Measuring and attenuating bridged/BTL speaker output?

[abza]

A quick sanity check is needed, if you'll be so kind

I've been using a Focusrite Clarett 2Pre to make measurements of single ended amplifiers - so far so good. I have a 300W, 8ohm dummy resistor and a simple, single-ended attenuator box to ensure the voltage at the interface inputs isn't too high.

However I've got a bridged amplifier I'd like to test, and just want to double-check my methodology. In addition, I'll need a new attenuator design.


My idea is to wire across the resistor and feed the output into one of the interface's balanced/differential inputs - either the microphone XLR input or the balanced TRS line input. I'll use the signal positive and signal negative, and leave the ground floating.


In cases where the output voltage would be too high, I plan on having a simple H-bridge attenuator, dropping the speaker output by 10dB before it reaches the interface.

Here's the master plan. Any problems/issues? Any "best practices" when it comes to measuring bridged amps in sound cards?

 

[Monte McGuire]

The circuit above looks good, and after grinding the numbers with a 300W RMS amplifier output, the attenuator produces a peak signal level of around +18dBV at the interface input, which is probably optimal for most line level inputs.

Most power amps and sound cards are not going to have all that low of distortion, so the noise produced by the 2k2Ω resistor won't be a limiting factor. However, if you test really clean amplifiers at lower signal levels, the distortion floor may be limited by the noise from the 2k2Ω resistor. This all depends on a lot of other factors, so start with the circuit above and see if you need to change the H pad to be able to push the noise floor down to accurately measure the harmonics.

To do that, you could scale down the resistor values by a factor of 4 to get 6dB less noise. But, when you do that, the power dissipated in each resistor goes up, so a simpler approach might be to use a parallel of 4 or 5 of each of the 10kΩ and 2k2Ω resistors to build that pad - the noise will drop and the power rating will also go up while the pad attenuation remains the same. You'll probably buy 10 of each resistor value anyway, so this might be a good way to use those extra resistors you bought that would sit around unused otherwise.

Also make sure to use quality metal film resistors for the 10kΩ and 2k2Ω resistors. With large amplifiers, you'll dissipate around 0.1W into those parts, so it's helpful to use slightly larger wattage resistors, like a 1/4W or 1/2W metal film part. Don't use thick film parts, or wirewounds, since wirewounds can be really dirty if you're unlucky. In the US, RN60 size or RN65 would be ideal.

 

[abza]

@Monte McGuire - wow, thank you so much for the clear and detailed reply.

Fortunately I do have access to Vishay/Dale CMF55 series resistors (metal film, 0.1%, 0.5W) which I'm sure will work like a charm.


I must admit I wasn't aware that larger value resistors would introduce noise - I know carbon and thick-film types are noisier than metal film, but hadn't really considered the actual resistance as providing noise to the signal. Now that I think about it, it makes sense! Is there a sort of "rule of thumb" when it comes to predicting the level of noise that would be introduced, or just a matter of experimentation/"the lower the better"?

I'll have to breadboard it and test before committing, but really like your solution of simply stacking them in parallel to decrease their combined value while driving up the power handling capability.

 

[phofman]

IMO power amp speaker output will very likely be more noisy than resistor noise.

I built a small balanced adapter with ranges matched for balanced inputs of ESI Juli@ Balanced Amp Outputs Measurement Adapter . The attenuator (2W resistors) in shielded enclosure does not generate any extra noise measurable by that soundcard (mid-fi by today standards).

 

[abza]

@phofman - ah nice, thanks for the link! Elegant handling of SE/Balanced toggle in your schematic. And that’s some pretty serious 3D printing work Appreciated!

 

[phofman]

Thanks The 3D printer has become an indispensable friend of mine for DIY work. Some hours of making the model spare me of filing unfitting holes and parts. Real-world parts all fit just like the model in the PC, I love that

 

[Elvee]

I must admit I wasn't aware that larger value resistors would introduce noise - I know carbon and thick-film types are noisier than metal film, but hadn't really considered the actual resistance as providing noise to the signal. Now that I think about it, it makes sense! Is there a sort of "rule of thumb" when it comes to predicting the level of noise that would be introduced, or just a matter of experimentation/"the lower the better"?

There is more than a rule of thumb, there are formula's:
Johnson–Nyquist noise - Wikipedia
But in your case, carb-comp, thick-film, etc. would make no difference: their pure thermal noise is the same as for a plain metal resistor.
The difference is the excess noise, when they are biased.
That said, I do not recommend low grade resistors for your application since they also have other imperfections like a significant voltage coefficient.

 

[Mark Tillotson]

I must admit I wasn't aware that larger value resistors would introduce noise - I know carbon and thick-film types are noisier than metal film, but hadn't really considered the actual resistance as providing noise to the signal. Now that I think about it, it makes sense! Is there a sort of "rule of thumb" when it comes to predicting the level of noise that would be introduced, or just a matter of experimentation/"the lower the better"?

In short the thermal noise is derivable from basic physics, and resistors where thermal noise is the main source of noise are as good as you can get (metal film or wirewound). The rules of thumb is √(4BRkT) for voltage noise, wikipedia gives all the details.


One interesting point is the thermal noise _power_ is the same for any resistance value, only the balance between voltage noise and current noise is affected by the value.


When measuring noise we are usually interested only in the voltage, but in circuit design performance is often mainly dependent on the input current noise spec of the active devices, current noise is often forgotten about alas.

 

[Mark Tillotson]

It might be worth considering a capacitive divider too - yes there can be phase shifts, and you have to avoid placing too much capacitance load on the amp, but the heat dissipation and noise issues can both be addressed.


The resisitive divider can be made very low noise if its incorporated into the load, by splitting the load resistance into two equal high power resistors and interposing a low value resistor of moderate power rating for the tap points.