Attenuator DC-RF

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2008
Paid Member
I've got a couple of signal generator boards with 50Ω output, one that goes to 10MHz and the other to 65MHz.

The first can't output signal less than half a volt p-p, the other is better - down to 1mV but still not to zero.

Knowing pretty much nothing about higher frequency stuff I'm wondering if it's worth building my own variable attenuator with a few resistors and/or a pot or if it's better to buy a used HP box off the bay.

Any info, advice, or good stories welcome .

Thanks
 
There are virtually no RF attenuators that will go down to zero RF output. There will always be some RF leakage getting past the attenuator. It is usually possible to get a low enough signal voltage for your particular application, but those last few dB may be hard to get or expensive. Even the top of the line HP signal generators of today start out around a tenth of a microvolt.

I spent 41 years of my life as a tech / engineer in a Motorola plant that designed and made two way radios and cell phones. Ten years of that were spent in the test equipment lab. It was a continuous problem until well into the 2000's that we had to modify and shield even the best signal generators that HP / Agilent / Keysight made.

In the 90's we sent a whole batch of HP8656A generators back to HP because the RF leakage was so bad that the radio would pick up the signal generator without a cable connecting them together. They sent an engineer to work with us to fix the issue. It required additional shielding, and redesigned attenuators. Around 2000 HP adopted our shielding methods as an option on their high end RF generators.

The internal oscillator in most RF generators runs in the 1 volt or so range. The sensitivity of our radio equipment at that time was such that a signal of 0.25 MICROVOLTS was clearly received. The leakage issue gets worse as the RF frequency is increased and by the 90's we were making radios and pagers at 900+ MHz.

The Elecraft design shown has a maximum of 41 dB of attenuation. It will turn a 1 volt signal into roughly a 10 millivolt signal. It would be possible to build two, or even three of these and cascade them. For this to work, each attenuator would need to be housed in its own RF tight box with RF connectors on each end.

I just finished repairing several HP RF signal generators. The good ones, HP8656A's from the batch sent back to HP, scrapped when Motorola closed down, use two separate 0 to 70 dB attenuators, each inside their own box, with both housed in its own shielded compartment inside the generator. The ultra high spec HP8664A's have three attenuators, one inside the RF generator module, and two in separate boxes inside the RF output module.

Tell me a little more about your application, maximum RF frequency, minimum and maximum RF level needed, and we can figure something out.
 
Tubelab is correct as DF96 said. I do nothing with RF myself, but I needed an attenuator to attenuate a 10 MHz signal for some older HP volt meter calibration. I used the pi calculator to make an attenuator that worked for my application. I installed it in an aluminum enclosure and it suits me just fine.

HP made a bunch of RF attenuators with models numbered 355. Here is a Google return for you: https://www.google.com/search?q=hp+...ved=0ahUKEwjYk9Cew7rRAhXrqVQKHf24BRAQ_AUIBygC

You can find these on eBay for a reasonable price.

The HP 353a is a transformer coupled impedance matching attenuator. The actual attenuator section inside is a shielded unit and makes a great 600 ohm unit by itself.

The HP 350d is a 5 watt 600 ohm attenuator.
 
The attenuator shown in the Elecraft link will probably work as designed up to 200 MHz or so, but may require an enclosure to be accurate above 100 MHz or so. The switches and resistors used DO matter.

Some leaded resistors become inductors in the VHF range. Any attenuator I made used SMD resistors. They are good to at least 1GHz if the board is laid out with 50 ohm runners above a ground plane. Contrary to popular belief ordinary FR4 will work up to 2.5 GHz with minimal losses if the layout is good.

The tiny toggle switches found on some of the Chinese web sites are good to 1 GHz as are some of the Panasonic tiny (non RF relays). For better RF performance RF specific reed relays are best.

At frequencies above a few hundred MHz the enclosure needs to have a separate compartment for each switch, especially 5 dB and above. Otherwise the RF will leak right past the resistors.

SMA or type N connectors are preferred above 500 MHz.

I use one of these. It's good to 1.5 GHz and you can't make one for $60. I just used this listing for example. I don't know anything about this particular unit, since it's too easy to blow these things with a radio transmitter.

KAY 837 Attenuator PMAX 3W DC-1.5GHz 50 Ohms | eBay
 
Member
Joined 2008
Paid Member
Skidave,
Thanks for the link. I'd forgotten about Elecraft and it was interesting to see their stuff. The schematic was interesting as well.

George,
I was kind of hoping you'd appear. I actually had you in mind when I wrote the "good stories welcome" above.

The generator boards are from a Cleverscope. The CS700 is installed but there's also a CS701 that does sine wave only (though it's not installed and not sure if it works yet).
 
I looked at the specs for both units. maximum frequency is 65 MHz and both units have a BNC output jack. There are two easy paths here.

