ECM8000 microphone measuring techniques and usage discussion

BoomieMCT wrote:

Yeah, position errors aside, the delta between my mike and Anthony's stayed the same every time we did different runs. Also the difference between horizontal and vertical was very repeatable.

To add another data point: I've been experimenting with various real time analyzers running on my laptop. While measuring pink noise with my ECM8000 I can clearly see that the 10kHz+ part of the spectrum shifts up by at least 5dB when I point the mic at the speakers, and goes back down when I orient it vertically. The rest of the spectrum doesn't change when I re-orient the mic, and nothing changes much if I move it around without changing the orientation.

[Dennis H]

Hey guys,

I don't get over here often enough. This thread and the ECM schematic got linked over at AVS and I was a bit concerned with the schematic. Siegfried Linkwitz has done a lot of experimenting with the Panasonic capsules and has some DIY designs on his web page.

http://www.linkwitzlab.com/sys_test.htm#Mic

In part, he says:

When the cartridge is connected as described by Panasonic it produces fairly high distortion at moderate SPL's and is marginally suitable for serious recording and measurement purposes. The microphone itself is extremely linear, but the built in FET amplifier stage is not configured optimally.

Fortunately, is it possible to modify the external connection to the FET. This involves some delicate work of cutting a trace on the tiny pcb in back of the cartridge and soldering thin, flexible wires to the standard two hookup points, and making a different connection to the capsule housing.

I noticed that the ECM has the capsule wired the conventional way. Obviously they aren't going to do surgery on every capsule. I wondered if their circuit might overcome SL's objections so I emailed him a copy of the schematic and asked. I received a prompt reply (the guy is a real gentleman) and here's what he says.

Hi Dennis,

This mic will definitely suffer from distortion.

Too bad,

SL

FWIW.......

[vrsound]

The ECM8000 is not alone in this phenomenon. In recording orchestras for film sessions we use the omnidirectional Neumann M150 (or if you can find one the original M50). One of it's characteristics is that the high frequency response is very non omnidirectional past 2kHz. Besides it's amazing sound quality this is one of the reasons it's used so much for the Decca tree recording technique. It is not a cheap microphone costing around $6,000 - $7,000. In the Decca configuration there are always 3 of them (That's around $20k for 3 mics). The M150 was never meant to be a measurement microphone as the frequency response chart shows, but it is also not truly omnidirectional.

If the ECM8000 exhibits this type of polar pattern as well then wouldn't it be correct to say the the horizontal orientation would be the only position that would be correct to measure HF response?

I guess the correct question to ask is what is the orientation for the flattest frequency response?
My guess would be on-axis to the intended target. Not really what you want for a measurement microphone but at this cost, beggars can't be choosers!!

Vince

[bjs]

I finally got around to checking out my ECM8000...the results are consistent with the data shown by Anthony in his first post. My mic has almost 5 db variation between horizontal and vertical and even in the best orientation still has several dB of peaking. The test results were entirely repeatable. Sadly, this mic is not suitable for full range speaker testing.

What is the least expensive *calibrated* mic available? Hopefully some resourceful person has found one that mere mortals can afford!

[crigztina]

