(bclee's) RAT DETECTOR PROJECT

Updated January 30, 2000.


INTRODUCTION:

From experience with my own rats, I've long been fairly certain rats made ultrasonic vocalizations (USVs). Among other things, there's a particular noise they make at the vet sometimes -- I can't hear the sound, but I can hear the air leaving the rat's lungs and feel their body, and it's gotta be loud. So I had occasionally thought about making a device to "downshift" a rat's ultrasonic vocalizations into my range of hearing.

Then on May 3rd, the National Public Radio program Sounds Like Science ran a short story about researchers Jaak Panksepp and Jeffrey Burgdorf of Bowling Green State University in Ohio who had interpreted a particular rat USV as laughter. It seems rats make this sound during play and when tickled. And the article mentioned that the researchers had used "bat detectors" to listen in on the rats. So that was it, I was determined to figure out how to make a bat detector.

Click here to see the stories and reference to the paper by Knutson, Burgdorf, and Panksepp.


BAT DETECTORS

Commercial bat detectors start at around $250, which was beyond the range of my curiosity. Plus I wanted to build one myself! Greg and Kathy who were on a similar quest, pointed me towards several amature bat detection sites, where designs and/or cheap detectors were available.

First, a brief summary: There are 3 kinds of bat detectors:
(1) Analog real time (heterodyne)
(2) Digital real time (frequency division)
(3) Analog (but usually recorded to digital) direct recording and playback.

Only the third preserves all information, but has to be played back later, and the systems are quite expensive. The heterodyne systems must be tuned to a specific frequency. Digital frequency division devices typically use microphones which are sensitive to a wide range of frequencies (as much as 20-100kHz) but they tend to loose all amplitude information -- output is a constant volume. Still, this kind of system seemed to have the greatest flexibility, so I thought I'd start there. (See the Bat Detectors page for what's certainly a better explanation than I've just provided.)

Tony Messina has schematics for a simple digital real time detector and many possible modifications, as well as complete kits ($35) and assembled detectors ($50) on his Bat Detector page. I contacted Tony, and he was and continues to be very interested in the project -- he has even started a Rat Detector web page! I got parts from Tony and started experimenting.

You can also buy a kit ($45) or a preassembled ($75) detector of a similar design from Bill Gerosa at his Bat Echolocation and Bat Detectors page. Greg and Kathy got a detector from Bill, which Bill tweaked somewhat to hopefully get better response in the rat range. (See below.)

There is also a page which will tell you how to convert a cheap Radio Shack radio into a heterodyne type detector. The model of radio may be difficult to acquire outside the UK, though.

A final note here -- if you decide to get one of these, you may want to wait until I'm done experimenting some. At the very least, be sure to tell the person you get it from what you'll be using it for. In order to get maximum performance for minimum price, some of these units are very sensitive to electronic noise -- they are usually used outside, not inside.

FYI, Jaak Panksepp (of the rat laughter paper) wrote me, "Most of the work we have done has used on-line counting using the Mini-3 Bat Detector (Ultra Sound Advice, London)--see Knutsen, et al., Journal of Comparative Psychology, 1998, 112, 1-9. More recently we have been using the Pettersson D-980 detector with heterodyne."


RAT USVS

I'll put more here eventually, but suffice it to say that rats make USVs which cover the entire range of 20kHz to 70kHz. Humans (with good hearing) can hear sounds from 20Hz to 20kHz. As for some of the specific noises, I've heard 22kHz for many common adult rat noises, 40kHz for similar sounds in young rats, 50kHz for certain aggressive noises, and 55kHz for rat laughter. All these figures are approximate.

One of the initial concerns both Tony and I had was that bats are quite loud, up to 100dB. Who knows how loud rats are? (Well, not me anyway.) So we weren't sure how well these simple detectors would work.


SUMMARY OF RESULTS

I thought I should add a short summary of my results before you try to wade through the long and incomplete journal entries below. The very short version is that these simple digital detectors are great for bats but not so great for rats. The difference is the environment -- bats are loud, can be observed outside in relatively quiet environments, and are even so kind as to fly through the air and not touch anything. Your rats are much more quiet, generally inside your house which is full of background noise, and inside a cage which makes all sorts of ultrasonic noise every time your rats move. Further, since the digital detector looses much of the quality of the original sound, it can be nearly impossible to distinguish rat sounds from background sounds. So the result is that this detector WILL pick up your rat, but it will pick up many more background sounds than rat sounds, and only very rarely will you be able to tell the difference and even know that your rat has actually made a sound.

