20dB 'RF' attenuator - seeking feedback to improve

[joeqsmith]

I have the 3GHz test set for the 8754A and a Mini-Circuits doubler that allows the system to reach a little over 2.6GHz. 

In the graphs the blue is phase and red is the gain. 

[joeqsmith]

In these plots I have changed it to 1dB/div and compared the homemade surface mount against the Midwest part.

The 8754A has basically no intelligence and there is no way for the software to know how the instrument is setup so the software doesn't rescale the graphs like with the Nano.    To make this measurement, the thumb wheel switch is changed to give it a 20dB offset. 

If you wanted to look at say a 50dB attenuator with the Nano at a GHz, you will be out of luck.   The Nano doesn't have a lot of dynamic range and is pretty noisy, but it's $50 so it's pretty hard for me to fault it.   The Nano V2 is supposed to be available soon.   This may be a better system for the hobbyist.

[enut11]

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.

Thanks @ogden (and @ Berni). I have purchased a NanoVNA online from a local Sydney dealer. He claims to sell the genuine model. Apparently there are [inferior?] clones available on eBay.

I will also bow to the collective wisdom of Forum members and order some SMA edge connectors and cables.

Thanks to all who have taken the time to reply to my request. This Forum is the place to gain rapid knowledge. 

[enut11]

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.

Thanks @joeqsmith. Lots to read and absorb in your post.
Very impressive set of test gear! BTW, I just found your Youtube review on the NanoVNA. Thanks for that.
Where did you get those small PCBs with the the SMA connectors. Looks like what I need to build my next 20dB attenuator.
enut11

[joeqsmith]

Thanks @joeqsmith. Lots to read and absorb in your post.
Very impressive set of test gear! BTW, I just found your Youtube review on the NanoVNA. Thanks for that.
Where did you get those small PCBs with the the SMA connectors. Looks like what I need to build my next 20dB attenuator.
enut11

Those are a custom PCB but I have to believe you could find something stock by now.   

I think you will be impressed with the Nano's performance and what it can tell you about your circuits.   

I would like to get something a newer that can be controlled similar to the Nano, and make some sort of comparison video between the NanoVNA, the Nano V2, the HP relic and what ever I end up with to replace it.
 

[ogden]

Those are a custom PCB but I have to believe you could find something stock by now.   

Such a simple layout can be made from double sided CC PCB using exacto knife or dremel or nail lacquer + etch_solution in comparably short time.

[fourfathom]

Those are a custom PCB but I have to believe you could find something stock by now.   

Such a simple layout can be made from double sided CC PCB using exacto knife or dremel or nail lacquer + etch_solution in comparably short time.

I sometimes make boards like this using a small woodworking v-gouge to cut between the copper traces.  Use a thick steel straight-edge to guide the gouge and be sure to cut *away* from your fingers.  Or just freehand it, but do be careful.

[joeqsmith]

I put together a 50dB attenuator using 0603s & 1206s built on the same section of microstrip.   I set the Nano to sweep to 1.5GHz and installed a Midwest Microwave 18GHz, 10dB attenuator.    I then installed the home made attenuator.  +/-25dB isn't very good but is it my circuit or is it the nano.   It's easy to measure 10dB but 50dB may be pushing the $50 pocket unit.

[joeqsmith]

The old HP8754A is a bit odd in that it uses eight vertical divisions rather than the typical 10.  To measure down this low, I need to add a 40dB offset  to make the top line 0dB.      The doubler is still installed and the instrument is sweeping to 2.6GHz.   

Installing the Midwest Microwave 10dB attenuator, we can see it is one division down or 10dB and is basically flat,  as we would expect. 

Now let's have a look at our home made 50dB attenuator.   I suspect the old HP is going to be much closer to reality than the Nano but it's only $50 and for that, a great learning tool.   

[Berni]

50dB attenuators do get a lot trickier.

The resistor values start getting very far from 50 Ohm so the parasitics of those resistors become much more visible. But also leakage can be a problem.

I did make one once when i needed a high value attenuator i didn't have on hand and ended up making it on the PCB as 3 stages of attenuators in series so that each one didn't have do do quite as much attenution. Giving it shielding made from thin sheet metal also improved the response since at these low levels any sort of leakage can be seen making its way across. Still was not quite all that flat tho so i suggest against DIYing attenuators this strong if you want it ruler flat.

