60 db attenuator help

[robrenz]

I need a BNC 60db attenuator for some calibration I would like to do. I want to take my FG output down by a factor of 1000.  I have found this minicircuit attenuator that is very inexpensive ($9.95)  I currently only need 600kHz so the specs seem extremely good  down there. I am thinking of using 3 of these since they don't sell any 60db attenuators. Is there a better way to do this for this price/accuracy?

EDIT: I was thinking of making my own but the parts would cost more than this and would probable not be as accurate.

[PA0PBZ]

If you can live with the maximum input power of 0.5W I see no reason not to do it, $30 for a switchable 0/20/40/60 db attenuator seems not too much.

[alm]

I would agree. A DIY design will likely perform worse, especially at frequencies beyond DC (eg. 2 GHz ). You're unlikely to find a 50 Ohm 60 dB step attenuator for under $30.

[robrenz]

I need it to get a 3µVrms 600kHz max signal to calibrate My 3410A on that full scale range. So the input would be 3mV rms from the FG thru the three 20db attenuators thru a 50ohm feed thru terminator into the 10Mohm input impedance of the 3410A. 

If I am thinking straight 0.5W would limit me to 5Vrms input.  But that is enough for me to measure the attenuation ratio more precisely by measuring the input rms from the FG and then the output rms after the 3 20db attenuators. 5mVrms is the bottom of the non de-rated specs of the 8846a so it just makes it.  If  I calculated correctly measuring the 5Vrms input and the 5mVrms output using the 8846A worst case errors in the worst combinations would still allow characterizing the attenuation ratio to +/-0.98% which is good enough for what I am doing.

Would I get a higer quality output from my Rigol DG1022 at 30mV vs 3mV?  If so should I stack 4 of these to get my 3µV from 30mV input?

Thanks all for the help

[blackdog]

Hi robrenz, :-)


I hope this excel sheet can help you.
Test the attenuator with DC, thats ease and it wil be the same for your 600Khz if cleanly built.

www.bramcam.nl/50-Ohm-2watt.xlsx

Kind regarts,
blackdog

[edpalmer42]

Instead of spending $30 on three fixed attenuators, you might consider putting the money towards a variable attenuator.  It seems like I'm always looking for some value or another and the variable ones have really been a lifesaver.  Choose continuous, stepped, variable, or programmable depending on your needs.

Ed

[robrenz]

Hi robrenz, :-)


I hope this excel sheet can help you.
Test the attenuator with DC, thats ease and it wil be the same for your 600Khz if cleanly built.

www.bramcam.nl/50-Ohm-2watt.xlsx

Kind regarts,
blackdog

With DC I could use ratio measurement directly and it would be much more accurate than AC.  Are there no AC impedance effects at 600kHz?

Thanks, the spreadsheet is very nice but I think I am after voltage ratio attenuation not power.  Is there one for voltage?

[robrenz]

Instead of spending $30 on three fixed attenuators, you might consider putting the money towards a variable attenuator.  It seems like I'm always looking for some value or another and the variable ones have really been a lifesaver.  Choose continuous, stepped, variable, or programmable depending on your needs.

Ed

A quick look turns up nothing close. If you know of something for $100.00 or less I might be interested.

[edavid]

1. Please fix the thread title (6o vs. 60).

2. You are not going to get 60dB attenuation out of 3 20dB attenuators... there will be too much leakage.  (Actually, any BNC attenuator may have too much leakage to get an accurate 3uV.)

3. You might want to shop for a used HP 355D:

http://www.ebay.com/itm/321175020424

[edpalmer42]

Instead of spending $30 on three fixed attenuators, you might consider putting the money towards a variable attenuator.  It seems like I'm always looking for some value or another and the variable ones have really been a lifesaver.  Choose continuous, stepped, variable, or programmable depending on your needs.

Ed

A quick look turns up nothing close. If you know of something for $100.00 or less I might be interested.

It never fails that when you want it, there's nothing that's quite what you want for the price you want to pay.  Anyway, here are a few that looked interesting for less than $100.  I have no connection to any of the sellers.

http://www.ebay.com/itm/221214145399
http://www.ebay.com/itm/370867531133
http://www.ebay.com/itm/190868257352 - only 50db.
http://www.ebay.com/itm/290908411116 - not sure about this one.  Maybe more research.

Ed

[robrenz]

Believe it or not I forgot about checking Ebay. Thanks, that has me looking in the right direction.

