Joss Research: Vacuum: Gross Leak Detection Techniques

A Joss Research Institute Informal Report:

Using an iPhone to Detect Gross Vacuum Leaks
and Making an Ultrasonic Sniffer for Smaller Leaks

(30 January, 2000)

I am debugging a largish nitrogen laser head that I have just built, and I wanted a good way to find the vacuum leaks that seem to be inevitable with any device of this sort. It is often possible to use a stethoscope with a small-diameter tube as its probe, but I have sometimes find it difficult to get good results that way, and last night it occurred to me that I should be able to use my iPhone with a hands-free device. (The opening for the microphone is no more than a millimeter across, and it is on a very small pod that sits on one of the earbud wires.) Rather than try to listen to the sound, however, I thought I would use an appropriate application to visualize the sound. (There are several of these; I chose one called Spectrogram.)

Sure enough, this technique is positionally sensitive, and at least for moderately large leaks it is quite good. As I continue working on this head I will have to deal with smaller leaks, and I do not yet know how well that will go, but I wanted to get this technique out on the Web so people could use it.

(Please note: in terms of the vacuum they require, nitrogen lasers are remarkably crude. The kind I am building operates at about 30 Torr. Detection of small leaks in actual high-vacuum systems requires techniques that are a lot more sensitive than the microphone of a hands-free.)

In the following pictures, the information at and below about 1800 Hz is mostly the sound of the vacuum pump, though there are occasional vertical bars that show either human speech or the microphone being moved around. When the microphone is stationary and is not at or near a leak there is very little going on above 2000 Hz, and that area of the display is mostly just empty.

From left to right:

  • Small: At the left edge of this photo the microphone is fairly close to a rather small leak. In the middle, I move it closer (vertical bar of noise), and then I move it again (smaller noises) until is is on top of the leak (pinkish horizontal line, a little above the “2500 Hz” legend). A truly tiny leak would probably emit primarily around 40 kHz, and would require the use of an ultrasonic “sniffer”, which I am building. (See below.)

  • Larger: This is what a somewhat larger leak looks like, though I actually took this photo as the microphone was getting close to a very large one.

  • Gross: The huge white cloud at the right edge shows the microphone being moved until it reaches the large leak. When it is even a few millimeters away, the change in the display is quite pronounced. (In other words, for sufficiently big leaks this technique is child’s play.)

(My apologies, btw, for the fact that the third photo is slightly out of focus.)


In principle you could use nearly any hands-free, and if you don’t happen to have a handy visualization app on your telephone you can call your phone from another phone and just listen. I strongly suspect, however, that having the visual display makes the task significantly easier.

A Homebrew Sniffer for Smaller Leaks

(1 February, 2009)

As I continue to make progress with the laser head, it becomes clear that I want to be able to detect leaks that are too small to whistle in the frequency range that is audible to humans. There are several ways to deal with this; I am starting to look into a device that is usually called a Sniffer, which downconverts some range of ultrasound, thus making it audible. I have read that vacuum leaks tend to emit around 40 kHz, so I will probably try to downconvert the 30-50 kHz range.

The first step is to find a sensor. I have some Panasonic WM-61A microphone capsules on hand, because I am learning about some interesting stereo microphone techniques, and it seemed to me that these should have at least some response above 20 kHz. I tested one last night, using a 555 timer to drive the piezo beeper from a parted-out microwave oven (thanks to Todd Johnson for suggesting this!), and although the mic circuitry detects the electrical noise from the 555 circuit better than it detects the acoustical wave from the beeper, there is definitely some sound pickup. Here is what I’m using:

This circuit is essentially the one that was developed by Siegfried Linkwitz. (You should go to his site and read what he has to say about microphones, as it is extremely cogent and informative. He also has some very interesting speakers.) WM-61A capsules, btw, are available from DigiKey, and they cost less than two dollars apiece. The ones from Knowles Acoustics, which are on the next page, are better, but they cost more than 10X as much.

The next step is to bring up the level of the signal from the mic. My first try will be with half of an INA2141 instrumentation amplifier chip. The INA2141 is extremely easy to use, has reasonable bandwidth, and can be set to either 10X or 100X amplification by means of a single wire or switch.