A nice thing that is easy to DIY however is a resistive 3 way power splitter/combiner. Made one by soldering 3 SMAs into a triangle and joining then together in the middle with a star configuration of the right value 0402 SMD resistors. It seamed to work great from DC to 3GHz (Didn't have gear to measure higher).

[joeqsmith]

50dB attenuators do get a lot trickier.

The resistor values start getting very far from 50 Ohm so the parasitics of those resistors become much more visible. But also leakage can be a problem.

I did make one once when i needed a high value attenuator i didn't have on hand and ended up making it on the PCB as 3 stages of attenuators in series so that each one didn't have do do quite as much attenution. Giving it shielding made from thin sheet metal also improved the response since at these low levels any sort of leakage can be seen making its way across. Still was not quite all that flat tho so i suggest against DIYing attenuators this strong if you want it ruler flat.

Good points.  My 50dB example used two stages (lazy) which is really pushing things.   My goal was more to show the differences between the Nano and my old HP VNA as we start working with lower signal levels. 

Op, you may be interested in the following article on construction.   It's very short and provides some practical information. 

https://www.electronics-notes.com/articles/radio/rf-attenuators/construction-practical-tips-guidelines.php

I'll try to make something a little better using a PI type and more stages and compare it against a commercial grade part.

[ogden]

My goal was more to show the differences between the Nano and my old HP VNA as we start working with lower signal levels. 

When you challenge noise floor of instrument, you better demonstrate it. Now it is impossible to know - poor isolation results are due to leakage through attenuator or instrument (NanoVNA) itself. Yes, we see better results with HP VNA, yet showing "open" S21 curve of your NanoVNA would be much more telling and educational.

[joeqsmith]

PIC1
Three stage PI configuration made with 0603 & 1206 parts, with shield.
 
PIC2
The S-parameter test set on the 8754A includes a 70dB step attenuator (manual of course).  This was set to 50dB and a thru was installed.   As previously, thumb wheel switches set for a 40dB offset.  The top of the graph is 0dB.  Gain is 50dB.   

PIC3
The internal step attenuator was set to 0 and the thru was replaced with a JFW attenuator.

PIC4
The home made three stage attenuator was installed.  At this time, no shielding was added.   Maybe 42dB at 2.6GHz.   

PIC5
A simple shield was added.  Maybe 45dB at 2.6GHz.   A noticeable improvement. 

PIC6
The shielded attenuator on the NANO, sweeping to 1.5GHz.


It appears to be a significant improvement in the flatness compared with the 2-stage one.  Shouldn't be too surprising. 

[enut11]

Op, you may be interested in the following article on construction.   It's very short and provides some practical information. 

https://www.electronics-notes.com/articles/radio/rf-attenuators/construction-practical-tips-guidelines.php

I'll try to make something a little better using a PI type and more stages and compare it against a commercial grade part.

Thanks. That is a good introduction to RF and attenuators.
enut11

[joeqsmith]

Glad to help.   Another pretty good resource is the ARRL Handbook.   Even the very old editions may be helpful.  They used to have a good section on construction practices.   

Just for fun, I made one last attenuator.   This time it's a 4-state T.    I set the 8754A to sweep to 200MHz.  One plot is 10dB/div, the other 1dB.   You can see a slight rise to it.   I then set the Nano to also sweep to 200Mhz.    You can see the two systems are fairly close when looking at the lower frequencies.   

Good luck with your projects.  Let us know how you like the Nano after you have played with it for a while.   
 

[enut11]

Hi @joeqsmith, @ogden and @Berni
The NanoVNA arrived today. Quick delivery from Sydney. Just connected it to the PC to charge the battery.
Switched it on and found it a little daunting!
I admit that I am out of my depth at the moment.
Looking around for PC interface software and any other info to soften the learning.

From what I have read so far, my next (SMA based) 20dB attenuator is better be done in stages, say 10dB +10dB.
Is there anything to choose between T and Pi arrangement of resistors?
enut11

[ogden]

Few cents: in pictures from @joeqsmith we see PCB edge-mount male connectors, male-female cables. That is very unusual approach - due to surplus/scrap/misordering(?). Better & widespread approach: female SMA receptacles on instruments/boards/devices and male-male SMA cables. Only case when you would want to use PCB edge-mount SMA male+female connectors would be small RF accessories like: attenuator, DC block, Bias tee and (maybe) LNA.