[Mechatrommer]

With DC I could use ratio measurement directly and it would be much more accurate than AC.  Are there no AC impedance effects at 600kHz?

from my little experience, i can hardly see any high frequency effect at this range (600KHz pretty much DC) even building from cheap material (except maybe wirewound type material which i believe you wont consider) but then how about DC power received by the attenuator? i guess it will be different from AC, temp rise could be different hence drifting the attenuator. the challenge with DIY is to properly and confidently specify it, ie requiring top level instruments.

i found this link descriptive, it has formulas and resistive values for common type attenuators. i hope it may help (if you havent found it)...
http://www.allaboutcircuits.com/vol_3/chpt_1/7.html

i just realized w2aew also made a video using SMD for RF work (havent watch, just downloading it now. afaik smd is worse than carbon layer)
https://www.eevblog.com/forum/beginners/new-video-basic-rf-attenuator-design-construction-and-testing/msg231040/#msg231040

and i've teared down one of those cheapo attenuator to reveal the carbon layer used... your link may has different layer network.

in https://www.eevblog.com/forum/beginners/replace-carbon-layer-with-passive-resistors-components-in-attenuator/msg177871/#msg177871
i'm not sure if the carbon material and proper fabrication method is available for diy purpose.

[robrenz]

Thanks Mecha,  I did see w2aew's video on attenuators but now that I see the step attenuators on ebay and since I can do a very accurate voltage ratio measurement for whatever setting I am using to compensate accordingly on the calibration I think I am going to go that route.

[vk6zgo]

1. Please fix the thread title (6o vs. 60).

2. You are not going to get 60dB attenuation out of 3 20dB attenuators... there will be too much leakage.  (Actually, any BNC attenuator may have too much leakage to get an accurate 3uV.)

3. You might want to shop for a used HP 355D:

http://www.ebay.com/itm/321175020424

I'm dubious about leakage being a big problem at 600kHz.
Certainly at VHF/UHF,but stray capacitances are going to have pretty large reactances at such a low frequency.
Radiation from lack of shielding?---maybe !

[G0HZU]

For the cost of the connectors plus maybe 20 cents for the SMD resistors you could make a low power attenuator that would be more accurate (at 600kHz) than the three cascaded MiniCircuits attenuators even if you visited Minicircuits and could test and pick out the 'most accurate' three Minicircuits 20dB attenuators from maybe 50 examples they had in stock on a shelf somewhere.

But if you just want the accuracy/uncertainty quoted on the Minicircuits datasheet then it will probably be fine.

[G0HZU]

If the meter works up to 600kHz then maybe you should calibrate it at lower frequency because the accuracy will be degraded at 600kHz.

I'd be tempted to make an attenuator from 3 x 20dB sections and solder them together in a chain and include the 50R termination at the far end. Then you only need one BNC connection at the meter itself.

Otherwise you could easily end up with a snake of BNC connections and you might get frustrated if the meter indication fluctuates as you wiggle the snake of BNC connections.

Also, it's worth checking the FG has lowish odd order harmonic content eg -20dBc or better or this will degrade the accuracy of the meter. I'd also be tempted to use another attenuator at the FG output in order to establish it has a decent 50R source impedance. So run it at a higher level and have a fixed attenuator on the output.

[robrenz]

I made and offer on a step attenuator but I think I may just do my own with this design.  Use a Pi attenuator design because the values are more friendly for 60db in one step using two 50.1 ohm and a 24999.975ohm. The resistor tolerance effect on the attenuation is less sensitive on the Pi also.  Put a 25 turn 100 ohm cermet trim pot paralled with a 100 ohm metal film for the two 50.1 resistors Do the same with a 50k cermet and 50k metal film for the 24999.975ohm. That setup gives a 4:1 adjustment sensitivity improvement and increased wattage.  Set the resistances after assembly to the nominal values with the 8846A measured at the bnc connections.  All inclosed in a full metal bnc in-out box.  Then  feed an exact 10V from a Power Designs supply to the input and adjust until exactly 10mV output.  Worst case error of the 10V is 29ppm and 387ppm for the 10mV.  Please comment if I am missing something.

The actual 3410A procedure for adjustment is done initialy at 400Hz which is in the most accurate +/-3% area.  Then you check to see if the 600kHz is within the +/-5% tolerance. all frequencies are checked and you try to balance the errors over the full frequency range.

