Testing mic preamp frequency response

[EmbraceRandom]

I've got a small broadcast mixer from the late 70s. Its mic preamps sound good but I suspect some of the channels have weak LF (possibly intermittently) so I'd like to test their frequency response.

I figure I could do the following:

• Set up a sine sweep on a channel in my DAW and route it to a separate channel with Voxengo SPAN inserted

• Also route the sine sweep to an interface output and use a TRS>XLR cable, Radial JDI Duplex and an XLR to patch the (mic level) signal to the mic preamp

• Route the mic preamp's direct output to a line level interface input, onto a DAW channel, and route that to the Voxengo SPAN channel's secondary inputs

• Compare the direct and signal-via-preamp frequency responses in SPAN

• Test multiple times at various mic pre input gain settings

Is there anything wrong with that process?

Thanks

[Wonks]

Sounds fine. Just ensure the number of samples per display update in Span is set as big as you can, to ensure you don’t get aliasing from higher frequencies boosting the low frequency display, or do an off-line frequency analysis of a recording of the sweep.

[EmbraceRandom]

Thanks. How would aliasing occur if its a clean sine sweep that would only go up to Nyquist anyway?
Any harmonics above Nyquist added by the preamps would be filtered out by the ADC, would they not?

[Wonks]

Think about how many samples it takes to represent a full sine wave at 20Hz at say 44.1kHz sampling.

20Hz means each cycle is 50ms long or 44100/20 samples = 2205 samples. If you are displaying after 1024 samples, then the software isn’t getting the full picture of those low frequencies so higher frequencies are also included in the low frequency amplitude values.

You are also often getting smoothing of the display curve (depends on settings and the real-time frequency display software used), so the inaccurate values of even lower frequencies can pull the slightly higher values up as well.

It’s why in your typical DAW real-time EQ frequency displays, there’s a load of low frequencies being shown that simply aren’t there, especially if you’ve stuck a 96dB/octave high pass filter in line first and you don’t see those low frequencies disappear.

A long sample buffer for the display does really blunt the screen refresh rate, so they normally forgo low frequency accuracy for showing the change in the higher frequencies very quickly.

But if you are interested in low frequency performance, you need to set the sample buffer size high or you won’t get the real values displayed.

A frequency bar graph can be used for comparison. Whilst it won’t be as accurate a display (as each bar is typically covering 1/3 octave) it is good for showing the disparity between bass levels on standard real-time displays.

[Hugh Robjohns]

As wonks says, you need to be careful if using any FFT type display as the low end data can easily become very misleading.

If you're just try to confirm incorrect LF performance, you could do it more reliably with static tones or a low frequency sweep from a frequency generator or test tone file* and a simple level meter.

For static test tones, I'd use 1kHz as the reference and adjust output/input levels to get a reading of, say, -10dBFS. Then change the test frequency to 500Hz, 250Hz, 100Hz, 80Hz, 60Hz, 40Hz and 20Hz.

*Even easier, use track 25 from the SOS Test Files download (https://www.soundonsound.com/techniques ... test-files) which is a sweep from 20Hz to 200Hz at a constant level.

Ideally, during the frequency changes or sweep, the level meter shouldn't change by more than, say, 0.5dB. If it does, there may be a high-pass filter engaged in the channel, or the AC-coupling electrolytic capacitors between circuit stages maybe drying out and need replacement.

[cashhewn]

As wonks says, you need to be careful if using any FFT type display as the low end data can easily become very misleading.

Is this always the case with FFT type EQ plugins? Are there any SoS articles on this to help understand how better to accurately interpret low end using these kinds of plugins?

[EmbraceRandom]

Thanks Wonks and Hugh, that's amazingly insightful. Really appreciate it.

I'll run the tests when I'm back in the studio next week and come back if I have any more questions

[James Perrett]

• Also route the sine sweep to an interface output and use a TRS>XLR cable, Radial JDI Duplex and an XLR to patch the (mic level) signal to the mic preamp

Is there anything wrong with that process?