I think the wide response range of this detector (which makes it great for bats) is the main problem, and that detectors with a tunable frequency response (such as the analog detectors) would be more productive -- you could tune to the specific frequency range of the USV you are looking for, and at least eliminate background noise at other frequencies. Tony is still working on this detector (for bats and other animals) and at some point in the future I hope to continue this research with new detector designs. I'll let you know here what I find out!


PROJECT JOURNAL
Archived project updates:

The primary purpose for constructing this page is so that people can keep track of various things I tried so that you can all find out how well various detectors might work and what you might expect from them. I'll post some version of this to the rats email list, but for the rest of you, you can find it here.

Please keep in mind that I'm trying to adapt Tony's detector, designed to operate outside in an EMI and noise free environment, to detect much quieter animals in a high EMI environment, rattling around in a metal cage and often stepping on crinkly and loud litter.

Eventually I may fill in some of the earlier emails here as well.

May 22, 1998 -- The first hints

Tony Messina writes:

"Yes, I have also been communicating with Brian C. Lee, so I see how the links came together. I'm copying him on this email also.

"Well, I emailed to let you know that one of the 2 stores I went to did have rats, and they were OK by letting me monitor them for a little bit. ( Though I did raise a few eyebrows ! )

"The first obstacle I had to deal with was the fluorescent lights they had on the enclosures. They generated a lot of electrical interference, which made it hard to use the detector at full sensitivity. I started out by using a less sensitive, low frequency ( 25 kHz ) detector to "listen in". You don't want to use florescents around the detector. If you do, a less sensitive version of the detector needs to be used.

"The rats were not going to be cooperative, as they had just had their cage cleaned out and had eaten. All they were going to do is sleep in the plastic tube. ( there were 5 of them in there ). After a short bit, a rat came out, a pretty white and black one, and sniffed around the food, and sniffed around the detector... but didn't seem to have anything to say - ultrasonically speaking.

"When the rat returned to the tube, it rustled the others up a little, and I did hear two strong clicks ... so I started to think there could yet be hope ... but alas no more sounds as they all just wanted to sleep.

"At this point the pet store worker decided that she'd stir them up a little for me ( something I had hesitated to do, as I know that this was their sleep period - and they would have been more active at night ). She picked up the tube and dumped out the lot of them ... they were only mildly disturbed - and still quiet.

"She removed the white and black one and held it in her hand and stroked it. As she did so the rat was facing away from me and the detector, and I did not hear anything. After asking if it would be ok I made a "kiss" noise - one that I knew was rich in ultrasonic harmonics. In the detector I heard TWO noises - the first, the one that I made, the second I startlingly realized was a response from the rat. I asked her if she heard the rat make any sound ... she said no ! ( I couldn't really tell, with the earphone in my ear. ) So I waited a few seconds and tried again ... again the rat "kissed" back at me - and there was no AUDIBLE sound. The rat was also clearly agitated by the fact that the noise I made was coming from behind, and made a concerted effort to squirm around to check me out. After a couple of these sound exchanges the rat stopped responding. I expect that the first few calls startled it into echoing an alert or alarm call - strictly in the ultrasound region. Once it was more comfortable in knowing the source and nature of the noise it chose to ignore me.

"Anyway that's the story. A brief experiment to be sure, and as I wasn't familiar to the rat, I don't know if the correct kind of environment could have been set up to elicit the tickle response. It was only our first encounter ... I'd have to know a rat a little better .

"That's the story. I ordered a few parts today to start building a prototype ultrasound monitor that could be used for rats, mice, insects, or any other ultrasound source. The bat detector did indeed hear the alarm call of the rat... but certainly didn't hear any playground of giggling rats, but then they wanted to sleep, and weren't exactly in a playful mood.

"Once I've put together a prototype Rat Monitor, I'll email you all with a breakdown of the costs, and possibly a picture of what it looks like. Brian was expecting to do some experimenting with one of my parts kits and get back with some more data. But, for now, I'd say the idea looks like it will work !!!

"Tony Messina Las Vegas, NV"

June 5, 1998 -- My first attempt

Hi everyone,

This is the first of what will probably be a few updates on my rat detector project. (Otherwise known as the quest to hear my rats laugh.)

