Variable Attenuator for Audio Measurements

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
3,266
Location
New Jersey, USA
It would be very helpful if I had a precision attenuator for measuring gain, looking at frequency response, etc.

I'm thinking of something like 10K input impedance, 150 ohms output impedance and selectable attenuation from 0 to -80dB with 1dB resolution.

What is the standard circuit for this?

A sequence of "T" attenuators switched using toggles? Like 40+20+10+5+3+2+1 using 7 toggles?
 
> 10K input impedance, 150 ohms output impedance and selectable attenuation from 0

Can't work as specified.

At "0" the output connects to the input. How can it be "10K in" with a 150 load?

It's not 1935 any more. The modern way uses an amplifier. You can be 1Meg in, "zero" out (capable of 300r load).
 
I'd use an attenuator, followed by  a voltage follower opamp 5534 etc. Simple front end attached, from my ancient Lindos LA1, (maybe with the TL072 replaced, ). Probably many more elegant solutions, but for general work I'd use this sort of  cheapo circuit
 

Attachments

  • Lindos_IP.jpg
    Lindos_IP.jpg
    146.3 KB · Views: 42
Ok. So clearly I need to relax my requirements a little.

But I don't want to use anything active because I want to use it to, for example, measure noise a microphone preamp so it cannot add noise itself and that is difficult and probably not possible with an active circuit.

So if I relax requirements in the following ways:

1) Input impedance is greater than 1K.
2) Replace "0" with bypass.
3) Attenuation starts at -20dB with input Z 1K and higher at higher attenuation steps.

So there could be a "course" rotary 4 pos switch that is just a rheostat on top of a 150R to make steps:

  bypass
  -20dB
  -40dB
  -60dB

Now the output of that stage is 150 so then I feed that into a standard 150R pi attenuator sequence of 5 separate toggles to add any combination of attenuation:

  -10dB
  -5dB
  -3dB
  -2dB
  -1dB

for a total range of -20dB to -81dB attenuation.

The most important things are 150R source so that it can be used to make comparative noise measurements and precision so that I can use it to measure gain.

So is this a plan or is there a better way?
 
I want to use it to, for example, measure noise a microphone preamp so it cannot add noise itself and that is difficult and probably not possible with an active circuit.

I have the feeling you are thinking wrong.

1  feed mic amp (under test)with say -40dB, trim mic amp gain =0 dB on millivoltmeter
2  disconnect oscillator,short mic amp input with 150R
3  measure noise of mic amp, this will require gain on the measuring meter, not attenuation, to get the meter to read 0 dB
 
There are lots of ways to measure mic pre noise / performance. But the way I usually do it is some permutation of using an audio interface and computer like:

1. Adjust mic pre gain setting to be measured like 60dB for example.
2. Generate a tone of a known level (preferrably an absolute scale like dBu but it can just be some reading on your audio interface software) like 0dBu for example.
3. Inject said tone through the attenuator and into the mic pre. Adjust the attenuator so that the output of the mic pre matches the level of the original tone. Now you can read the gain off of the attenuator. In this case it should be 60dB. You need to know the gain if you want to compute things like EIN. If the attenuator is precise, it is a more reliable way to measure actual gain.
4. Capture 10 seconds of said tone and plot the spectrum using software. Use the peak of the tone as "0dB" on your plot. Depending on said "known level", this could be 0dBu or the clipping point of the mic pre.
5. The previous spectrum will give you a good idea of noise by itself. But you can also simply disconnect the source to the attenuator and capture again without the tone to get just the noise.
 
IMO you should disconnect the gain and noise measurements. The best way for measuring noise itself is presenting the input with a floating dummy load. For gain measurement you must make sure the attenuation is exactly as dialled, that's why I would rely on a low-Z driver; that would make the attenuator independant of the  generator's impedance.
You need a switch that toggles between the  attenuator and the dummy load. This switch can also have a third position for CMRR evaluation.
 
abbey road d enfer said:
You need a switch that toggles between the  attenuator and the dummy load.
Ooh yeah, you're so right. I don't want the USB ground connected to the DUT when testing noise! I've just been physically unplugging it but I'm doing this to make things cleaner.

So one pole to get the attn. input from either the input connector signal pin or ground and a second pole to connect the input connector ground pin to either local ground or nothing so that the generator ground is not connected to the DUT. Fortunately the RCA input connector I'm going to use has isolation washers. The output connector will be connected to the chassis ground and thus to the DUT.

Here's my current schematic:

AttenVariableMeasure.png


I partly did it this way because I happened to have little PCB mount DP4T slide switches.

I haven't agonized over the resistor values but I think it should get me within 2dB of any value between -20dB and -78dB.
 
squarewave said:
I haven't agonized over the resistor values but I think it should get me within 2dB of any value between -20dB and -78dB.
Then you should ask yourself why you want to have a 1dB resolution if the accuracy is only +/-2dB. I'd rather have less-than-1dB accuracy and 5dB resolution. Whatever instrument you use for level measurement should have less than 1dB error over a 10dB range. A simple VU-meter is better than that.
Please don't take it bad, but I think you are in need of a refreshment about metrology, particularly the distinction between accuracy, precision and resolution.
 
Well I don't like being accused of not understanding "metrology" (study of subway systems?) so I simulated this to get the right attenuation / output impedance and now each step is well within ~+-0.5 dB. I had to add an extra switch for 1dB.

-20dB to -80dB
1 dB steps
Output impedance 150 ohms

Here's the schematic and a pic of my implementation:

Atten0.png
 
squarewave said:
Well I don't like being accused of not understanding "metrology" (study of subway systems?) so I simulated this to get the right attenuation / output impedance and now each step is well within ~+-0.5 dB. I had to add an extra switch for 1dB.
You have three attenuators in series. Assuming you use 1% resistors, that results in +/- 6% precision. That's another 0.5dB you have to add to the result. So, unless you individually trim each resistor, the 1dB resolution is some kind of wishful thinking.
I have a gen with a precision attenuator, it has two switches, one from 0 to -90 followed by one from +5 to -6. I never trust the last digit, when I need to dial an exact value I always control it with the meter.
 
abbey road d enfer said:
So, unless you individually trim each resistor, the 1dB resolution is some kind of wishful thinking.
I trimmed them. I just used two resistors in most places. You can see the numbers I got from my QA400 on the piece of paper in the pic. So precision is actually closer to +-0.1dB over 90% of the range (assuming perfect precision of the QA400). Although I designed the -20dB setting (which only uses a 15K upper resistor) for the 68 ohms output impedance of my QA400. So with another device it may not be as accurate. I'm also not sure about the initial attenuation however. Meaning attenuation between steps is accurate but the entire attenuator might be shifted down / up by a few tenths. Also, I have not measured the accuracy of the output impedance. But these are pedantic details at this point. I think this is going to be a very useful device.
 

Latest posts

Back
Top