# attenuator

• posted

I would like to design a 50 ohm attenuator that I can adjust to have 3 attenuations of 1, 2 and 4. I would like it to be able to handle various types of signals (pulse, sine etc) with frequencies between dc and 250 MHz. Does anyone have any pointers? I have seen some circuits that use a ladder configuration that may be suitable, but I am unsure of how to make the output adjustable.

TIA

Jon

--------------------------------------- Posted through

• posted

First thing to do is to write an unambiguous specification. Does attenuation of 1 mean gain of 1 = no attenuation? OR 1dB or what? How much power does it have to handle? Anything that don't come out the end gets dissipated in the attenuator.

How accurate do you need the 50 ohms to be?

Do you care what happens to the input impedance and attenuation when the output load isn't exactly 50 ohms? Same for the source impedance when the input ain't from 50 ohms?

Do you care if it's symmetrical? Work the same way if you reverse the input and output?

Answers to those questions will narrow your options. Typically, you'll end up with a T or PI arrangement of three resistors. You now have a schematic with only resistors on it.

Problem is that real components have significant parasitics in the range of frequencies you contemplate and don't (yet) show up on the schematic.

What's your definition of "handle"? If the bandwidth is 250 MHz., the output will be attenuated by 3dB at

250MHz. That good enough?

It's relatively easy to build an attenuator that will pass sine waves up to 250MHz., within some bandpass ripple specification. With pure sine waves, your attenuation may be a little off, but it's still a sine wave.

If you expect to pass a 250 MHz. squarewave, without ringing or flatness issues, your problems multiply.

Attenuator vendors do it by minimizing the parasitic elements. You'll find rod resistors in series and disc resistors in shunt...all arranged in a shell that keeps the impedances matched throughout the attenuator.

If you're building it out of discrete components, you need to worry about the physical arrangements of the components. For example, for low power attenuators, you may be able to use surface mount resistors. You might find that two 100 ohm resistors, one on each side of a 50 ohm microstrip line may have better transient response than a single 50 ohm resistor.

You probably don't want to use wire wound resistors for higher power. Although I once used a wire wound resistor because I experimentally determined that it had exactly the right ratios of resistance, capacitance and inductance to critically damp the output of a pulse generator.

We haven't even started to talk about switching. You don't want transmission line stubs hanging around when the switch is open. A SPST switch can be made by enclosing a reed relay in a copper cylinder of the right diameter.

If you can get by without the switch, I suggest you build three separate attenuators for your first foray into this realm. Do you have the signal sources and measurement apparatus (oscilloscope) with specs better than you want for your attenuator so you can tell if it works?

Attenuator design is more art than science.

• posted

Several issues here.

To handle frequencies up to 250 MHz, and maintain something close to a

50-ohm characteristic impedance (i.e. not introduce reactance and raise the SWR) you are going to have to pay very careful attention to component selection and layout. Even a relatively small amount of lead inductance could add enough reactance to notice, up at the higher frequencies. It may be necessary to use surface-mount components and microstrip construction techniques.

I've got some switchable attenuators designed for this sort of use at home - bought them surplus last year - I don't know what their useful frequency range actually is. They're hefty beasts - silver-plated brass cylinders a couple of inches long, with BNC jacks on opposite sides of the cylinder wall, and a big switching mechanism which rotates inside the cylinder. It appears to me that they actually rotate a whole different "T" attenuator into place for each desired attenuation, using big silver-plated contacts... the attenuators not being used are actually rotated away from the signal path by an inch or more to minimize capacitive coupling. In other words, building something like this is a job for somebody with a machine shop and some serious metal-working experience.

A simpler-to-build approach along the same lines uses simple miniature DPDT toggle switches, to switch in or bypass a series of T attenuators. If you want high attenuation levels to work reliably it's important to have signal isolation between the switching stages (e.g. put them in separate compartments, or separate them via shields). For smaller attenuation levels such as you are proposing, that's probably not necessary. I'm sure there are projects along this line documented in the ARRL Handbook, QST Magazine, and all over the net... ham-radio "fox-hunters" often use such switchable attenuation boxes.

You could probably build an electrically-switched discrete-attenuator design on a medium-sized PC board. What I think you'd want to do, is lay out three or four separate T-attenuator circuits (three resistors each) using microstrip layout techniques, and then use tiny little DC-operated reed relays to connect precisely one of these to the common input and output points at a time. There would be a little bit of capacitive leak-through into the unused attenuator sessions but it might not be enough to matter.

