Jack, do you have any graph over your findings? If 2-3 uOhms is true, I'm extremely happy.
No, that was for the Jung-Didden Super-regulator, and I am left to believe that it was a fluke -- since I can't replicate it. As Jan has pointed out, the measurement setup is very critical.
AP has a test procedure for impedance measurement which nulls the impedance of the cables and analyzer input impedance, but I haven't been satisfied with the results thus far.
Last edited:
Baseline Sanity Check -- Here's my baseline impedance -- in the 1995 articles I think the baseline was between 3 and 4 milliOhms.
An externally hosted image should be here but it was not working when we last tested it.
If I read it correctly it's about 30u ohms - close to what I found at a real implementation. Correct?
But it is too flat with freq, up to several kHz, which suggests to me there's still some parasitic resistance in the measurements that's not from the DUT.
What were you actually measuring here?
jd
This is frustrating, what does it mean at all, baseline impedance? If you take the time to report results, please take another 30 seconds to say a few words about the setup. Otherwise your results report is wasted, unfortunately.
This is the "Baseline Impedance", i.e. the impedance of a jumper across the impedance measuring nodes from 10Hz to 200kHz at the point where it would make contact with the SR. The jumper is flat copper tape, a foot of which (30.5 cm) has DC resistance on the order of a milliOhm, obviously there's less distance involved here. There's only a tiny bit of the insulation of the two conductors between them (I stripped it out) so if there's any leakage, it's the conductivity of "air". It would appear from the article that WJ's impedance at 20 Hertz is 4.3 milliOhms. Mine is on the order of 32microOhms.
An externally hosted image should be here but it was not working when we last tested it.
I also ran the test with a 100 milliOhm non-inductive current measuring shunt and it looks exactly like series "C" in the article.
Last edited:
The setup (sorry Ikoflexer) -- as Jan had suggested I used the 20R output impedance of the AP generator, calculate the difference between the generator's output, and the output at its BNC jack to calculate the current (1V - V(GenMon)). This has a 50 ohm cable going to a 10,000uF/50V capacitor to inject the perturbation signal into the DUT. I download the voltage measurement into Excel and change it to current.
The DUT is connected to an amp very similar to the SSM2017 shown in the 2/95 article -- only difference is that an SSM2019 is used as the '2017 is obsolete. this amplifies the output across the node so that the impedance of the cables connecting to the AP is nullified. I can verify that the noise of the SSM2019 amplifier (powered by batteries) is less than 2nV/SQRT Hz. at 10Hz. (The input impedance of the AP isn't ideal -- it's 100KOhms/100pF or therabouts and add to this the impedance of the cables.) I can select gain of 10,100 or 1000 with the amp, and check to make sure that the measurements are consistent.
The DUT is connected to an amp very similar to the SSM2017 shown in the 2/95 article -- only difference is that an SSM2019 is used as the '2017 is obsolete. this amplifies the output across the node so that the impedance of the cables connecting to the AP is nullified. I can verify that the noise of the SSM2019 amplifier (powered by batteries) is less than 2nV/SQRT Hz. at 10Hz. (The input impedance of the AP isn't ideal -- it's 100KOhms/100pF or therabouts and add to this the impedance of the cables.) I can select gain of 10,100 or 1000 with the amp, and check to make sure that the measurements are consistent.
Thank you for the explanation; if I understand correctly, in the attached image, U3 is the SSM2019 preamp, and the shunt copper tape is connected between DUT1 and DUT2?
This part is not clear, sorry. The generator output is not exactly at the BNC jack? Or do you mean the reported/set generator output, which is meant to be some value, but at the BNC jack something else is measured?
What do you measure the rms voltage with? I'm finding myself not trusting very much the values I measure with my instruments, as they don't agree with one another even on one decimal place.
This part is not clear, sorry. The generator output is not exactly at the BNC jack? Or do you mean the reported/set generator output, which is meant to be some value, but at the BNC jack something else is measured?