The harmonic distortion of the DG1022 is supposed to be <=55dBc in the range I would be using. THD DC to 20 kHz,1Vp <0.2%  Spurious non harmonic DC to 1 MHz < -70 dBc  The 3410 is relatively immune to even harmonics and somewhat effected by odd harmonics depending on the phase.

I plan to use my 8846A as a level meter to monitor the AC rms output of the FG as it goes into the attenuator. That is part of the procedure to keep the output leveled as the frequency is changed. The 8846A specs are only good to 300kHz but that is better than nothing. Stated typical error at 1MHz is 30%.  My interests would be below 300kHz anyhow so I am not concerned about it. 

How can I tap into the FG output as it enters the attenuator with the 1Meg input impedance of the 8846A without disturbing the attenuation?

[Mechatrommer]

How can I tap into the FG output as it enters the attenuator with the 1Meg input impedance of the 8846A without disturbing the attenuation?

just tap it? you are tapping 50ohm impedance circuit, so tapping in a 1Mohm circuit should affect it by 0.005% of deviation in accuracy, am i correct?
add: but hypothetically to become more "T&M'phobic", i think you need the input capacitance too of your 8846A for anything other than 0Hz

[robrenz]

Spice model of proposed attenuator. R7-C2 represent the 8846A input impedance (not connected) and R6-C3 represent the 3410A input impedance (connected).  FG output will enter thru a female bnc connector and I will provide a female bnc "input tap" connection for the meter that is represented by R1-C1. when the meter is moved to measure the output (meter shown as R7-C2). A dummy load of 1Meg shunted by 100pF will be in its place on the input tap to keep the loading the same while measuring the output. R3-R4-R5 are the 60db attenuator and R8 is another paralleled 100 ohm pot (not shown)and resistor for a built in adjustable 50 ohm termination that feeds the male output bnc connection. The attenuation will be adjusted by setting the input to an exact value like 10Vrms and then the meter will be moved to the output bnc and the dummy load will be put in its place while tweaking R8 until the output is 9.9988mV.  If  my spice modeling is correct at 600kHz that value will give 10mV when the R6-C3 input impedance of the 3410A is connected. This will give me only one signal cable from the FG to the attenuator and the attenuator will attach directly to the 3410A bnc input.

criticism welcomed

[Jay_Diddy_B]

Robrenz,

I have had a look at your proposal for a 60dB attenuator. My main concern is the 24.999K resistor.

I have put together a LTspice model that allow the comparison of 3 cascaded 20dB with a single 60dB attenuator.


I added C7 to the 60dB design to illustrate the effect of stray capacitance.





I have attached the LTspice file.

Regards,

Jay_Diddy_B

[robrenz]

Thanks Jay_Diddy_B, I will study that tonight and then ask some more questions.

[GK]

A standard "RF" technique is to cascade "Pi" pad attenuators.

For example, see the "Example 1" circuit here:

http://www.eetasia.com/ARTICLES/2001OCT/2001OCT08_AMD_RFD_AN1.PDF


That's a 6dB attenuator that presents a 50 ohm load to the source when loaded with 50 ohms. Put 10 of them in series and you will still have 50 ohms z-in / z-out with 60dB of attenuation without any bandwidth issues (due to the low resistances used throughout) for most applications. If you want to get really fancy you can use DPDT switches for each series-connected pi pad to switch it either in or out, thus setting the attenuation in 6dB steps.

 

[robrenz]

Ok, I see I was not considering the higher thermal noise of the 25k resistor but I cant find anything that says the parasitic capacitance would be higher in a high resistance. Won't the 247.5 resistors have parasitic capacitance that needs to me modeled also?

[alm]

Imagine (or plot in something like LTspice) the impedance of 250 ohm // 1 pF and 25k // 1pF as a function of frequency. At what frequency does the impedance start to drop say a factor two for each of the combinations? At what frequency does the reactance become equal to the resistance?

[robrenz]

alm, thanks for the homework assignment.  Figuring out those answers finally turned the light bulb on. Using 1pf on the series resistors only, the -60db 1stage drops to -30 at 200MHz The 3 20db's drop to -30 at 2.5GHz.  1pF is 247.5 ohms at 643.050MHz.  1pF is 24999.975 ohms at 6.336MHz.  So the 247.5 ohm when paralleled with the 1pF at 600kHz only drops to 247.269 ohm (only a .093% drop) while the 24999.975 ohm drops to 22846.724 (a 8.613% drop)