I realize this thread is old, but I think many people may arrive here looking for info about the ecm8000. I can help with the explanation of the schematic posted and something more...
In order of the signal path:
- C1 (3n3) is a low pass filter, 6dB/octave, cutoff or -3dB point at about 16KHz. Changing it for 1n or less improves the high end response without any undesired side effects. If you're going to change this, use a ceramic 330p to 1n capacitor.
- The intended function of the first transistor, BC118 or BC331 (Behringer has used both), is to make the unbalanced signal a balanced one. At the emitter and collector we have the same signal as in the base (input), but at lower impedance and one inverted with respect to the other. The gain is 1 (0dB). But there is a flaw here: to work as intended C6 should be connected to the emitter of this transistor instead of to ground. I have not opened any of my ecm8000s to see if actual circuit has this mistake or is only the schematic that has the mistake, but no doubt it's a mistake. The effect of this flaw is a 6dB loss of level and SNR when connected to balanced preamps. As is in the schematic, this transistor would do nothing except adding some noise.
- The last two transistors are emiter followers for each of the two balanced signals. Their only but very convenient function is to lower the output impedance. As emitter followers, the gain is obviously 1 (0 dB). C6, C7, C9 and C10 make low pass filtering, the cutoff frequency is 200KHz, well above the audio band. R12 and R13 improve the slope of the low pass filter and provide provide stability and protection against highly capacitive or inductive loads at the expense of raising the output impedance a bit. Still the output impedance remains at an excellent low value. So everything right at this stage.
- So the total gain of the circuit is 1 (0 dB), so it does not have nor need any feedback loop. The design is very good except the flaw described above, that most probably is only in the schematic and not in the actual circuit (I may confirm this if I open my mic sometime).
--------------
Some people modify the capsule to use its FET as a drain follower instead of a common drain amplifier to raise the maximum SPL (sound pressure level) handling. This is completely unnecesary for any purpose other than close drum micing or measurements at levels above 115dB SPL (permanent ear damaging levels). If you intend to make this modification, note that you must also add an ultra low noise amplification stage immediately at the capsule output. When using the capsule's FET as a drain follower, we loose all the amplification we had as common drain amplifier, and because the very low level provided by the electret mic, we need to correctly amplify it very close to the capsule to avoid a severe degradation of SNR. Regardless of author's claims, every amplification circuitry I have seen designed by people that does such modification is simply lame and will convert your good balanced mic into a lot noisier, much higher output impedance and unbalanced ouput one, incompatible with pro mic preamplifiers, only with a bit higher SPL handling and that only in some designs. If you really need to improve the maximum SPL handling of this mic and modify the capsule as drain follower, keep using the original balancing buffer, that's pretty good, and simply add a common emiter amplifier with some feedback between the capsule and the balancing buffer. Use a bipolar transitor, not a FET: after the capsule's FET, we have a low enough impedance so we want here the smallest equivalent input voltage better than the smallest equivalent input current, so a bipolar is better here). Some modification is also needed to the power suply scheme: the added amplifier stage and the BC118 should be powered by more voltage than 6V, but can't be powered from 48 since most small signal transistors can't handle that, ideal would be around 24v given that most small signal transistors specify a maxVce of 30v. A simple zener and a capacitor would do it well. If done this way, the result would be a very similar balanced low impedance mic, with 12dB higher maximum SPL handling and most probably (depends on the added stage) only a bit noisier than the stock ecm8000. Even further you may add a switch to bypass the added amplifier stage for really high SPL handling above 130dB, supposing that the electret itself could handle that. Note that you increase the minimum phantom power requirement to 24v, and that the mic pre you connect it to must be able to handle very high levels. The maximum output of the mic this way would be near +20dBu, while even the highest end pro Yamaha preamps clip at -14dBu without padding or -4 with padding, but many other manufacturers allow near 0dBu levels at mic inputs without padding, and some even handle more than +10dBu. Anyway, all this is only is to use this mic for purposes very different for which it has been designed.
-------------
About the frequency response of this mic, what I have found is that there have been very different production series. My mics are all from the very first series, and sure don't match any of the responses I have seen reported by other owners in the net. I have not sent them to a lab for accurate calibration but I have a high end B&K 4007 measurement mic and by comparison I'm sure my ecm8000s don't show shuch an exagerated peak near 10KHz, but something between +1.5..+2.5dB. So what I got was cheap and reasonably good measurement mics, but you can not trust the high frequency part of the actual piece you get without calibration or at least comparison with another mic you can trust. And if you have to calibrate, you can use actually any small condenser or electret. I purchaed the ECM8000 to avoid moving my B&K (I drop it once and repair cost was $1200), and I was satisfied with its performance, but because I'm using a portable computer for measurement, now I find much more convenient one I made based in a subminiature $1 capsule mounted in a thermoretractile tube of only 5mm diameter. When I calibrated against the B&K, I found this $1 capsule was excellent, even better than my ECM8000s, and nicely fits in a small pocket of the portable's suitcase.