[tkamiya]

I just wanted to say:

Getting certain attenuation value is relatively easy but getting flatness across wide range is half art, half science.  HP attenuators for example, uses copper "flaps" to purposely generate stray capacitance and they are hand tuned.  Also, registers like that aren't inductance free.  Often carbon/metal coating on ceramic core is cut in spiral to establish values.

Also, OP's construction has way too much stray inductance and capacitance.  I don't think suspending resisters in air is bad thing but choice of connectors are not optimum.  If BNC, center conductor attach point should be much shorter, and I don't mean resister leads.  Internal to that connector is a long enough lead wire between gold plated brass center pin and where it comes out of the metal shell.

Argh....  I should stop reading every post on EEVBLOG!  Now I want to replicate this experiment myself and see how I do!

[joeqsmith]

I just wanted to say:

Getting certain attenuation value is relatively easy but getting flatness across wide range is half art, half science.  HP attenuators for example, uses copper "flaps" to purposely generate stray capacitance and they are hand tuned.  Also, registers like that aren't inductance free.  Often carbon/metal coating on ceramic core is cut in spiral to establish values.

Also, OP's construction has way too much stray inductance and capacitance.  I don't think suspending resisters in air is bad thing but choice of connectors are not optimum.  If BNC, center conductor attach point should be much shorter, and I don't mean resister leads.  Internal to that connector is a long enough lead wire between gold plated brass center pin and where it comes out of the metal shell.

Argh....  I should stop reading every post on EEVBLOG!  Now I want to replicate this experiment myself and see how I do!

Normally I turn to foil tape and razor blades or magnet wire stubs to trim them.   Shown is a small load with a tap that I wanted some decent flatness.   

https://www.eevblog.com/forum/projects/home-made-pa-and-load/msg635851/#msg635851

I modified a cheap DC current clamp a while back to increase it's bandwidth.  That thing was a nightmare to get a decent response out of and I was only running at a 100KHz or so.     

[tkamiya]

Very nice, indeed! 

[ogden]

Accidentally stumbled on this $1.33 (+0.88$) set of attenuators. I doubt it can reach 4GHz, but whatever:

https://www.aliexpress.com/item/4000333427604.html

[joeqsmith]

This is the 4-stage T attenuator after a few mods.    There are 18 resistors on the board with a layer of Kapton, then some copper foil strips that I trimmed very carefully.   It's well within a half dB up to 1.5GHz.     

[EEEnthusiast]

Accidentally stumbled on this $1.33 (+0.88$) set of attenuators. I doubt it can reach 4GHz, but whatever:

https://www.aliexpress.com/item/4000333427604.html

How can these guys sell them for $1.33 with 8 of those SMA connectors.

[tkamiya]

I have a many cheap Chinese made no-brand-name SMA connectors.  The "Gold" plating is so thin if I attach anything and remove it, metal underneath will show.  Then it will start to turn black and corrode.  I don't think plating is actually gold. 

That attenuator is cheap even by China's standard.  I have to wonder what's going on....

[joeqsmith]

It seems they are on eBay and Amazon for about 5$-$15 USD.  Maybe someone will test what you get for your cash...    I would expect its like a Chinese 1000V CAT IV meter with glass fuses or a 5000mAh 9V transistor battery. 

My home made ones are good to 8GHz.        

https://www.ebay.com/itm/DC-4-0GHz-23dBm-RF-Fixed-Attenuator-SMA-Double-Female-Head-0dB-10dB-20dB-30dB/174018514007?hash=item28844fcc57:g:qwkAAOSwBbBdbshD

A different style
https://www.ebay.com/itm/23dBm-DC-4-0GHz-RF-Fixed-Attenuator-SMA-Double-Female-Head-0dB-10dB-20dB-30dB/173343871488?hash=item285c199200:g:8q8AAOSw7m9bEB9s

******
Just looked at the second ad.   So 30dB with a flatness of 1.5dB up to 4GHz. 

Direct: 0dB (reference)
Level 1: 10 0.8 dB
Level 2: 20 1.1 dB
Level 3: 30 1.5dB

*****
And ... Standing wave ratio: 1.20

Made another attempt to bring it in tighter.   Within 0.2dB.   Here is the gain and SWR.