switched attenuator for measurement

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
5,951
Im having loads of fun with my budget Feeltech Fy6600 and Fy6800 Arb gens .
It occured to me that when you dial in a signal say in the milli volts region to test audio the units bit rate/dynamic range is severly limited
Like has been suggested in other posts, running your digital generator at full level ,then using an attenuator on the audio output is a good way to maximise signal quality . I remember Ian.T.B made a special attenuator for audio testing especially for this purpose but cant find the thread .

The sig gens are the usual 50 ohms output so I guess its a good plan to keep the attenuators to be able to opperate properly into that impedence . Something like five steps of ten db coarse adjusment and fine control in one db steps would be nice .

Of course Ive attended to other areas in order to allow sustained opperation of the unit at full power with more headroom ,new high power output op amps (THS3095 ,THS3491),powersupply upgraded ,+/- rail voltage raised etc.

I also note people recomending an output transformer in relation to testset outputs to prevent any chance of phantom power damaging the dc coupled out of the generator .  What would be a good choice of transformer in this application ? I'd imagine mic input would be the lowest impedence  you'd be expecting to drive with it  , ballanced line level inputs at 20 k also . Could the attenuator be placed between the generator and the transformer ,say a smaller mu metal 1:1 10kohm to avoid core saturation due to the generators  high output level? 

Another suggestion someone made to improve the signal quality of a  generator for single frequency distortion measurement  was to include a low pass in the output in order to attenuate 2nd and 3rd harmonics etc .So, by scrubbing off a predefined number of db of output in a low pass filter can we improve on the theoretical  limits set by the 14 bit dacs in the unit ?
Maybe a simple switchable  low pass at a handfull of common test frequencies might work well here?

I have a couple of small die cast boxes , BNC connectors, some nice mini rotary switches 2x6ways etc ,I can patch in any one of a number of transformers as the need arises . 20V peak is quoted as maximum output from the generator ,into 50 ohms 10 volts is maximum . If anyone knows software that would allow me to calculate  the resistance values ,impedence,number of steps and db loss that would be great ,Im aware there are also several different types of attenuation network which kind makes most sense for my application ? its a simple unbalanced signal ,but ladder ,pi etc , I'm aware of the different impedence charachteristics of the various types of attenuator ,but I cant clearly visualise which is best for the job

Thanks again,

 
Every DUT is different so you will invariably end up with a different rig for each.

For example, for a mic input test, the ideal rig would be a balanced attenuator that attenuates by slightly less than whatever it takes for the output to be near peak dynamic range for whatever the mic gain setting is. You can always fine tune the level with the output adjust of the FN gen so that the output of the mic pre is just below the clipping point. And you probably want the output impedance of the attenuator to match a typical microphone but the most common used (and perhaps slightly low) impedance would be probably 150 ohms (not 50 because it would yield unrealistically good results).

But even I don't get into it that deep. I just use a cable like this:

attn40mic.png


Of course if the FN generator output is balanced (mine is not), then you can do a basic 3 resistor balanced attenuator.

As for phantom, I wouldn't worry about it. Phantom has 6k8 in series with each leg so your FN gen will slurp that down like a frosty shake. The only concern would be if the FN generator was powered off. In that case, I'm not even sure if protection diodes would help (maybe low capacitance TVS diodes to ground but that might affect distortion measurements). But my guess would be that even an inexpensive unit like that can handle a measly 48V.
 
Here is the schematic of my home made test interface with -40dB and -60dB source pads, 600 ohm and 20dB pad 600ohm loads.

Cheers

Ian
 

Attachments

  • CTC-Audio-Test-Interface-MK2.png
    CTC-Audio-Test-Interface-MK2.png
    35.9 KB · Views: 77
So I put together my attenuator box ,as of yet no resistors are wired in ,

Im a bit uncertain about charachteristic impedence i need ,sig gen output is 50 ohms ,but I probably dont need to load down the output to that degree which will only cause more thd .

I have a dual pole 12 way switch on the 1db range ,would a H network be usefull here ?

I found a few good attenuator network calculators ,but they only work for single values , theres also several different variations on each kind of network  and Im a little unsure how to best to implement it , three equal networks of 10db two switched by toggle switch and the third switchable in 1db steps ,with  off and 0db to -10db positions

This guy has a nice set of attenuator calculators ,
http://www.chemandy.com/calculators/t-attenuator-calculator.htm

Advice on best network types to use and impedence would be greatly appreciated .
 

Attachments

  • Test set.JPG
    Test set.JPG
    742.2 KB · Views: 24
Thats some 1:1 10k:10k transformers I have in the first photo  , 600ohm impedence attenuator output should adequately drive 10k transformer with 20k termination ?, I have some other 1:4 input transformers ,could one of these be used in reverse on the attenuator output to supply lower voltages at lower  impedences?