(That same evening)

I changed the capsule connection to coax in order to decrease its sensitivity to electrical noise, and ran the output of the mic circuit to one side of the INA2141, with its gain set to 100X. That didn’t give me as high a signal level as I wanted, so I added the second side at 10X. Here is a scope photo, with the 555 timer driving the piezo beeper at about 40 kHz:

As you can see, I now have reasonable signal level with the microphone about 5 inches away from the beeper. I will confess that this is about the best-looking waveform I’ve seen from this setup so far; I took advantage of that fact to photograph it.

Here is a revised schematic:

(I have omitted various connections that are present in the application notes in the datasheet for the INA2141 but are not immediately relevant.)

The next step is to build a downconverter circuit.

I have (albeit with some difficulty) found an article written by James T. Hanson, in which he describes his ultrasonic powerline arc detector, and also a set of notes with a slightly different but basically equivalent schematic. Starting from the article, I have worked up a hybrid between Hanson’s unit and my existing microphone circuit. (Note: if the large image is not quite large enough, change “11c” in the filename to “22c”.)

Don’t yet know how well (or even whether) this will work, but there is one fairly easy way to find out. I may even have an MPF102 transistor here already.

(9 February, 2009)

The circuit is partly working now, but quite a bit of the output from the 555 is showing up in the output of the audio amp, so I am thinking about ways to prevent that. I also want to filter the 555’s output a little, to get rid of some of the high-frequency components. I should note, btw, that the microphone connections are all done with coax, to minimize pickup of electrical noise. This includes the cable from the capsule to the board with the capacitor and the two resistors; the cable from that board to the board with the INA2141; and the cable from the INA2141 board to the board with the mixer and the LM386. (The 555 is, at least for now, on a board of its own.)

(Later that afternoon)

Careful placement of a capacitor has given me something that looks about like a triangular wave and has more reasonable amplitude. I think this will probably do, and I have noted it in the schematic. In addition, I am thinking about increasing the value of the coupling capacitor between the stages of the INA2141, but I haven’t done so yet.

(Early AM, 11 February, 2009)

I am having no end of trouble with this circuit. Yesterday and earlier this evening, the microphone section displayed an uncanny ability to pick up one of the local radio stations; I think I’ve mostly taken care of that issue, though I still hear occasional bits of song in the background. In addition, the 555 circuit has been injecting noise into the LM386 audio amplifier. I am not sure whether the 386 is oscillating or exactly what’s going on, but the result is horrendous roaring hash in the headphones I’m using to listen to the device. This clearly isn’t being caused by the headphones, as I also get it if I use the piezo beeper as an output device. Worse, the hash is intermittent; at some settings of the volume control it comes and goes unpredictably, and there are settings where is not present at all.

(Early afternoon, 11 February, 2009)

It occurs to me that I can probably just filter the output of the mixer FET and feed it directly into the iPhone over a cable, which would simplify matters. I would want to be very cautious, so as to avoid damaging the phone or the FET, but it might be a viable option. OTOH, it is less worrisome to use the hands-free, and I could probably just use a single transistor to amplify the mixer output enough for a small earphone. A 2N2222 or 2N4401 (both of which I have) would probably do just fine for this, and I could still use the potentiometer to adjust the gain. Whether that would still create the horrendous noise remains to be seen.

…More as it transpires…

This work is supported by
the Joss Research Institute
19 Main St.
Laurel  MD  20707-4303   USA

Contact Information:

Email:, where you can replace b with joss (as in Joss Research Institute), and a with my first name (no “h”, only 3 letters).

Phone: +1 240 604 4495.

Last modified: Wed Feb 11 13:24:05 EST 2009

Share this:

  • Facebook
  • Twitter
  • Delicious
  • LinkedIn
  • StumbleUpon
  • Add to favorites
  • Email
  • RSS

I am a Researcher of the Joss Research Institute. I work primarily on lasers and ceramics, with occasional excursions into other areas.

Leave a Reply

WP Socializer Aakash Web