Or, use the modern approach. Go to Mini-Circuits, and buy yourself a "digital step attenuator" IC. For under \$5 you can buy an IC which will let you select attenuation in 1 dB steps from nond down to 30 dB or more, and works from DC up to 2.4 GHz or above. Add a PIC or similar micro to read pushbuttons and send the attenuation-selection codes to the chip, and you're good to go.

```--
Dave Platt                                    AE6EO
• posted

Here's a good one:

Watch the wrap.

• posted

Watch the red flag: "We do not have the necessary resources in our facility to test this unit." That's often code for, "it's broke but we are betting that you don't have the resources to test it either."

I'd also watch the 1/2Watt power limit. May not be a problem for you, but one of the last ten people who owned it might not have been so gentle.

"Non-functional items may be returned for a replacement, when available, or a refund within 30 days of our ship date", suggests that your risk is limited to \$12.35 shipping...twice...If you're in the UK, that number might be much larger.

If you need it NOW, ebay may be your only option. If it's a hobby, I'd contact a local ham radio club. Somebody may sell you one and show you how to use it, verifying it works. Or they might help you build and test one.

• posted

Yes, but also "The item may have some signs of cosmetic wear, but is fully operational and functions as intended. This item may be a floor model or store return that has been used." ISTM _that functions as intended_ means it meets manufacturers specs. But, then again, I somewhat of an idealist.

The listing was just the first one of many. Guarantees of performance can be had if the scratch is available.

You are absolutely correct. The first thing I did was test mine when they arrived.

They are in San Jose. But, again, there are quite a few listed.

That's true, too. Maybe even pick up a book or two from

If this is a learning experience, then I encourage going with your suggestion.

• posted

Ok, but the OP was from jbean@n_o_s_p_a_m.beandigital.co.uk

• posted

Be careful about what they mean by "DC." Most of these attenuators are spec'd with coupling caps on the input and output, with any real DC specs missing.

Some people are being honest and using "LF" instead of "DC."

John

• posted

I missed that. Sorry.

• posted

various

MHz.

find

Sorry for the vague spec. I need an attenuator that divides the incoming signal by 1, 2 or 4 and the 250 MHz will be the 3dB bandwidth of the attenuator. I am mostly concerned about the attenuator passing square waves. I have seen some attenuators that you can get in an IC package, but wasn't sure if these are only suitable for RF type signals. I really need something small that will fit onto a circuit board and that I can select the attenuation factor with a digital signal.

Jon

--------------------------------------- Posted through

• posted

I may have overstated the complexity of the design. For a general purpose attenuator, you want the input and output impedances to be purely resistive independent of mismatch on the input or output. You want a bandwidth at least 5X the maximum frequency square wave you want to put through it. You can't stand any feedthrough at high frequencies. For sinewaves, bandwidth pretty much describes the problem. For transients, risetime and aberrations are critical. This can be modeled by phase shift vs. frequency, but it's easier to evaluate in the time domain.

Input minimum risetime is a critical specification. Bandwidth limiting the input makes the attenuator design easier. You need to model the attenuation component parasitics, including the circuit board and relays, and arrange lumped or distributed elements that make the total into a low pass filter. The filter characteristic is described as bessel, gaussian, or maximally flat envelope delay. You can design the filter in the frequency domain, but test/tweak it in the time domain.

There are a zillion other potential gotchas. If you're attenuating some signal that needs to maintain a constant phase relationship with some other signal, you might find that constraint dictates your design topology.

In your case, you seem to have a controlled environment. If you need only 250MHz. bandwidth, you're talking about square waves 50 MHz. or less, with risetimes that make 'em look more like sinewaves than squarewaves at 50 MHz. At those frequencies, you can get by with much higher level of parasitics.

It is possible to get by with relays that you'd think would never work at. The attenuator for a number of

500MHz. TEK oscilloscopes uses rather large relays. They stand them up on their leads in a puddle of solder paste on a micorstripline. I would have bet money that it'd never work. I'da been wrong. Problem is that a typical relay is not specified for that. You have no guarantee that the same part number purchased next week will have the same high frequency characteristics as the one you got yesterday. Not a problem if you're only building one. I can't count the number of times I got burned when the purchasing department switched vendors on me and the production line went down. If you think you can fix that on the spec, think again. If you spec'd what you really needed, you'd not get any takers at a price you could afford.

If money is no object, you can likely buy what you want from Mini-Circuits as suggested elsewhere in the thread.

The following is a RANT on engineering...

I used to do system design reviews. Typically, somebody has a concept of the system. They do the top level architecture and partition it into modules. They may have no clue about what it takes to do some module. The engineer who gets assigned the module does his best to do what he's asked.

If the system designer had a better idea of the cost and difficulty of designing a seemingly trivial switched attenuator, he might have modified the architecture.

I encourage engineers to push back so the architecture and partitioning can be reviewed for lowest total cost/effort. Tunnel vision is not always optimal. Not every system engineer is a genius.

• posted

maxascent a écrit :

Use that

6dB atten. 6dB atten. 6dB atten.

o-----------o o----------o o----------o ---o--__ ___ __--o--o--__ ___ __--o--o--__ ___ __--o--- o-+-|___|-+-o o--|___|-+-o o--|___|-+-o | R2 | | R2 | | R2 | .-. .-. .-. .-. .-. .-. R1| | R1| | R1| | R1| | R1| | R1| | | | | | | | | | | | | | '-' '-' '-' '-' '-' '-' | | | | | | GND GND GND GND GND GND

R1 = 150 R2 = 37.5 ( 2x75 paralleled)

