advertisement


measurement microphone calibration

gustav_errata

pfm Member
File under: "you get what you pay for"

TL;DR: Is there a DIY/hobbyist-oriented measurement-microphone calibration service in the UK, similar to what Cross Spectrum does in the US?

I wanted to do some basic room correction for the low-end in my yet-untreated bedroom studio. Given that money in the long-run would be better spent on some basic treatment, I just bought a cheap measurement mic (Red5 RV200). Now I'm thinking maybe I should have just ponied up the extra cash for a Dayton Audio EMM-6.

In short, the Red5 doesn't come with a calibration file and the company never responded to my query about it. I did some measurements of my room in REW, which showed rather elevated low-end (50Hz - 500Hz) relative to the rest of the response (i.e. when I set the target level in the REW EQ tool using the full 50Hz-22kHz sweep, and then tried to match the response on the above low-end frequency range, it complains about something like 80% of the response being above the target). The anomolies were greater than what could be fixed with the EQ knobs on the back of my studio monitors (Focal Shape 50).

I generated impulse files from the filters estimated for the left and right channels and loaded them up in a convolution plugin. Upon listening, the result feels really anemic. However I can't tell if I just had already gotten used to a large bass response or whether something else over-emphasized those frequencies during measurement. So, naturally, I start to worry about the calibration of the mic and that maybe it's not so flat afterall.

So, basically, I'm wondering if there's a cost-effective way in the UK to have the mic measured and a calibration file generated so I can redo the measurements with confidence, or whether that cost is just going to end up matching that of a DA EMM-6 or a UMIK-1 in the end. All I can find are industry-oriented, certified calibration services, which are way beyond what I need.
 
You are just comparing relative values so calibration is unnecessary.
Keith

I'm not sure I understand. Can you please clarify?

If the microphone is more sensitive in the 50-500Hz range, for example, then on a frequency sweep using REW in a perfectly neutral room, those frequencies would show up as accentuated, would they not? So then how would I differentiate room-induced bass bloom from microphone inaccuracy?
 
We get mic's calibrated at work and it ain't cheap.

I noticed, which is why I was looking for a more "hobbyist"-oriented, non-certified service like the one in the US. I certainly don't need anything certified.

If calibration is irrelevant for home room measurement and EQ, why does REW make such a big deal about it?

FWIW, here are the responses of the left and right channels, averaged each over 5 sequential measurements, for the 50Hz - 1kHz range (the monitors' specified frequency response range low-end is 50Hz):

freq-response-50-1000Hz_small.png
 
I would change the y to 5dB increments, keep the 50dB range and perhaps add some smoothing 6/12 it will make it easier to read, I would also measure from 20-20kHz to give you an idea of the whole response.
Keith
 
desk.jpg


Focal Shape 50, lifted up with ISOAcoustics stands, but sitting at the back of a relatively deep desk. As you can see, the monitors are up against the wall, which Focal claims is not a problem. Well ok maybe "not a problem" is stretching it, but given that they're not rear-ported but instead use passive radiators on the sides, they're a bit less sensitive to placement near a wall.

I'd rather not post a picture of the room, but the dimensions = 420 x 270 x 235. The desk is situated at the midpoint of the longer wall (firing across the shorter width), which I know is not ideal. The room is carpeted and has a lot of furniture in it. It has a deep bay window on one of the 270cm walls.

I took a bit of a shortcut in measuring, which was that I only had a desktop mic stand, so I, um, "creatively placed" a table on its end in front of the desk to hold the mic stand to position it at my listening location. On the chance that this induced additional reflections (doubtful considering the wavelengths in question here, but I'm nowhere near an expert), I've just ordered a different stand to repeat the measurements more cleanly. Obviously I wouldn't have taken that shortcut if this were a professional studio and I wanted to treat the whole range of frequencies, but I was just aiming for "quick and dirty" to clean up the low-end.....but maybe I went too dirty...

Thanks for your advice.
 
Wow somehow I never noticed that the amroc calculator plays tones. Yes, with the convolution filter off, the 145Hz mode is definitely prominent. The filter evens it out very nicely. It's a bit harder for me to evaluate the flatness by ear at freqs < 150Hz
 
A mic that's known to measure flat is important if your aim to achieve an in-room response that conforms to a target. If, however, you are simply wanting to measure the effect of moving listening position, speaker position, the effect of adding room treatments, etc, then as long as you use the same mic for all measurements it doesn't matter if it's calibrated.