The easiest is to purchase a few 20 dB fixed attenuators from a reliable source. Use one or more at a time connected end to end, each one will reduce the output voltage by a factor of 10. One attenuator will reduce 1 mV to 100 uV, two of them, 10 uV, three of them 1 uV.....Brand new ones cost $10 each from Mini Circuits, model # HAT-20+ They are good to 2 GHz.

I used the SMA versions at work, and there were no issues with any of them unless someone screwed one of them into a 5 watt radio and transmitted through them. I did this on purpose to emulate weak signal testing, and even still they don't blow, they just lose accuracy.

The other option is a step attenuator like the KAY (or HP) unit I shower earlier. Again each 20 db will drop the signal voltage by a factor of 10. The switches will eventually wear out, but it took several years in a daily use situation (with the Kay units) but they don't like transmitters.
 
Member
Joined 2008
Paid Member
Thanks George.

The SMA's are clean and take little space but I know that if I have to put several in series I'll always end up standing there with one in hand, trying to figure out where I put the other two.

For a while I thought it would be cool to build my own until I saw this post on QSL.net where it shows the machined compartments for each stage of attenuation. That together with your story about the HP 8656A's put things in a different light.

If the Kay is what you use who would argue? There's a place that caries some surplus here in town I'll give a call tomorrow.
 
If the Kay is what you use who would argue?

I used the Kay attenuators at work because that's what I could get my hands on. When I started at Mot in 1973 there was a 60's vintage 500 MHz rated switch attenuator mounted to the communications analyzer in every test position in the factory (100+). After a year and a half of working in that factory hand tuning the old HT220 two way radios, I got a job in the test equipment lab maintaining that old stuff while building new test systems for each new radio introduced until 1984, when I got a job in engineering.

Equipment in the factory ran 24 / 7 in the early years. The techs were not particularly gentle on the equipment. Stuff broke, a lot. This was often because it was old junk. Those old Kay attenuators did not break much, and when they did, it was because a bat handle got broken off a switch, or someone transmitted into them for a while.

The later factories were more and more automated, and radios required less and less manual tuning, so the old equipment was scrapped including all those old attenuators.

After working in several different engineering groups and collecting 2 college engineering degrees, I got a job in the advanced development / research group at a time when several engineers had left. There was a room full of test equipment that "wasn't good enough" for those who were still there, and the old paging / cellular plant had been shut down, so I set up my own lab space with all the "cast off" equipment, including 6 of those smaller 1.5 GHz Kay attenuators. I never had an issue with any of them in the 12 years that I worked in that lab. I think that they were cast offs because they required manual switch flipping at a time when most engineers wanted automated workstations.

It's a 20$ kit but not sure if that would do the job up to 65M.

I doubt that you could really get 101 dB of loss at 65 MHz with all switches on without some additional shielding, but it might be good enough for some work depending on the absolute accuracy needed.
 
Last edited:
Member
Joined 2008
Paid Member
I made the seller of the Kay an offer and it arrived, clean and tidy. Unfortunately clean and tidy has nothing to do with function, and as you and the fellow at qsl have both said, it's the switches. (Thanks for the link, Frex.)

In this one's case there's a switch with variable attenuation - ie, different each time you switch it in - and not only that, it avoids being singled out by randomly trading places with its neighbors. Hmmm.

On the page here it says the switch actuators drop down onto board mounted contacts and if mine turns out to be the same construction (and I can get through the fabric the previous owner glued to the bottom plate - over the screws,) I'm hoping I can get it working properly by cleaning and perhaps "respringing" the contacts.
 
and I can get through the fabric the previous owner glued to the bottom plate

If it's a grey or beige felt material, they come from Kay that way.

Sorry to hear you are having trouble. I haven't opened mine up since it works good. The older green 500 MHz Kays with the large toggles have sealed switches that must be replaced as a unit. At least these look repairable.
 
Member
Joined 2008
Paid Member
If it's a grey or beige felt material, they come from Kay that way.

I soaked it in alcohol and it didn't give a millimeter. I thought the owner must have done it.
I'll try to go through just at the screws. It would be nice to leave the rest on anyway.


Sorry to hear you are having trouble.

It's not trouble. It's motivation. I'm lazy by nature and not smart enough for LTspice. Stuff like this is where most of the little I know gets found out.

I have a good friend who, whenever I'd get frustrated by obstacles to plug & play smoothness, liked to smile and say; "Well, looks like yer fixin tuhlern sumpm." Over time I've learned to say it to myself and enjoy it. What's harder is getting the time to actually get to it. I think most people here know something about that. :)
 
Stuff like this is where most of the little I know gets found out.

That's the way I learn stuff the best. I walked into a Motorola plant in 1973 and got a job assembling / fixing / whatever it took to make it work, the two way radios of the day as they came off the assembly line. Anyone remember the HT-220? About 400 people answered a full page ad in the Miami newspaper. The HR girl told the crowd that three people would be hired that day and two would be gone (quit or fired) within a month. Then she showed us what we would be working on (about 300 tiny leaded through hole parts on a 2 by 3.5 inch PCB). About half the crowd left. Three of us were hired......two were gone after about a month or so. I left after 41 years.

