On Mon, 26 Feb 2018 21:16:03 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon@On-My-Web-Site.com> wrote:>On Mon, 26 Feb 2018 18:06:12 -0800 (PST), George Herold ><gherold@teachspin.com> wrote: >>Hydraulics... one use and you are spoiled for the rest of your life. >>:^) GH>Yep, Hydraulics are your friend... ><http://www.analog-innovations.com/SED/BBQ_Project.pdf> > ...Jim ThompsonMuscle power is cheaper. Personal forklift. No test equipment collector should be without one: <http://4.bp.blogspot.com/_uuOo8x3WXWE/RnvztdIVNpI/AAAAAAAACA0/EBYvrnoag5k/s1600-h/human_forklift.jpg> <http://www.soceadth.co.jp/index2.html> <https://www.youtube.com/watch?v=NlLLEdXUEWY> (1:34) -- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
FFT analyzer recommendations
Started by ●February 16, 2018
Reply by ●February 27, 20182018-02-27
Reply by ●February 28, 20182018-02-28
On 02/23/2018 03:21 PM, Gerhard Hoffmann wrote:> Am 23.02.2018 um 20:15 schrieb Phil Hobbs: > >> So my new-to-me HP 89441A analyzer is due to be here Tuesday. I noticed >> that it doesn't have the serial cable--is it a regular straight-through >> 9-pin cable, or something weird? > > I also had to make the cable myself. > It wasn't much work, and it is known how to do it. > >Yeah, it's just a straight-through D-9 as it turns out. I see what you mean by "1/f challenged"--it's about 7 nV/sqrt(Hz) in the flatband, but 100 nV/sqrt(Hz) at 1 kHz, which is about where it flattens out. The corner seems to be about 26 kHz, which isn't great. However, with a decent preamp it'll be a great advance over analogue RF spectrum analyzers and scope FFTs. I'm looking forward to trying out that code that you kindly sent me. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●March 13, 20182018-03-13
On 24/02/2018 23:32, Phil Hobbs wrote:> On 02/24/2018 06:29 PM, Phil Hobbs wrote: >> On 02/24/2018 06:23 PM, JM wrote: >>> On 23/02/2018 17:11, Phil Hobbs wrote: >>>> On 02/23/2018 11:15 AM, JM wrote: >>>>> On 21/02/2018 01:29, pcdhobbs@gmail.com wrote: >>>>>>> I have written a program that controls that thing over the network >>>>>>> and measures spectra from 0.1 Hz to 1MHz, one FFT per decade, reads >>>>>>> the results, combines them and plots them with gnuplot. >>>>>> >>>>>>> One needs a converter box from coax ethernet to contemporary >>>>>>> network, >>>>>>> then one just opens port 5000-something on 192.168.178.111 and >>>>>>> simply >>>>>>> reads and writes GPIB-strings. And the coax needs 2 terminations, >>>>>>> even >>>>>>> when the "cable" is only 5 inch long. :-) >>>>>> >>>>>>> No need for GPIB cards and semi-supported drivers. >>>>>> >>>>>>> These measurements of voltage regulator noise have been done with >>>>>>> it: >>>>>> < >>>>>> https://www.flickr.com/photos/137684711@N07/24070698809/in/album-72157662535945536/ >>>>>> >>>>>> >>>>>> > >>>>>> >>>>>> >>>>>>> Just in case you win it and are interested. >>>>>> >>>>>> Hi, Gerhard, >>>>>> >>>>>> I did, and I am. I'll have to gin up a nice low-1/f-noise preamp for >>>>>> it, but that'll be fun. >>>>>> >>>>>> Thanks >>>>>> >>>>>> Phil Hobbs >>>>>> >>>>> >>>>> Note that the software options can easily be unlocked on these >>>>> (although >>>>> they would be of little use for your intended application). >>>> >>>> The cell-phone stuff won't help much, but I'm hoping that I can turn on >>>> the AYB (spectrogram and waterfall) option without scrooching it. >>> >>> You'll be able to do that. >>> >>>> (Option AY8, the internal source, would be useful too, but it's >>>> probably >>>> a HW option.) >>>> >>> >>> Oh, you got an 89431A as well as a 89410A? You'll soon be welding a >>> couple of 42H racks on top of each other to save on floor space. >>> >>>> The software one should be no big issue as long as the floppy drive >>>> works reliably. >>>> >>>> Cheers >>>> >>>> Phil Hobbs >>>> >>> >> I got the complete 89441A 2.65 GHz gizmo. I'll move my spare >> Tek11801C out of the rack to make space. ;) >> >> You can fit a lot in a 7-foot EIA rack. >> > > Speaking of which, I could use a shorter four-post instrument rack with > adjustable rails, but they never seem to come up on eBay--just the > flimsy audio and gigundo server racks. > > Any sourcing suggestions? I bought my present HP one from long-lost SED > regular ecnerwal in about 2009. > > Cheers > > Phil Hobbs > >I like the Zpas SRX open racks, but they not be available in the US. Having said that, I also thought it would be nice to have a couple of smaller (24U) racks, but in practice I thought they were just too bulky. I ended up replacing them with a couple of carts, which seem to be happy enough with upto three boat anchors, two on the table and one on the bottom shelf. https://imgur.com/D6hhXi9.jpg https://imgur.com/s78dEM8.jpg
Reply by ●March 13, 20182018-03-13
On 25/02/2018 16:14, Winfield Hill wrote:> JM wrote... >> >> On 23/02/2018 20:17, Gerhard Hoffmann wrote: >>> Am 21.02.2018 um 02:29 schrieb pcdhobbs@gmail.com: >>> >>>> >>>> I did, and I am. I'll have to gin up a nice low-1/f-noise preamp for >>>> it, but that'll be fun. >>>> >>> >>> I'm working on a chopper preamp that should be flat down to at least >>> 100mHz, but it seems that I cannot verify that with the 89441A. >>> >>> It works with parallel ADG819 analog switches, step up transformers, >>> more low noise gain with ADA4898s on 500 KHz, synchronous demodulator >>> back to baseband and some more gain. >>> >>> A Xilinx Coolrunner II generates the timing from a 100 MHz osc. >>> It looks like I get 120 pV/rtHz. >> >> I designed a CSEM receiver 20 or so years ago with sub 100pV noise >> (0.01-15Hz BW and 120dB overall gain) with a similar architecture. >> These worked with Ag-AgCl field sensors with a source resistance of >> about 5 ohms. > > Just as a reminder, Paul and I made a simple 65 pV/rt-Hz preamp, > all the how-to details reported in AoE III, pages 505 to 508. > >I think Gerhard is interested in measurements in the 0.1Hz region and below, where your preamp would be about 26dB noisier than the one he's working on.