Personally, however, I'd prefer to use a mic I knew had an essentially flat response, that's why I bought a UMIK-1. I compared it against my previous mic, an uncalibrated ECM8000, which is criticised by many for having poor unit to unit consistency, and found the two mics responses matched to within a dB. Perhaps I just got lucky?! I still prefer the UMIK-1 however since it is plug and play and has more than paid for itself given the number of measurements I've taken with it over the years (easily in the thousands!).

ecm8000%20vs%20umik1%20left%20smoothed_zpsjtohe2ku.jpg~original
 
A mic that's known to measure flat is important if your aim to achieve an in-room response that conforms to a target. If, however, you are simply wanting to measure the effect of moving listening position, speaker position, the effect of adding room treatments, etc, then as long as you use the same mic for all measurements then it doesn't matter if it's calibrated.

This was my understanding too but maybe I wasn't clear in my original post since I mentioned room treatment. My goal today is to implement a convolution filter to match a target curve in the low/low-mid frequencies. "One day" I will buy some room treatment to improve things. I know that usually it should be done the other way around but funding and domestic peace dictate otherwise at the moment.

This is why I didn't understand the "just comparing" comments. My listening position, my monitors, and their position are all fixed for now, so I don't understand what I am comparing.
 
AKG used to offer a calibration service for their measurement mics but those are pro mics - you could buy three UMIK-1s for the cost of the AKG and for home room setup the former is more than good enough.
I've used the UMIK during 'speaker design with the calibration file enabled and no issues.
 
I wanted to do some basic room correction for the low-end in my yet-untreated bedroom studio. Given that money in the long-run would be better spent on some basic treatment, I just bought a cheap measurement mic (Red5 RV200). Now I'm thinking maybe I should have just ponied up the extra cash for a Dayton Audio EMM-6.

I bought an EMM-6 a few years ago because it came with a calibration curve. The calibration was nonsense and clearly taken in the presence of significant reflections. I guess one has to expect this sort of thing when buying extraordinarily cheap generic hardware from Asia. Dayton to their credit offered to get the microphone calibrated at CSA but I needed to use the microphone (which was fine) and had had problems with shipping audio equipment from the states before in terms of long delays, driving to warehouses to argue with customs about costs, import taxes, etc... It didn't seem worth it for a cheap microphone.

The easiest way to calibrate a microphone is to compare it to a calibrated microphone. If you know someone with a calibrated microphone then this is what I would suggest. An idea of the shape of the calibration curve and the variation in response from mic to mic can be seen in this old blog post from CSA. Armed with an idea of the smooth shape of the calibration curve if you have speakers in good condition with competent measurements published on the web you may be able to use them to estimate an approximate calibration curve. Be aware of the difference in near field vs anechoic measurements at low frequencies. Not perfect but likely to be an improvement over a non-calibrated mic.

In short, the Red5 doesn't come with a calibration file and the company never responded to my query about it.

Like Dayton, Red5 will be almost certainly paying to put their name on a generic microphone and so if they haven't ticked the box and paid extra for a calibration curve of some form from the manufacturer then there won't be one.

Apologies, I'm still learning my way around REW. Here you go.

If you want to use your measurements to back out the effects of low frequency resonances the last thing you want to be doing is destroying the shape of those resonances by averaging frequency bins. Ensemble averaging is good. Frequency averaging is not. I think Toole's book contains an example of the consequences of doing this.

If you want to learn how to use a microphone the Primers and Handbooks section of the Bruel&Kjaer library is an excellent resource and free if you register. I have a set in book form on my shelves but, inevitably, the microphone one has been borrowed and not returned.
 
@h.g. thank you for the detailed response.

Like Dayton, Red5 will be almost certainly paying to put their name on a generic microphone and so if they haven't ticked the box and paid extra for a calibration curve of some form from the manufacturer then there won't be one.

This is definitely the case, which I had already figured when I was still looking at mics, since you can see a bunch of identical mics with different branding in different shops.

An idea of the shape of the calibration curve and the variation in response from mic to mic can be seen in this old blog post from CSA.

