Resource for ADC quality? Recording interfaces

[617]

We have a great variety of DACs tested here, but it may be in the interest of a lot of people to know about the performance of recording interfaces. Specifically, microphone preamps. Is this an easy thing to test? In the same way that our hi-res music formats are limited by our DACs we use for conversion, our recordings are limited by ADCs, and microphone self noise as well.

If there is any interest I am sure we can put together a collection of mic preamps for test if there is community interest. I know this would be of considerable interest to journalists, music producers, and other people using inexpensive mic preamps for field recording.

What would such tests look like?

 

[scott wurcer]

A decent survey here, https://www.avisoft.com/recordertests.htm . I'm not sure if Rob Danielson's (professor of film recording) pages survived his retirement. He had a more extensive survey of microphones.

If you look on the Linear Audio site you can find some recordings that SIY and I made in Austin. They were DIY mics (details in one of the articles) and the Fostex FR2-LE field recorder.

 

[617]

A decent survey here, https://www.avisoft.com/recordertests.htm . I'm not sure if Rob Danielson's (professor of film recording) pages survived his retirement. He had a more extensive survey of microphones.

If you look on the Linear Audio site you can find some recordings that SIY and I made in Austin. They were DIY mics (details in one of the articles) and the Fostex FR2-LE field recorder.

That's an excellent resource! These EIN figures are closer than I would have thought, ranging from -130 to -100 or so. Looks like the noise level of a recording system is as much dependent on the microphone's self noise. Dynamic mics are really good for field recording but getting enough gain out of one of these recorders for a low output omni dynamic is not trivial.

Would it be reasonable to assume that my ~-126dbu EIN tascam unit can capture a good 20-22 bits dynamic range with the right mic?

 

[scott wurcer]

Would it be reasonable to assume that my ~-126dbu EIN tascam unit can capture a good 20-22 bits dynamic range with the right mic?

I have two recent Tascam field recorders, if you are talking A weighted noise floor probably but I don't know how you would measure THD via a mic and an acoustic source. Other points, the controls are for "tiny hands", and the monitor outputs are not via spectacular DAC's.

You can get a Rode NT01 for <$250 and you will have a hard time finding any situation where the venue has less noise than the mic. I have some field recordings from 3mi offshore in Casco Bay that I made at an empty golf course at 4AM. Unfortunately the putt-putt of the lobster boats carries like crazy over open water.

 

[617]

I have two recent Tascam field recorders, if you are talking A weighted noise floor probably but I don't know how you would measure THD via a mic and an acoustic source. Other points, the controls are for "tiny hands", and the monitor outputs are not via spectacular DAC's.

You can get a Rode NT01 for <$250 and you will have a hard time finding any situation where the venue has less noise than the mic. I have some field recordings from 3mi offshore in Casco Bay that I made at an empty golf course at 4AM. Unfortunately the putt-putt of the lobster boats carries like crazy over open water.

I'd love to hear those. My concern is field recording interviews. You need a mic that sounds natural with good rejection of background stuff. Dynamic omnis sound good but need lots of gain, lavs can work but handling noise is difficult.

 

[Blumlein 88]

That's an excellent resource! These EIN figures are closer than I would have thought, ranging from -130 to -100 or so. Looks like the noise level of a recording system is as much dependent on the microphone's self noise. Dynamic mics are really good for field recording but getting enough gain out of one of these recorders for a low output omni dynamic is not trivial.

Would it be reasonable to assume that my ~-126dbu EIN tascam unit can capture a good 20-22 bits dynamic range with the right mic?

I think maybe the EIN spec is a little different than how you are looking at it. There is a good explanation of EIN (and other specs) down this page.
https://www.rane.com/note145.html

Normally this EIN is measured with a 150 ohm source impedance. Thermal noise means the lowest noise you could have is -131 dbu. These are usually measured with 60 db of gain from the microphone pre. So a pre adding no noise of its own would apply 60 db gain, and record -71 dbu. Your Tascam would add 60 db of gain and record 66 dbu as it added 5 db of noise that wasn't in the signal. Which isn't a bad result at all.