I wouldn't want to put a transformer in the signal path if I was trying to do low frequency measurements. It would be much better to use a simple resistive attenuator - or even no attenuator at all and just keep careful tabs on the signal level.

[Hugh Robjohns]

^This is a very good point. The transformer in the reamper could cause measurement problems.

It wouldn't be a problem if you're just comparing channels, but could upset any absolute measurements.

Instead, it would be better to turn the mic preamp gain to minimum (phantom OFF) and run the test tones at -40 or -50dBFS with the interface balanced output connected straight to the mixer mic input. That wouldn't be great for noise or THD testing, but it won't affect frequency response measurements

[EmbraceRandom]

Great to know, thank you both. I was planning on testing it that way too, just to see if impedance happened to be a factor (perhaps there's no way it would be?) and generally testing for scenarios where I run line level sources (e.g. a mix) through the mic preamps.
I'll just make that my main method for testing now.

[Hugh Robjohns]

I was planning on testing it that way too, just to see if impedance happened to be a factor (perhaps there's no way it would be?)

It won't be...

Almost all current analogue audio gear employs a voltage-matched interface paradigm, and that requires a low source output impedance feeding a substantially higher destination input impedance (typically 10x).

This ensures the transfer of the maximum possible signal voltage from the source to the destination, which is what we want. Or, put another way, it avoids losing signal voltage as it passes through the interface.

M ost microphones have an output impedance of 150-200 Ohms, and mic preamps typically present an input impedance of between 1500 and 5000 ohms — in other words, at least 10x higher than the output impedance.

Line level equipment output impedance is often very similar to a microphone at around 75-150 Ohms, but may be as high as 1000 Ohms in some cases. Line input impedance is usually between 10,000 and 50,000 Ohms — so substantially higher than most mic preamps.

Obviously, this is a lot more than 10x higher (in most cases), and that's intentional — it's to allow one output to feed multiple inputs simultaneously (eg. mixing desk out feeding three tape recorders in parallel) with a negligible loss of signal level. (This arrangement is called bridging, or a bridging input).

Nevertheless, a modern line output with a source impedance of around 100 Ohms will still be perfectly happy feeding a mic preamp input of more than 1500 Ohms as it is still over 10x higher.

So, as said previously, for the majority of analogue audio connections, impedance is only very rarely an issue.

Exceptions include electric guitars and passive loudspeakers, amongst others....

[Wonks]

As wonks says, you need to be careful if using any FFT type display as the low end data can easily become very misleading.

Is this always the case with FFT type EQ plugins? Are there any SoS articles on this to help understand how better to accurately interpret low end using these kinds of plugins?

It’s normally the case that they favour higher frequency update times than LF accuracy. If you want to test an EQ’s frequency display, then a pink noise source then a high pass filter with say a 48dB/octave (or steeper) slope and then your EQ plug-in’s display should give you an indication of how low you can trust the EQ’s frequency display by adjusting the LF cutoff frequency.

Start quite high, say 1kHz, with the high pass filter, and then lower the frequency. You’ll discover a point where the filter starts having very little effect on the frequency display.

It’s normally quite low, 50Hz or so, before the frequency plot becomes inaccurate. But that is the reason why you seem to have a lot of low frequency sound in your recordings when looking at EQ displays despite your ears telling you otherwise.

Most DAWs allow you to select a block of recorded sound and display a frequency analysis of it. This should be accurate. It’s just the real time displays that can be inaccurate at low frequencies. EQ frequency display parameters are generally fixed, whilst frequency scopes like Span normally have adjustable parameters. Which is why it’s important to use and trust your ears whilst dealing with EQ displays. If you are looking at low frequencies, then having Span (or other frequency scope-style plugins - I generally use the Melda scope) after the EQ, set to a large sample size so you can view the LF content, is very useful.

Alternatively use a graphic bar graph frequency display, where the sample period is optimised for each frequency band.

[cashhewn]

Many thanks for this thorough information Wonks!

[EmbraceRandom]

Thank you both again! Very grateful for you both.