I was thinking of using the off position on the fine range switch to apply 150 ohms to the input of the DUT for noise testing purposes ,inspired by Ians design ,Ive read through some of the other posts on attenuators ,but each case has a unique set of requirements so its only of limited help ,I did see PRR suggested 'bridged T' when a 0db position is required, as I have the dual pole 12 way on the fine range ,how is this best implemented ?  I can see an unbalanced  series attenuator  like half of Ians 40/60db switch but -10 and -20db is doable on a  3way (on-on -on) toggle switch.  Im unsure how to scale the impedence of the bridged T properly , in relation to the series attenuator ,  and how impedence relates to bandwidth.Preferably I want to be able to drive signals well into the 100's of khz  through the network.

The layout is nice and simple the way it is but I could put in an extra toggle very easliy , Im aware that trying to cover all eventualities  in terms of an attenuation box is impossible ,  simple logical user controls in an all passive box are a pleasure to use .

Would it be possible , keeping things all passive , to add a sharp LC filters and make them switchable into my attenuation network in such a way that when testing single frequencies for distortion ,say 1khz, the 2nd harmonic ie 2khz and above was sharply filtered from the output?

The FY6800 dacs are 14bits ,by attenuation and filtering external to the generator can I reduce  the the THD content of single sine waves effectively by this method specifically for the purpose of distortion measurement ?.
At -20db from full output digital  ,thd rises to around 0.25% in the generator ,were at full output 0.03% or less was measured .

I'll do out a 'back of a fag packet' sketch of my idea so far and post it up in a while ,any help or ideas graciously accepted .
 
Tubetec said:
So I put together my attenuator box ,as of yet no resistors are wired in ,

Im a bit uncertain about characteristic impedance i need ,sig gen output is 50 ohms ,but I probably dont need to load down the output to that degree which will only cause more thd .
There are two important things. The source impedance of the generator (50 ohms in your case) and the source impedance you want to present to the unit under test (UUT). You are unlikely to want to present the UUT with a source impedance less than 50 ohms.

You are unlikely to want steps finer than 20dB because in between even a digital output resolution will be sufficient not to introduce any measurable distortion.

So for 0dBu out use the generator as is. If you want its source impedance to be greater than 50 ohms, stick a resistor in series with it.

For -20dBu output use a 450 ohm and 50 ohm attenuator. The gen is loaded with 500 ohms (no problem for it) and the -20dBu source looks like 50 ohms. If you want the -20dBu output to be greater than 50 ohms stick a resistor in series with it.

Repeat for other values of attenuation.

If the generator output is balanced and you want to maintain CMRR then split the 450 ohms into 2 x 225 ohms and make it into a U attenuator.

Cheers

Ian
 
Thanks for your help Ian,
I can see how holding the output impedence low is a good move for all kinds of reasons when your trying to deliver a test signal .

As far as adjusting the sig gen for db range its do-able but means you need a table to convert vp-p on the screen to dbu, its cumbersome at best .

There is some extra complication having 1db steps on the attenuator, but  for ease of use, counting clicks of a switch  beats  having to look in a screen, menu and opperate  buttons and encoders any day.
 
Tubetec said:
As far as adjusting the sig gen for db range its do-able but means you need a table to convert vp-p on the screen to dbu, its cumbersome at best .

I thought you said the sign gen was digital?? So some software somewhere sets the level? With the right program it is easy to change level in dB or have I missed something?

Cheers

Ian
 
Hi Ian,
I found a combination of 10/20 db attenuator ,at the -20db position it output impedence is 50 ohms ,at -10 its 100 ohms , and if I use the one db switch is higher output impedence ,but thats ok , its still handy for a quick check

Theres one or two small calibration issues with the Feeltech FY6600/6800 , mainly to do with the tollerance of the components in the output stage . The display is in volts p-p and in order to adjust levels in db, a db to voltage chart would need to be reffered too each time you wanted to change level .

The unit does have its own software which allows computer control of all the parameters on the front panel and allows creation of arb waveforms and upload to the machine via usb ,it was definately well worth the 60 euros or so it cost for the basic 15mhz model ,and Feeltech sent me out the new 6800 60mhz version FOC for my contributions to the EEV blog on their product.

 
Tubetec said:
Theres one or two small calibration issues with the Feeltech FY6600/6800 , mainly to do with the tollerance of the components in the output stage . The display is in volts p-p and in order to adjust levels in db, a db to voltage chart would need to be reffered too each time you wanted to change level .
It's not crystal clear to me why you need to know dB at all. The dB scale is relative. So even if you know what Vpp corresponds to 0 dB-whatever, it's actually not that useful to come up with one dB value by itself. What is helpful is knowing the difference between two levels like the maximum level a device can handle vs the noise floor. For that, I would want the sig gen to be ~-10dB from it's maxium output and have the DUT be near it's peak level. For most devices this means you need an attenuator. Then you adjust the sig gen output +-10dB. Then the attenuator doesn't need that many steps. For a mic pre you only need 2-3 steps maybe. My example cable uses -40dB but it would be nice if I had a cable that could switch between -40dB, -50dB and -60dB for measuring mic pres. But then I need a switch with a metal box and more spots where cable is unshielded and all of that comes at a price.

