different resistor types as well.   You may also want to try using a Pi arrangement as well.     

I agree with the above comment about buying the NanoVNA as a starting point.   Looking at your 20dB example wouldn't be a problem with it.   Even at 1GHz, I imagine you would at least be in the ballpark with it.   

https://youtu.be/mKi6s3WvBAM?t=2021

[David Hess]

Any respectable test instrument that goes into GHz is fitted with N type connector. BNC was designed, according to Wikipedia, for applications up to 2GHz but I wouldn’t use it in a test system that goes close to that frequency. SMA can go up to 18GHz. I use SMA and SMA to N adapters in most of my test equipment.

For a while there was push to replace BNC connectors with SMB connectors in test instruments.

1) So one must sacrifice power handling for frequency performance by using smaller resistors?

At higher power and frequency, transmission line construction would be used for the resistors so that they could be made much larger.

On a more practical level, multiple attenuator sections made from discrete parts can be used in series to distribute the heat.  So for example stacking 3dB, 14dB, and 3dB sections will double the power over a single 20dB section because the first stage only absorbs half of the power.  Other circuit layouts are possible to increase power handling.

In general typical BNC starts having problems once you approach 1GHz, fancy good quality BNCs go higher into a few GHz but you also need the nice high quality connector on the cable too.

Tektronix discovered that when designing their 7A29 1 GHz vertical amplifier for their 7104 1 GHz oscilloscope mainframe.  No available BNC connector at the time had a uniform impedance so Tektronix ended up making their own precision BNC connectors.

[ogden]

Just equip all your equipment with SMA adapters, use SMA cables. I don't even reuse "made in China" low cost SMA PCB connectors, just throw used testboard in the "afterlife box" as it is. You get bag of 100 edge SMA sockets for 10$ 20$. Try this with BNC.

[joeqsmith]

Test using the Nano, sweeping to 1GHz. 

IMG_4453, test setup
Nano_thru, showing the thru
Nano_MWM20dB,  DC-18GH MidWest
...

Also for fun, showing the MidWest's gain on my 1970's HP8754A sweeping to 1.3GHz.  Also shown is the surface mount attenuator.   Normally we would look at more than gain but I think this should give you some idea what to expect.


The Nano is a really nice tool for getting started with RF.