Dear colleagues As some of you may remember I built my own DMTD in December and I struggled with some problems. It had quite a bit of stability issues. A noise floor test with some 10 MHz sources gave a 2e-12 noise floor at 1sec, not so good! [image: grafik.png] Well, after I had this result, I was a bit disappointed and set the project aside for a couple weeks (months, actually) and worked on other stuff. Recently, I decided to go back to this DMTD project and tried different things. I found out that my first OpAmp in the amplifier/limiter for the zero crossings didn't like the fact that it was clipping! so I reduced its gain slightly. My zero cross detector has now a gain of 20 in its first stage (previously: 100) and a gain of 100 in the second stage (unchanged). The IF bandwidth is 16 Hz. And after this small modification, the first OpAmp in the zero cross detector does no longer saturate, stays in its linear region, and I get a quite good result: around 2.5e-13 at 1sec! [image: grafik.png] also note that this test was a very "quick and dirty" test, i.e. nothing is mounted in a housing, and I didn't let my instruments warm up before. I used a HP 5335A as TIC, and a HP 8663A 10 MHz out for the DUT, and let the HP 8663A generate 10MHz+10Hz for the REF. Just connected everything and recorded from the beginning. In my opinion, under these circumstances, 2.5e-13 is a quite acceptable result, isn't it? it might not be good enough to measure a BVA, but it should definitely be OK to characterise some "normal" OCXOs. After all, I am quite happy with the result, or should the floor be even lower? The next step is definitely to build all four channels. With a 4 channel DMTD, I can compare 3 DUTs against each other (three corner hat!) and use the fourth channel to estimate the noise floor! BR Tobias HB9FSX

Hi At the top of this paper, http://www.stable32.com/A%20Small%20DMTD%20System.pdf <http://www.stable32.com/A%20Small%20DMTD%20System.pdf> there is a plot of what the noise floor of a DMTD *should* look like. The key feature to note is that each time tau goes up by 10X, the floor goes down by 10X. Eventually, you will hit some sort of floor due to thermal issues, so it doesn’t just keep going down forever and ever. At short tau ( say < 1,000) it *should* follow the 10X rule. The key in looking at any ADEV information is that the entire plot is what matters. You can’t just look at one number and move on. Something is still going on in your setup at short tau that is degrading the noise floor significantly. Bob > On Apr 26, 2021, at 1:21 PM, Pluess, Tobias <tpluess@ieee.org> wrote: > > Dear colleagues > > As some of you may remember I built my own DMTD in December and I struggled > with some problems. It had quite a bit of stability issues. > A noise floor test with some 10 MHz sources gave a 2e-12 noise floor at > 1sec, not so good! > > > [image: grafik.png] > > Well, after I had this result, I was a bit disappointed and set the project > aside for a couple weeks (months, actually) and worked on other stuff. > Recently, I decided to go back to this DMTD project and tried different > things. I found out that my first OpAmp in the amplifier/limiter for the > zero crossings didn't like the fact that it was clipping! so I reduced its > gain slightly. My zero cross detector has now a gain of 20 in its first > stage (previously: 100) and a gain of 100 in the second stage (unchanged). > The IF bandwidth is 16 Hz. And after this small modification, the first > OpAmp in the zero cross detector does no longer saturate, stays in its > linear region, and I get a quite good result: around 2.5e-13 at 1sec! > > [image: grafik.png] > > > > also note that this test was a very "quick and dirty" test, i.e. nothing is > mounted in a housing, and I didn't let my instruments warm up before. I > used a HP 5335A as TIC, and a HP 8663A 10 MHz out for the DUT, and let the > HP 8663A generate 10MHz+10Hz for the REF. Just connected everything and > recorded from the beginning. In my opinion, under these circumstances, > 2.5e-13 is a quite acceptable result, isn't it? it might not be good enough > to measure a BVA, but it should definitely be OK to characterise some > "normal" OCXOs. After all, I am quite happy with the result, or should the > floor be even lower? > > The next step is definitely to build all four channels. With a 4 channel > DMTD, I can compare 3 DUTs against each other (three corner hat!) and use > the fourth channel to estimate the noise floor! > > BR > Tobias > HB9FSX > <grafik.png><grafik.png>_______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com > To unsubscribe, go to and follow the instructions there.