[Anthony]

Good insight. Thanks for the information.

And welcome to the Shack!

[hifisponge]

Anthony wrote:

Okay, so BoomieMCT and I did a battery of tests on our ECM8000 mics.

The setup was my M-Audio MobilePre preamp, test amp, Magnepan MG10.1. ...

The Top trace is comparing his and mine horizontal. The Bottom traces are the vertical ones.

Anthony -

I came across your thread while trying to troubleshoot what I see as a problem with the HF response in my own measurements. I own speakers that use a Scan Speak Revelator 9500 tweeter that is supposed to have flat FR up to 30KHz, yet when I measure them, there is a sharp roll-off above roughly 15KHz. I see a similar problem with the graphs you published.

Here's what I get when I measure one of my speakers at 1M using the ECM and no cal file (vertical mic position).


Here is the response plot of a typical ECM mic as provided by Cross-Spectrum labs.



There is nothing that I can see in the above plot or in the generic cal files that would indicate that the ECM mic rolls off sharply before 20KHz. I also use the M-Audio Mobile Pre, so I wonder if this is contributing to the issue. I would love to get to the bottom of this.

[Anthony]

Hmm, I had not thought of that. I tell you what, I have another calibrated mic on loan to me right now. I will run a couple of sweeps with that in the mix and see if that rolloff is still there. If so -- an M-Audio problem. If not, still and ECM8000 problem, but at least it's narrowed down.

[brucek]

If so -- an M-Audio problem.

How could it be an M-Audio problem, if a full loopback soundcard calibration file was created and returned a flat line after a measure?

brucek

[Anthony]

The loopback is done on a 1/4" to 1/4" unbalanced connection. Not exactly the same circuit as the balanced mic and phantom power. So if that part of the circuit is causing problems it would be impossible to "cancel" it out with the loopback channel.

There is no balanced out of the MobilePre, so you can't even loopback XLR to XLR on the other channel.

[brucek]

The loopback is done on a 1/4" to 1/4" unbalanced connection. Not exactly the same circuit as the balanced mic and phantom power.

Loop unbalanced out, into the positive differential amp (pin 2) of the XLR mic-in using a pad.

brucek

[Anthony]

What about the phantom power? IIRC, that was the original concern that having phantom power on that circuit would be be a problem for the unbalanced input. I seem to remember asking M-Audio tech support and they never got back to me (which is odd because they usually got back to me with answers pretty fast).

I have the cables to do that with (XLR to RCA with RCA to 1/4 adapter). If I can verify that it's not putting 48V DC on the unbalanced connection I'll give that a shot.

[brucek]

What about the phantom power?

Just shut it off with the front panel switch.

You also need to make the little pad to lower the level from line to mic level... I can advise you on that.

When all is said and done, you'll probably find that the mic amp offers a small drop off in the lower end and the upper end will be fine. But, I suppose it never hurts to test it, and of course when you're done you'll have the file to use.

For example, here is my XENXY showing the line-out to line-in (yellow) and then compared against the line-out to mic-in through a pad (green). See how it drops a bit extra at the low end (inconsequential, really, since 10Hz is almost 0dB difference))) and then see no change at the upper end.



brucek

[Anthony]

I thought you meant to run the loop that way during measurements (in which case the phantom power had to be on).

what kind of divider? I have a test jig for speaker workshop that does a voltage step down while impedance matching. But now that I think about it, it matches the power amp and line out of the Mobile Pre. Hmm, I'll have to think about this more.

[Anthony]

I found this:
http://www.uneeda-audio.com/pads/

I think I can make the 100:1 pad at the bottom using a barrier strip and some open ended wires to suit. Then run a calibration sweep and check the results.

I'll still run that second mic just to be sure, but this should only take me about 10 minutes to gin up.