During that time I worked my way up to product design engineer with no formal electronics education other than a high school electronics class that taught tubes.

Most of what I knew, I learned by doing or watching. Now that we have internet, I read, and watch Youtube videos. After about 25 years, Motorola paid for me to get two college degrees. Only a small part of that material was actually useful on the job. It did help me a lot at knowing how to train new engineers that just graduated......Step one, This is a soldering iron.....keep your fingers away from this end! Seriously there is a lot of IMPORTANT STUFF that is NOT taught in school. Working with modern test equipment is one, SMD another, and so is common sense!

"Well, looks like yer fixin tuhlern sumpm."

When you stop learning.....you start dying.

Unfortunately I find myself relearning some things. Either I have forgotten, or the world has changed since I learned it. When I got my computer engineering degree a fast PC ran DOS on a 16 bit '286 chip at 16 MHz, and the Arduino wasn't even a dream yet. Now I'm learning how to program those things.

Learning.....I just bought a basket case Hewlett Packard RF signal generator on Ebay. Original cost in 1997, about $25K. My cost on Ebay, $200....Yes it was broken, and had been wet too. All that was disclosed up front.

Unfortunately UPS managed to bust it up some more. They smashed in the front panel breaking one of the glass VFD displays. I have the "service manual" from HP. There are no schematics, it's "too complicated" for component level servicing. They were repaired at the board / module level only. HP discontinued this generator some time ago and now offers no parts or service, so parts can only be obtained by stripping other units, which is why I bought this one, for parts......but it became a chance "tuhlern sumpm." At worse it would be a $200 lesson with lots of useful parts left over, I've done worse.

I stripped it down to bare chassis, cleaned each board, and all the ugly stuff, and even painted a few things, and in the process found and fixed some bad modules (by testing them in a working unit). Now, I have a functional unit that passes all the diagnostics, and is just waiting for a replacement front panel.......keep learning.
 

Attachments

  • P1080533_x.jpg
    P1080533_x.jpg
    585.3 KB · Views: 126
Member
Joined 2008
Paid Member
Well, as usual, I'm impressed. What are you going to use it for?

As for my little project, I have to thank you for putting me onto the Kay. I got it opened up last night and discovered my interpretation of the qsl photos was wrong. It was easy to clean the contacts with alcohol and I pinched the spring contacts together a little to make them stand a little more proud for better contact with the board.
I didn't find the source of the scoring on the board pads but did find a couple of finer than hair thin strands of aluminum (likely from machining the enclosure) lying on two of the actuator bodies, intermittently shorting contacts when they moved.

I found it interesting that the roughness I felt in the toggle movement actually turned out to be from the feel of the contacts, which are so delicate I wouldn't have thought they'd make that much difference.

Anyway, they all run smoothly and the attenuator as a whole works fine now!

Thanks George.
 

Attachments

  • Kay 837  board view.jpg
    Kay 837 board view.jpg
    403.5 KB · Views: 113
  • Kay 837  switch contacts .jpg
    Kay 837 switch contacts .jpg
    554.6 KB · Views: 111
Member
Joined 2008
Paid Member
One thing I missed before posting above was the change in pattern for the ground traces on the pi filter. Can you say something about how it works?
I'm guessing that on the left end the fat bars on the 30,30,and 20dB steps are shielding the two sides from each other (or is it capacitive coupling), but then why dropped on the 10dB and then after getting thinner for the right hand four (3, 2, 1 and 0.5dB) turning into a meander pattern. Is it all to tune out specific resonances?
 
The meanders are inductors. This sort of thing is very typical in microwave circuitry. Often you'll design in a lump on a component pad to add a little capacitance, or throw in a length of narrow trace to give some inductance. At even higher frequencies you can just go a quarter wavelength away and do the opposite.

I'm thinking the inductors balance out the parasitic capacitance associated with the pads for the shunt resistors on the low-value attenuators, presenting a more constant impedance with frequency.

Microwave low-pass filters are a good example - they're essentially just 50 ohm trace, fat trace, thin trace, fat trace, thin trace, fat trace, 50 ohm trace. Fat traces are shunt caps, thin traces are series inductors. Instant filter.

You can go quite nuts balancing every little parasitic out.
 
Can it be done in the 'on paper' design or is it by empirical testing of a built prototype?

Yes, certainly. There are lots of software packages available for microwave simulation. At the extreme end HFSS is a good example - it creates a 3d mesh of the design and uses Maxwell's equations to solve for the electromagnetic field. You just give it a geometry and material properties (and a whole pile of money) and it gives you a very accurate representation of your circuit.

There's also a huge amount of empirical knowledge to draw on for this stuff. Microwave circuits are just resistors, capacitors, inductors and transistors same as anywhere else, and every component has a model that you can design around. https://www.microwaves101.com is a good start for reading up on this stuff.

Even spice is very useful - just don't use idealised resistors and capacitors, but choose the proper models with parasitics baked in.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.