Now this doesn't indicate your mic pre has 126 db of dynamic range. It could have only 90 db dynamic range. The gain to the noise from the 150 ohm source impedance pushes the signal above the noise floor of the circuitry of the microphone pre. Microphone pres can have SNR or dynamic range values around a 100 db or a bit more. I don't know of any with more than 120 db.

Microphone self noise would be above the noise from the source impedance. Though a few like the Rode mentioned, some Lewitts and a few others like a Shure KSM44 get noise very low nearing the equivalent to 0 db SPL. You are never going to find a real venue that isn't noisier than that. But the self noise, and source noise and venue noise all get pushed up when you use gain in the microphone pre. It is pretty hard to get a recording with a noise floor lower than - 70 dbFS. A microphone pre with a good EIN and 100 db of SNR is not a limiting factor in the recording.

 

[amirm]

Was the interest in OP in preamps or microphones? If the former, it is pretty easy to measure them.

 

[617]

Was the interest in OP in preamps or microphones? If the former, it is pretty easy to measure them.

Amir, preamps. Mics are harder to measure and their noise performance is only part of the picture so you have to make compromises.

I would bet there is considerable interest from the home studio community to see the input performance of say a focusrite audio interface versus a behringer and and an apogee or whatever. Not sure how you'd test it; you'd supply a test tone through the mic xlr connectors and play with the gain setting I suppose.

 

[Cbdb2]

First mic pres are analog, second, most decent mics show measurements, third and most important, music recording mics are not flat, usually directional, which is frequency dependent, and its there imperfections that are sought after. This is why bigger studios have dozens of different mics. And hardly any B&K ( measurement ) mics. For example, the most popular guitar amp mic is the cheap lowely Sure Sm57, vocals Neuman U87.

I worked as a dialogue/sfx editor for many years and for field recordings where you dont want to see the mic shotguns are usually used, unless you want all the other noises around you ( the ones your brain filters out untill you playback a recording). For field interviews where you can see the mic use a shotgun, lavaliere or get the guy to hold the mic.

Unless your recording in a quiet room the self noise of the mic/electronics will be buried in the ambient noise.

 

[scott wurcer]

It is pretty hard to get a recording with a noise floor lower than - 70 dbFS. A microphone pre with a good EIN and 100 db of SNR is not a limiting factor in the recording.

As I mentioned before I'm not a huge fan of the total integrated noise referenced to a full scale tone as a sole measure of DNR. It ignores specific spectral properties of the transducer/system and any impact of masking at particular frequencies. I prefer data presented as spectra normalized to a 1Hz BW, you can get to any other measure from there, just as I would prefer THD vs level (not THD + N) with noise measured separately.

 

[Tks]

As I mentioned before I'm not a huge fan of the total integrated noise referenced to a full scale tone as a sole measure of DNR. It ignores specific spectral properties of the transducer/system and any impact of masking at particular frequencies. I prefer data presented as spectra normalized to a 1Hz BW, you can get to any other measure from there, just as I would prefer THD vs level (not THD + N) with noise measured separately.

Sorry if this is a wierd question but I'm new to the science and was wondering if you could give a perhaps laymen response as to why THD vs Level instead of + N?

 

[scott wurcer]

Sorry if this is a wierd question but I'm new to the science and was wondering if you could give a perhaps laymen response as to why THD vs Level instead of + N?

At low level the integrated noise obscures the THD. In general the input signal has no effect on the noise so on the usual up sloping as level goes down plots the information is redundant/useless.

 

[Blumlein 88]

As I mentioned before I'm not a huge fan of the total integrated noise referenced to a full scale tone as a sole measure of DNR. It ignores specific spectral properties of the transducer/system and any impact of masking at particular frequencies. I prefer data presented as spectra normalized to a 1Hz BW, you can get to any other measure from there, just as I would prefer THD vs level (not THD + N) with noise measured separately.