What do you measure the rms voltage with? I'm finding myself not trusting very much the values I measure with my instruments, as they don't agree with one another even on one decimal place.
Attachments
I was using that setup -- the amplifier OPA544 or whatever doesn't have such hot bandwidth and isn't that quiet, so I just took up Jan's suggestion. Something like an LME49600 has plenty of bandwidth and is very low noise but as of yet I am too lazy to install it. You have to remember to bring the DC on the electrolytic low by shorting the leads, the AP isn't happy with DC on the generator output.
The SSM2019 has a pair of 49.9 ohm current limiting resistors, some back-to-back eletrolytics (220uF Elna Silk) for bypass and diodes to clamp DC at its inputs. It amplifies the voltage across the 100 ohm resistor
I took a look at an LM317 with 120uF bypassing the adjust pin, and the results aren't dissimilar from the 1995 article -- now "lemme guess" -- did the authors adjust for the complex input impedance of the AP-One?
The SSM2019 has a pair of 49.9 ohm current limiting resistors, some back-to-back eletrolytics (220uF Elna Silk) for bypass and diodes to clamp DC at its inputs. It amplifies the voltage across the 100 ohm resistor
I took a look at an LM317 with 120uF bypassing the adjust pin, and the results aren't dissimilar from the 1995 article -- now "lemme guess" -- did the authors adjust for the complex input impedance of the AP-One?
An externally hosted image should be here but it was not working when we last tested it.
I didn't - I did use the 600 ohms input impedance option for the AP.
To Iko - the AP has a built-in Zout. Say you use the 50 ohms setting, and set the generator output (which is what the AP sends into the Zout) as Vgen. Then you measure the output voltage (which is what comes out of Zout, into the BNC which sends it to the DUT), call it Vdut. Then you calculate the current into the DUT as (Vgen-Vdut)/50 ohms
jd
The 600 ohm inputs also have 100k/100pF bypass -- there's some protection circuitry as well. 600 ohm input should be much less noisy! 3nV vs 40!
There's a tiny issue with the analog bandpass filter -- it's a 4th order filter -- at the top and bottom of the range the bandwidth limitation of the AP generator cuts off the tails of the curve, so the area under the curve really isn't comparable. I calculated the Q to be 5, but I may be wrong. I think you can pretty much throw out the results below 20Hz and above 100kHz. To calculate the root Hertz noise WJ used the approximation "V(nV/ SQRT Hz) = V/ (48 * SQRT f). The digital filter is 10th order, but it only ranges to about 22kHz.
Looks like I should hang on to my '3577A!
There's a tiny issue with the analog bandpass filter -- it's a 4th order filter -- at the top and bottom of the range the bandwidth limitation of the AP generator cuts off the tails of the curve, so the area under the curve really isn't comparable. I calculated the Q to be 5, but I may be wrong. I think you can pretty much throw out the results below 20Hz and above 100kHz. To calculate the root Hertz noise WJ used the approximation "V(nV/ SQRT Hz) = V/ (48 * SQRT f). The digital filter is 10th order, but it only ranges to about 22kHz.
Looks like I should hang on to my '3577A!
Sounds like a good opportunity to hack the AP. I have schematics and board layouts...
Sold mine yesterday. Getting something better.
jd
The '3577a seems to be around 15nV/rtHz. The SSM2019 amplifier is pretty consistent at a little over 2nV/rtHz.
If we can measure nanos and femtos and picos maybe we can estimate the number of angels which will fit on a pinhead.
Jan: I wouldn't touch the innards of my AP -- and I only very gingerly attacked the innards of my Boonton 1120.
Last edited:
I think I was wrong about my 3585A 8nV/rtHz. How did you verify that the SSM2019 is around 2nV/rtHz?
I looked at the output using the bandpass function, then take the output at 100, 1K, 10K, 100K Hz and divide by 48 x SQRT f. The amplifier is in the chassis for an HP 465A chassis. WJ said 1nV/rtHz, but with the switching, caps and diodes I get a little over 2nV/rtHz. Close enough for government work.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.