Your ability to measure performance of a device is going to depend on the performance of your test apparatus. So if you start adding transformers and op amps to your test rig, you have to ask yourself of those components are really exceeding the performance of whatever it is that you're measuring.

So again, I like to just evaluate each DUT and devise a rig specifically for that device or class of device. In practice, I use 2-3 cables with different ends and levels of attenuation. With RCA cables and some adatpers (see MPJA.com), I can get to BNC, XLR and 1/4" phono. I could make everything balanced too but unless you're measuring the performance of a balanced in / out, unbalanced is fine and in theory should perform better actually (circuits are almost always unbalanced internally).
 
dB just suits me ,I'm used to dealing with it ,and its a lot easier than having to refer to a big string of numbers .

Its the output op amps I have upgraded in one of my sig gens ,THS 3002 was the original ,Ive tried THS 3095 ,and will try the new 3491 op amps soon .

The transformers Id only use if balanced working is nessesary , and of course Im aware they impose there own limitations on any test signal .

Thanks again for taking the time to comment Squarewave.

 
Tubetec said:
Hi Ian,
I found a combination of 10/20 db attenuator ,at the -20db position it output impedence is 50 ohms ,at -10 its 100 ohms , and if I use the one db switch is higher output impedence ,but thats ok , its still handy for a quick check

For most tests as long as the source impedance is 100 ohms or less the exact value does not matter.

I just looked at the Feeltech user manual so I now understand what you mean about the display. ans setting the output voltage always being  peak to peak. If you set the peak to peak sine output to 2.19203 then the output will be 0dBu. The use your dB switches to set any required level.

Cheers

ian
 
ruffrecords said:
Here is the schematic of my home made test interface with -40dB and -60dB source pads, 600 ohm and 20dB pad 600ohm loads.

Cheers

Ian

I am interested in a attenuator for this purpose as well. I just would like to ask about the schematic. Are you using BNC connectors or XLR? And how is it all hooked up with tone generator and Oscilloscope? And I do not understand how the switch (s1A) works? It seems like the signal is cut when switching? Or am I missing something obvious here?

Very interesting though.

best regards

/John
 
johnheath said:
And I do not understand how the switch (s1A) works? It seems like the signal is cut when switching? Or am I missing something obvious here?
That's for measuring noise. You measure with signal to figure out what the peak level is and then remove the signal without touching any of the DUT controls to see where the noise floor is. Terminating with 0 tells you the absolute lowest noise floor of the circuit. Terminating with 150 ohms is much more realistic.
 
squarewave said:
That's for measuring noise. You measure with signal to figure out what the peak level is and then remove the signal without touching any of the DUT controls to see where the noise floor is. Terminating with 0 tells you the absolute lowest noise floor of the circuit. Terminating with 150 ohms is much more realistic.

Thank you sir

Then it makes sense to me. Thanks.

An could you tell me about the connectors?

Best regards

/John
 
Ians attenuator is set up for balanced working ,so xlr's or three pole jacks there.
I need some good literature on building attenuators ,theres just so many different schemes its hard to figure out the best for a particular purpose ,each kind has its compromises and plus points too.
Seems small 1db steps on a string attenuator force upping the output impedence ,which wont improve results much .
 
Tubetec said:
Seems small 1db steps on a string attenuator force upping the output impedence ,which wont improve results much .
It doesn't have to. You could switch the outer legs of the attenuator and keep the center resistor constant.
 
I just discovered the HP 350 series audio attenuator , 600ohms, quite a brick of a thing ,dual switched bridged T sections with two switched elements ,I really didnt imagine when I put my simple switch box together what was involved in a proper matched attenuator, ;D   

Model 353 is even nicer incorporating ,transformer ins and outs ,

I see them for 40 to 50 bucks on ebay ,by the time shipping and customs is added, 100 if your lucky to Ireland, Customs have the neck to calculate your charges not on the value of the item but shipping too .It kills the good in it completely.
Regular US airmail is saucy enough for any kind of weight , although stuff can get through without customs noseing in, heavy items tend to attract their attention ,Still if its marked of a nominal value under 22 euros,I think  its not applicapble for import duty .

The 200 series tube oscillator is another thing Id have on my bench in a second only for afore mentioned charges .I love that industrial feel and simplicity .big analogue scale and no blinking lights or other distractions to the eye.


 

Latest posts

Back
Top