I ordered various parts from Tony Messina to build and try a few variations on his design. Tony has been experimenting as well. I got the parts on Wednesday and late last night constructed my first detector. Technical details: I had to take out the detuning coil to get rid of an annoying hum, I suspect emi. Keep this in mind of you decide to get one of these eventually -- the detuning coil may not work indoors. So the design I tested last night was Tony's "Enhanced simple bat detector."

By either unplugging everything in a room or by making sure there was a solid object between it and the transducer, I was able to eliminate all the background noise. The detector seemed reasonably sensitive -- it was able to detect my fingers rustling together almost a meter away. Making a "kissing" noise anywhere in the same room was easily detected.

The rats, on the other hand, were somewhat more disappointing. I played with them for about 15 minutes, but did not detect anything convincing. Once I got one of them (Yeti) playing, I'd hear a few clicks every time I touched her, but I could get similar clicks sometimes by touching the cage or just moving. I'm fairly certain it was her, because if I touched the other two rats, nothing similar happened, and it happened every time with her, especially once she was chasing my hand around and nipping at me. But the result was not astonishing. There were other cases where a rat would bump into another one, and there would be a click, but again it wasn't enough to be convincing.

The rats were far too interested in me and my machine which until I removed the coil and unplugged everything was emitting a constant noise which Amy says they had heard even when I was down stairs. So pretty much all they would do was watch me, and I couldn't get them to interact with each other or even to play with me much. So I don't think they had much to say, in terms of USVs or otherwise. As I say, the device seemed fairly sensitive, so if they had been making much noise I'd think I would have heard it. I'll experiment more this weekend, and see what happens.

Tony adds: "The clicks are most likely rat generated ultrasonics ... unless they are sustained or prolonged sounds, they will be just clicks."

June 10, 1998 -- Further attempts, possible detections

I did a little work on the detector over the weekend, here are some details.

Technical details:

I need to edit this down a lot, but no time now. Basically I had a lot of trouble with noise this time around. All the noise problems vanished when I took the detector outside, so I suspect most of it is EMI.

Previously, I had used an 18k potentiometer for the volume control. I put in the 1k volume pot one evening, and I noticed the detector was very noisy, with an almost constant noisy/staticy hum. Changing the volume also changed the pitch of the noise. As there was no apparent source for the noise, I decided it must be emi. Our townhouse shares walls with other townhouses, so who knows what devices are humming away just on the other side of the wall. My current setup has lots of wires that are a lot longer than they would be if I used the board, so my guess is that by changing the resistor value, I was changing the frequency of noise that resonated best in the circuit.

Next I put the 18k pot back in. The system was still noisy; the louder I turned it, the less noisy it was, but the effect was still similar: every time there was a detected noise, it would be followed by about 1 to 2 seconds of a constant pitch staticy hum.

Tony suggested that "most of the situations you describe sound like problems with feedback and oscillation of the high gain amplifiers. All wiring around the 386 and transducers must be kept very short !!! A genuine ultrasonic noise will generate a signal in the amplifiers that in turn starts the amplifiers oscillating until they damp off and once again go silent. You should have a lot less of this kind of problem when you use the board."

Another curious problem I've noticed before -- often when I start, touching the top of the cage (or rats on the top of the cage) results in a hum, but after 5 to 10 minutes (with no changes to the circuit) the same action no longer results in a hum.

The next day (during the day!) using the UST-2, the system was still somewhat noisy. But I found that if I grounded the circuit to me, the noise stopped. I also found that if I placed the battery so that its case was grounded to the circuit, the noise stopped. Hmm, ground loop problems? The UST-2 seemed quite sensitive, although no one was in a talking mood. I did notice that for one rat (Yeti) the UST-2 would detect its sneeze, while the UST-40 didn't. (The UST-40 did detect a larger rat's sneeze.) This says something about the frequency of rat sneezes, I guess!

Tony adds, "I've gotten the same kind of results on a detector with the gain set a bit too high. Usually backing off the gain helps a little ( by gain, I mean the value of the R2.) Also, since the bat detector is usually a handheld unit, capacitive coupling through the case to the hand of the person holding it helps to reduce hum if too much gain is in."

Back to the UST-40 -- when I installed it, the system noise had gone away, and nothing I did (including un-grounding the battery case, but not including turing the volume way down, see below) brought it back. See below for results with the rats using this setup.