```--
Thanks,
Fred.```
• posted

Fred Bartoli a écrit :

Ooops! That one will give you 1,2,4,8 attenuation. Loose the last section.

```--
Thanks,
Fred.```
• posted

```--
http://www.minicircuits.com/pdfs/DAT-31-PN.pdf```
• posted

I would rather use an IC if I can as I think it would be a lot less hassle. But does anyone know what the step response of these attenuators are like? I would imagine that the majority of people are using these for RF applications.

Jon

--------------------------------------- Posted through

• posted

-3dB at 250Mhz isn't that hard with surface mount resistors and low power. If you want to handle watts as opposed to milliwatts, then it becomes tricky and expensive.

Switching is where it gets problematic. Take a look at the way oscilloscope attenuator switches are designed, even in instruments as low as 100MHz BW. Switching from a digital signal, my first reaction would be to use those nice subminiature RF relays that (IIRC), TRW make. They're expensive, however.

Resistor values aren't difficult:

50 ohm in and out, divide by 2 (-6.02dB) gives for a T section: R1=16.67 ohms, R2=66.67 ohms, R3=16.67 ohms. (theoretical) Nearest E96 values 30.1, 66.5, 30.1 respectively.

And for a Pi section: R1=83.34 ohms, R2=93.74 ohms, R3=83.34 ohms. (theoretical) Nearest E96 values 83.5, 93.1. 83.5 respectively.

For divide-by 4, just cascade two divide-by-two sections.

I wrote a spreadsheet, many moons ago, to calculate matched attenuator values. If you have access to alt.binaries.schematics.electronic, I can put a copy on there.

```--
"For a successful technology, reality must take precedence
over public relations, for nature cannot be fooled."```
• posted

My bad - that's for -12dB (divide by 4)

-6db is 150.01 - 37.5 - 150.01

Sorry.

```--
"For a successful technology, reality must take precedence
over public relations, for nature cannot be fooled."```
• posted

I don't believe that you've stated the impedance for the attenuator. Is there one? What is it? 50 ohms, to match RF practice? 75 ohms, to match video practice? 200 ohms or some other measure, to match the traces on your board? Does impedance matter? Can you survive having some loss in the nominal "divide by 1" case (i.e., can you start with a signal that's a bit extra strong?)

If you do want controlled impedance, two 6-dB pads (I assume that when you say "divide by two" you mean "divide voltage by two", not "divide power by two"). For 0dB of attenuation bypass both pads; for 6dB bypass one, for 12dB go through both.

You could do this with a multiplying DAC if you can find one with enough bandwidth (which is iffy).

If you could stand some attenuation in the "no attenuation" position then you could do it with analog switches to bypass the attenuators. Otherwise you'd need those relays that are mentioned.

You could use a multiplier.

If you actually told us what you're trying to do then your question will resemble a technical question more than it does a Rorschach blot.

```--
www.wescottdesign.com```
• posted

He did - 50 ohms.

```--
"For a successful technology, reality must take precedence
over public relations, for nature cannot be fooled."```

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.