This points to my nightmare scenario, linked in the second update on the article:

fr_plot_351.jpg


...a frequency amplification right around where I measured my peaks to be! I know that's unlikely to be the case but the uncertainty of it gets me. As mentioned in my first post, I know: you get what you pay for.

If you want to use your measurements to back out the effects of low frequency resonances the last thing you want to be doing is destroying the shape of those resonances by averaging frequency bins. Ensemble averaging is good. Frequency averaging is not. I think Toole's book contains an example of the consequences of doing this.

Just to clarify, by "ensemble averaging", you mean to take multiple measurements and average across those (i.e. arithmetic mean response at each sampled frequency)? In that case, fortunately that is what I did (average across 5 measured responses).
 
Just to clarify, by "ensemble averaging", you mean to take multiple measurements and average across those (i.e. arithmetic mean response at each sampled frequency)? In that case, fortunately that is what I did (average across 5 measured responses).

Yes but you tend to need orders of magnitudes more than 5 if the grass is to be reduced to a low level. You also need to make sure the frequency bins are narrow enough to resolve the high Q resonances of interest. At low frequencies this means accurate measurements take a long time to obtain and the software default settings can be inappropriate. A fair proportion of people that casually use microphones in the home for measurements have, perhaps understandably, not taken the time to learn how to use them effectively which is why it is fairly common on the web to come across suggestions to smooth curves to make them look nice when it is inappropriate.

If you want to show-off the frequency response of your speakers then smoothing can be helpful. If you want to use the details in that response to calculate an inverse response then destroying the detail in it by smoothing is not sensible because the calculated inverse response will be wrong (see Toole's book for an example). Of course there can be exceptions if what is being used to perform the inversion can only do so in a coarse rather than a resolved manner so it is not black and white.
 
Yes but you tend to need orders of magnitudes more than 5 if the grass is to be reduced to a low level. You also need to make sure the frequency bins are narrow enough to resolve the high Q resonances of interest. At low frequencies this means accurate measurements take a long time to obtain and the software default settings can be inappropriate. A fair proportion of people that casually use microphones in the home for measurements have, perhaps understandably, not taken the time to learn how to use them effectively which is why it is fairly common on the web to come across suggestions to smooth curves to make them look nice when it is inappropriate.

If you want to show-off the frequency response of your speakers then smoothing can be helpful. If you want to use the details in that response to calculate an inverse response then destroying the detail in it by smoothing is not sensible because the calculated inverse response will be wrong (see Toole's book for an example). Of course there can be exceptions if what is being used to perform the inversion can only do so in a coarse rather than a resolved manner so it is not black and white.

OK that all makes sense, thanks again for the advice. I did notice that 5 reps didn't do much to reduce the variation, so I'll try to aim for a lot more (I'll have a bit of time off work from tomorrow). As for the sliding-window smoothing that @Purité Audio suggested, I had assumed it was just to be able to visually inspect the response more easily. I hadn't yet considered trying to build a filter from the smoothed response, but your objection to the idea made it immediately clear to me why it's a bad idea (just by working through it in my head, no need to see Toole's result...I can already imagine the negative effect!).

I'm trying to strike a balance between my pathological need to do things with utmost correctness and my financial and domestic need to cut corners and aim for gradual improvement. In any case, I think it's time to pick up a book on the topic.
 
Yes but you tend to need orders of magnitudes more than 5 if the grass is to be reduced to a low level. You also need to make sure the frequency bins are narrow enough to resolve the high Q resonances of interest. At low frequencies this means accurate measurements take a long time to obtain and the software default settings can be inappropriate.

Could you elaborate a little on this, @h.g. ?

When you say the measurements take a long time to obtain are you referring to a stepped sine measurement? REW's default automated sine sweep takes only 6 seconds to complete and can be viewed without smoothing, and REW's RTA function can be displayed with 1/48th smoothing which shows almost as much detail at low frequencies as the unsmoothed sine sweep.

With regards to the number of measurements, is there not a risk that taking too many measurements spread out around the MLP will essentially mask the specific frequency and amplitude of the resonances heard at the MLP?
 
Most of the cheap measurement mics just use a small off the shelf £0.50 electret capsule. The Panasonic WM-61 was the go-to one back when I made my own with all the supporting electronics. Although dirt cheap they were known to be very accurate up to about 10 -12KHz. They go down to a few Hz at the bass end.
 


advertisement


Back
Top