After all this I experimented more with the (18k) volume pot, and found that at very low volumes the noise and symptoms from the previous night returned, and that adjusting the volume around low volumes dramatically changed the pitch of the hum.

Later that day, I had a chance to use the detector outdoors on a litter of approx. 3 week old rats, whose cage was (temporarily) under a tree in a backyard. The good news is that there was no background noise at all -- clearly emi is the major problem with using it indoors, and the next time I try it, I'm going to rig up some shielding and see how it works. It worked way better outside that it has ever worked indoors! Tony believes this cinches the noise problem as EMI.

So that's about it for now. I picked up a 10k pot at Radio Shack, and I'll try that and shielding. If I get the shielding to work, I'll probably try experimenting with the detuning coil and the UST-2 some more, but right now noise seems to be the major problem.

Tony suggests, "Keep R1 at about 150 ohms and delete R2 / C3 from your breadboarded circuit. Try to shorten your wires ! Don't let the wire between the transducer and 386 be any longer than 2" - twist the transducer wires together, or use shielded wire.

"Another idea that may reduce EMI - try wiring the transducer directly to the input pin (3) of the 386 - leave out C1 ( you can't do this with the coil in the circuit ! ) The original Simple Bat Detector did not use the coupling capacitor on the input, and it seemed to be more immune to induced interference. I had to add the capacitor when I started using the detuning coils."

Rat Detection results:

Using the UST-2 the first night (very noisy system) I was briefly able to replicate getting the clicks from playing with the Yeti, so it's becoming more convincing.

The next day, with the UST-40 (basically the enhanced simple bat detector again), when Mori woke up, Yeti groomed him, and I detected several clicks.

Then I gave all the rats treats. Afterwards, Mori came back to the cage door to beg for more, and I detected several clicks as he stood there, not moving or making any obvious noise. The pessimist in me says it could have been the cage door creaking, or his teeth grinding (rats grind their teeth when they are happy, the equivalent of a cat purring) but it didn't happen when either of the other two rats returned to the door.

I also found that the water bottle (metal tube with a metal ball) makes terrible noise! And I could detect the noise from Zotz chewing on some hard food, which probably means I'd hear something from a rat grinding its teeth.

Later that day, I was able to try the detector outdoors (At Jennifer's rat get-together) on a litter of approximately 3 week old rats. The rats were in a nest of shredded paper, so it was hard to tell what noises might be the paper crinkling and what might be the rats. Whenever the mother either started nursing or decided it was time to stop, there was a lot of noise (clicks mostly, but almost constant) in the detector, but these also tended to be times when there was a lot of movement in the cage.

June 19, 1998 -- Shielding

I haven't had much time to write or work on the detector. I tried a few things, but probably the most important is shielding -- I wrapped a cardboard box in aluminum foil, and put it inside that. Since then, I haven't experienced any emi problems! Seems to work a little better when I ground the circuit to the shielding.

My big rats just aren't very talkative. I guess if they were fighting or breeding or leading more turbulent lives in general, I'd hear more. I did get a new pair of rats (Phobos and Deimos) but they're still timid and tend to stop whatever they are doing and freeze when I come near, so nothing yet, but they should get over that in a few days. I'm waiting to find out what I hear both when they play and when I introduce them to the big rats in a few weeks. I'll keep you posted.

Amy later tells me she has been playing with the detector and the older rats, and has identified a small number of noises that seem fairly reproducable. Amoung other things she usually detects a sound (a reply?) from Mori whenever she calls his name.

June 20, 1998 -- SUCCESS!

The new rats got over their shyness yesterday, and are definitely detectable -- I have heard my rats laughing! They do it when playing with us or each other, detectable as single or a short series of clean, single clicks.

James Kittock asked about the sound being put up on my web page, but the sound itself isn't that amazing, just clicks. The rat laughter has been described as short chirps, and with the digital processing, the detector is just that, a detector. You can tell when the rats are laughing, but you don't hear the laughter itself. (Well, even if you could, it might still be just clicks!) The effect is sort of like a Geiger counter.

July 8, 1998 -- Testing components

This is mostly a technical update -- my only non-technical note is that even over this short time, it seems to me that the new rats (Phobos and Deimos) are becoming noticable more vocal in the sub-20kHz range (as in eeps I can hear without a detector) and less vocal in the ultrasonic range. They still make a good bit of noise up there, but by now maybe half what they did the first time I detected it.