[brucek]

I think I can make the 100:1 pad

Here's a very suitable -30dB attenuator that will lower line to mic level from unbalanced to a balanced XLR mic.

I drew it for you in Visio below. Very simple.

Hook it up and create a soundcard cal file and save it.

Then measure to be sure of your flat line.

Remove the loopback and use that channel for measuring.





When you're done, tell us how much different the soundcard file is from the previous line-out to line-in file.

brucek

[Anthony]

Alright, I'll give that a try tonight and post the calibration results (1/4 to 1/4 and 1/4 attenuated to mic). thanks.

[brucek]

I'll give that a try tonight and post the calibration results

Great, whenever you get the time, I think it will be a benefit to others who will ask if they should be concerned about the mic preamp portion of the M-Audio Mobile Pre. I suspect that the mic preamp will not be significant in relation to the standard line-in to line-out calibration that most do.

I've proven that the XENYX802 preamp is not worthwhile including in the soundcard cal routine, and this will add another to the list (or not). Now we'll know for sure.

Many users aren't really capable of this fabrication (as you are), so this will be a help... If possible, it would be good to see the graph in the same scale as mine above, with vertical = +/-30dB and horizontal of 2Hz-25KHz.

brucek

[hifisponge]

Anthony wrote:

Alright, I'll give that a try tonight and post the calibration results (1/4 to 1/4 and 1/4 attenuated to mic). thanks.

Thanks for taking this project on Anthony. I certainly don't have the experience needed to construct the attenuation circuit to test the mic input.

All the best,

- Tim

[Anthony]

Okay, some results:
Dashed is the Line to Line (1/4" to 1/4" unbalanced), Red is Line to Mic with an attenuator circuit to match the levels.

The only difference in the two is the scale. I was expecting differences in the high end and found none, but the low end was pretty interesting. Not a huge difference, but measurable for 20Hz up a bit.

[hifisponge]

Anthony -

So I guess that rules out the M-Audio sound card / preamp. Must be user-error on my part somewhere.

What do you make of the HF roll-off in the graphs you displayed in your first post? What might be the cause?

- Tim

[Anthony]

Nope, see the new thread I've started. I have more results that just add to the confusion. I split this thread into a new one to discuss just the MobilePre and the ECM8000 so we can figure out what's going on.

[brucek]

Okay, some results:

Nice work Anthony. So, the results are as expected. The low end is a bit off and the upper end is much the same. Even down to 10Hz, it wouldn't matter too much which file you use, so I think we're safe in advising members that a line-out to line-in for external soundcards is suffice for the soundcard calibrate loopback test.

brucek

[Anthony]

Okay, more weirdness. Below is a series of measurements I made with the new soundcard cal file (which seemed to only affect the low end anyways).

Speaker is a Magnepan MG10.1. Measurements were at the listening position. Red is the current ECM8k cal file, Blue is the old cal file, and Green is the Rat Shack meter used as a mic (with appropriate cal file and C weighting)


So with 2 different mics, I get the same fast rolloff in the top octave. It looks like the new mic file fixes the baffle "bump".

To see if it's a vertical/horizontal thing, I ran a semi-nearfield measurement of the speaker both horizontal and vertical using the new "split" ECM8k cal file.


Purple is vertical, yellow is horizontal. Both have the rolloff at 15kHz.

So I believe we've ruled out the mic, and via nearfield measurements ruled out the speaker. It's not the calibration or loop, but it seems to be the soundcard.

Any thoughts folks?

[brucek]

Any thoughts folks?

Well, it's as expected. The old ecm8000 cal file did not account for baffle step, and the new one does. The plots of the cal files themselves easily reveals why the two ecm graphs of the speaker are different. I do have more confidence in the newer ecm file (as far as generic files go).

The RS meter reveals what we already know about the RS meter, in that somewhere between 3KHz and 10KHz they all demonstrate a peak. Simply not usable above 3KHz.

brucek