I agree with you completely. I've complained about the same thing elsewhere on the forum. But on a forum it is easy to complicate things to the point it doesn't help communicate. On this specific point about recordings, I posted elsewhere how I've not seen recordings better than -65 dbFS noise floors, but if you look at that noise floor in terms of ERB where our hearing is most sensitive around 3-5 khz, the noise floor will be well into the -90 dbFS range. So thinking about the -65 db noise floor is rather deceptive.

 

[Tks]

At low level the integrated noise obscures the THD. In general the input signal has no effect on the noise so on the usual up sloping as level goes down plots the information is redundant/useless.

Generally speaking wouldn't that occur always? As in, all noise obscuring the THD at low level?

EDIT: idk why the post submitted, but I meant to say, wouldn't that make THD + N redundant always from the current aspect with respect to the microphone topic?

 

[Blumlein 88]

Sorry if this is a wierd question but I'm new to the science and was wondering if you could give a perhaps laymen response as to why THD vs Level instead of + N?

To expand upon Scott's answer take a look at this chart.

Notice the THD+N is higher at lower signals. What does this show us? About noise not much. Most likely the actual noise separate from the signal is pretty much the same on the left hand side as the right hand side. What does this show us about distortion? Also not much. Where the curve levels out or begins to rise then THD is the predominant signal. On the left usually the THD will be very low and the lower the level possibly the lower the percent THD, but we cannot know that for sure because noise is covering up what distortion is doing vs level. Most of that chart tells us nothing useful.

See this chart using an FFT.

Because the FFT digs deep into the noise and with regard to frequency, we see the individual harmonics separate from the noise. If we had several of these starting at max signal level and dropping each step maybe 6 db at a time, we could see what happens with distortion vs level separate from noise. Eventually even with this you may eventually see the harmonic distortion drop into the noise. The THD+N vs level gives you the idea the signal isn't as clean in distortion terms as level drops. In fact distortion likely drops with dropping level, and you'll be able to see that. Noise itself likely drops very little with level. So that THD+N curve sloping upward with lower level is deceptive in what it shows, and obscures much that is in fact possible to know about a given piece of gear. THD+N makes a simple looking graph, but it really isn't worth much.

If I've messed any of this up, Scott can point it out.

 

[Tks]

Thank you guys, this makes far more sense now

EDIT: Just one final question, wouldn't THD+N at least determine where both are at the very least are in terms of their worst amount of presence?

 

[Blumlein 88]

Thank you guys, this makes far more sense now

EDIT: Just one final question, wouldn't THD+N at least determine where both are at the very least are in terms of their worst amount of presence?

Could you re-phrase the question. I'm not clear on what the question is exactly.

 

[scott wurcer]

To expand upon Scott's answer take a look at this chart.
If I've messed any of this up, Scott can point it out.

Fine, certain class A/B amps can have rising THD at low levels a test for this would be nice. A more difficult thing to test for is thermal de-biasing, for instance, a quiet passage immediately after a loud passage.

 

[LTig]

[..]
See this chart using an FFT.
View attachment 35860

Because the FFT digs deep into the noise and with regard to frequency, we see the individual harmonics separate from the noise. If we had several of these starting at max signal level and dropping each step maybe 6 db at a time, we could see what happens with distortion vs level separate from noise. Eventually even with this you may eventually see the harmonic distortion drop into the noise.

Yes, but you can increase the number of averages and/or the FFT size to lower the noise even more and then the harmonics will stand out again. It just takes much more time (doubling for each 3dB noise reduction).

The THD+N vs level gives you the idea the signal isn't as clean in distortion terms as level drops. In fact distortion likely drops with dropping level, and you'll be able to see that. Noise itself likely drops very little with level. So that THD+N curve sloping upward with lower level is deceptive in what it shows, and obscures much that is in fact possible to know about a given piece of gear. THD+N makes a simple looking graph, but it really isn't worth much.

If I've messed any of this up, Scott can point it out.

Sound's all correct for me.

 

[Tks]

Could you re-phrase the question. I'm not clear on what the question is exactly.

The THD + N measurement itself, does it at least provide a worst case scenario when both noise and distortion are measured together. So not one or the other can be higher on their own when compared against the THD + N measurement itself?