Tony suggested that I use the divide by 8 output of the CD4024 (pin 9) instead of the divide by 16 (pin 6) to see if I could hear more than just clicks. He also sent me a new coil and transducer (henceforth refered to as the Mouser transducer) both of which should be more impervious to EMI.

I tested out 4 configurations, all using the divide by 8 instead of divide by 16.

(1) UST-40, R1 = 150, R2 = 0
(2) Mouser transducer, R1 = 150, R2 = 0
(3a) UST-2, R1 = jumper, R2 = 0
(3b) UST-2, R1 = 150, R2 = 0
(4) UST-40 w/new coil, R1 = 150, R2 = 0

I had noticed that the sesivitity seemed to be less than before, but I had made a number of (supposedly unimportant) changes (improving wiring, etc.) so I wasn't sure what had done it. I verified that the detector response is (of course) no different on divide by 8 than divide by 16. The "standard" level of sensitivity through these tests was finger rustling detected up to 20" away.

I do seem to be having a bizarre problem where the volume dips after a while. I adjust the pot louder, and it suddenly jumps up to normal again. I turn it down to a reasonable level, it's fine for a while but then fades. I've tried two very different 10k pots, as well as a 20k pot, the same thing happens with each. There don't seem to be any loose connections that could easily explain it, but with the rats nest of wireing it won't surprise me if it goes away once eveything's on Tony's printed board.

A note on "signal to noise": I define the signal to noise ratio as the ratio of probable noise clicks (noises most likely made by rat vocalizations) to the number of clicks detected from any other source. The cage rattling, grunding problems, my fingers rustling, and many other things can produce noise that is detected. Thus, if I hear 5 clicks I know had to come from the rats, then one from the rats bumping the cage, the S/N is 5. If I hear one click that might have been from a rat for every ten I know came from somewhere else, S/N is 1/10. Early on I wasn't sure about the detector because my older rats don't make much noise, so there was very little signal. When I use this thing with Yeti, who makes a few clicks, S/N is about 1. With the baby rats, it's been easily as high as 10 at times.

Oh, and just before I started the real tests, Phobos managed to get INSIDE the rat detector, pull out the earphone jack leads, and chew on the insulation for the ground line to the shielding! This took him mere seconds. The dangers of working with electronics around rats.

Results:

(1) Detected several clicks, a few longer noises, including one extended "grunt" which was entirely in the ultrasonic from an older rat (Zotz) encountering a younger rat. So I'm guessing I was detecting both noises n the 50-something kHz and 20-something kHz ranges. The divide by 8 seems to be the way to go. Approximate signal to noise level of 5.

(2) Range to the finger rustling test seems the same as the UST-40, but I don't pick up a single confirmable noise from the rats. It is more sensitive to cage noises than the UST-40 and (especially!) to my own speech -- I have to be careful not to talk close to it. One, maybe two possible clicks, but the approximate signal to noise level was well under 1. Considering that the rats were very vocal moments before, I'd have to guess this transponder just isn't sensitive in the right bands, but I'd like to give it a few more tries before I say that for sure.

(3a) Using the UST-2 with the jumper in for R1, finger noises are picked up in excess of a meter away. I find also that it easily picks up the noise of rat fur -- rats brushing past each other, me touching a rat, etc. -- way too sensitive for close work! I'll never get a signal over that much noise. So I took out the jumper and went back to R1 = 150.

(3b) With R1 = 150, range is about the same as the others. I pick up a few noises, but again no great results. Signal to noise less than one. By this point, the rats have settled down some, the older rat probably isn't making any 20ish kHz noise, and if the young rats are making any noise it isn't or is just barely being detected.

(4) I put back the UST-40 with the new coil. Not much luck picking up any rat noises, but switching between this and the no coil set up (1) which worked so well initially, I don't pick up anything either way, so the rats must have just quieted down. (I wore them out!) The differences I notice: (4) is slightly less sensitive to the finger test, at worse it has 75% of the range, so I should probably boost he gain slightly. (4) is less sensitive to the cage rattling, more sensitive to the water bottle, so the frequency response has changed noticeably. Also, I was operating unshielded (before dark) and the new coil picked up no noise at all, so that at least works.

That's it for now. I want to run through the sequence a few more times this week before I cone to any real conclusions, but that's the first pass. For now my pick of transducers is still the UST-40, but Tony doesn't want me to give up on the Mouser just yet.


[home]



[rat page]

Why are you here? / What does it all mean? / bclee@umich.edu