Reply by ●March 17, 20182018-03-17
I've used a number of 89410A devices, and they have exhibited some spread in the noise floor. Usually the flat part is around 8-9 nV/sqrtHz, and the 1/f corner could be in the kHz or tens of kHz. One device had popcorn-like noise in one channel. With a preamp built of 4 paralleled ADA4522s and LT5400 resistor arrays, I managed to get noise floor of about 5 nV/sqrtHz flat down to a few mHz. If there's interest, I can dig up those measurements. I used it to measure excess 1/f noise in resistors hooked up in a biased Wheatstone bridge, like this: https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf Another nice trick with the 2-channel FFT is cross-spectrum measurement. With two good preamps and some patience it buys you another 10-20 dB downwards: https://www.researchgate.net/publication/3622215_Measurement_of_voltage_noise_in_chemical_batteries Cheers, Nikolai
Reply by ●March 17, 20182018-03-17
Am 17.03.2018 um 21:16 schrieb Castorp:> I've used a number of 89410A devices, and they have exhibited some spread in the noise floor. Usually the flat part is around 8-9 nV/sqrtHz, and the 1/f corner could be in the kHz or tens of kHz. One device had popcorn-like noise in one channel. >That would actually be good news, that the noise comes from the input amplifiers. That would mean that one can use cross correlation on the two channels. I had some fear that it might be phase noise from the sampling clock since it is probably synthesized. That would be common to both channels and would not average away.> With a preamp built of 4 paralleled ADA4522s and LT5400 resistor arrays, I managed to get noise floor of about 5 nV/sqrtHz flat down to a few mHz. If there's interest, I can dig up those measurements. I used it to measure excess 1/f noise in resistors hooked up in a biasedYes, especially the setup of the 89410. Wheatstone bridge, like this:> https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf > > Another nice trick with the 2-channel FFT is cross-spectrum measurement. With two good preamps and some patience it buys you another 10-20 dB downwards: > > https://www.researchgate.net/publication/3622215_Measurement_of_voltage_noise_in_chemical_batteriesThanks to plowing with the power of a thousand chicken my setup is only abt. 10 dB worse than that of Fred Walls, but without cross correlation. I still can add that. 1000 chicken means averaging over 20 ADA4898 op amps :-). If you want numbers on batteries that you actually can buy and not some anonymous thingies that they happened to have in the lab, there is < http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf > It is still limited by the 1/f of the 89410A and the undersized input capacitor of the preamp (160 uF foil). I have replaced the foil caps with wet slug tantal, 4700 u IIRC, but that is only a good decade better, just a drop in the bucket. An array of organic polymer electrolytics was too leaky. An array of wet slug tantals costs too much if I have to pay for it from my own money. :-( BTW, two 3.7V Panasonic lithiums are quite OK, noisewise. They seem hard to top. < https://www.flickr.com/photos/137684711@N07/39056813010/in/album-72157662535945536/ > 0 dB is 1nV / rt Hz, DC voltage is 7.3V. The 60 Ohm 1nV/rt Hz reference resistor and the short switch are on the amplifier side of the coupling capacitor, so everything looks much better. The picture of the batteries is to the right. cheers, Gerhard
Reply by ●March 17, 20182018-03-17
On Sunday, 18 March 2018 00:20:48 UTC+1, Gerhard Hoffmann wrote:> Am 17.03.2018 um 21:16 schrieb Castorp: > > I've used a number of 89410A devices, and they have exhibited some spread in the noise floor. Usually the flat part is around 8-9 nV/sqrtHz, and the 1/f corner could be in the kHz or tens of kHz. One device had popcorn-like noise in one channel. > > > > That would actually be good news, that the noise comes from the input > amplifiers. That would mean that one can use cross correlation on the > two channels. I had some fear that it might be phase noise from the > sampling clock since it is probably synthesized. That would be common > to both channels and would not average away. > > > > With a preamp built of 4 paralleled ADA4522s and LT5400 resistor arrays, I managed to get noise floor of about 5 nV/sqrtHz flat down to a few mHz. If there's interest, I can dig up those measurements. I used it to measure excess 1/f noise in resistors hooked up in a biased > > Yes, especially the setup of the 89410. > > Wheatstone bridge, like this: > > https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf > > > > Another nice trick with the 2-channel FFT is cross-spectrum measurement. With two good preamps and some patience it buys you another 10-20 dB downwards: > > > > https://www.researchgate.net/publication/3622215_Measurement_of_voltage_noise_in_chemical_batteries > > > Thanks to plowing with the power of a thousand chicken my setup is only > abt. 10 dB worse than that of Fred Walls, but without cross > correlation. I still can add that. > 1000 chicken means averaging over 20 ADA4898 op amps :-). > > If you want numbers on batteries that you actually can buy and not some > anonymous thingies that they happened to have in the lab, there is > > < > http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf > > > > It is still limited by the 1/f of the 89410A and the undersized input > capacitor of the preamp (160 uF foil). I have replaced the foil caps > with wet slug tantal, 4700 u IIRC, but that is only a good decade > better, just a drop in the bucket. > > An array of organic polymer electrolytics was too leaky. An array of > wet slug tantals costs too much if I have to pay for it from my own > money. :-( > > BTW, two 3.7V Panasonic lithiums are quite OK, noisewise. They seem > hard to top. > < > https://www.flickr.com/photos/137684711@N07/39056813010/in/album-72157662535945536/ > > > 0 dB is 1nV / rt Hz, DC voltage is 7.3V. The 60 Ohm 1nV/rt Hz reference > resistor and the short switch are on the amplifier side of the coupling > capacitor, so everything looks much better. > > The picture of the batteries is to the right. > > cheers, GerhardThanks, Gerhard. You've done quite a lot of good work in this field. I've always tried to avoid AC-coupling and its associated problems. With near-zero voltages (e.g. balanced bridge), you can survive with some amount of offset, or fix it in between the gain stages. For batteries I've tried (with variable success) to measure two of them back-to-back, in counter-series. The point of my 4xADA4522 preamp was not to get super low white noise (5 nV/sqrtHz is nothing to brag about), but to have no 1/f. It was all DC-coupled, including the 89410A in its most sensitive range (10 mV I think). With an upper frequency of 1 Hz and 3201 frequency bins (the maximum), it takes one FFT in more than an hour. With overnight averaging it gets smooth enough to make sense. I'll find those details in the next days, some of them are in my work labbooks. Cheers, Nikolai
Reply by ●March 19, 20182018-03-19
So, here's the device with popcorn noise in one of the channels: http://nbeev.web.cern.ch/device1_psd.png http://nbeev.web.cern.ch/device1_td.png The second device is my own 89440 - the only one I have access to right now. Channel A is clearly better in terms of LF noise. The magenta curve is cross-spectrum. http://nbeev.web.cern.ch/device2_psd_hf.png http://nbeev.web.cern.ch/device2_psd_lf.png All these measurements are on the finest range, 1MOhm inputs with external short, DC-coupled. I coulnd't find the ADA4522-1 preamp measurement files, but I found notes on the setup. The preamp was followed by an SRS amplifier, and the total gain was 10000. That was enough to ensure flat floor down to a few mHz, but around there it did intersect with the 89440A 1/f noise. For my purposes back then it was fine. All of the batteries I tested showed steeper than 1/f behaviour at low frequencies, so they were not good enough as a DC source for my tests. And I did it carefully - temperature-stabilized oven, no stress, mechanical relief, plenty of time to settle, etc... Cheers, Nikolai
Reply by ●March 24, 20182018-03-24
Am 19.03.2018 um 11:28 schrieb Castorp:> So, here's the device with popcorn noise in one of the channels: > > http://nbeev.web.cern.ch/device1_psd.png > http://nbeev.web.cern.ch/device1_td.png > > The second device is my own 89440 - the only one I have access to right now. Channel A is clearly better in terms of LF noise. The magenta curve is cross-spectrum. > > http://nbeev.web.cern.ch/device2_psd_hf.png > http://nbeev.web.cern.ch/device2_psd_lf.png > > All these measurements are on the finest range, 1MOhm inputs with external short, DC-coupled. > > I coulnd't find the ADA4522-1 preamp measurement files, but I found notes on the setup. The preamp was followed by an SRS amplifier, and the total gain was 10000. That was enough to ensure flat floor down to a few mHz, but around there it did intersect with the 89440A 1/f noise. For my purposes back then it was fine. > > All of the batteries I tested showed steeper than 1/f behaviour at low frequencies, so they were not good enough as a DC source for my tests. And I did it carefully - temperature-stabilized oven, no stress, mechanical relief, plenty of time to settle, etc...Hi, Nikolai, thanks for the information. I'm just plowing through the 89441A programming manual to upgrade my control program. The manual leaves of lot of open questions, the only thing they say again and again is that I cannot use options that are not built-in. regards, Gerhard
Reply by ●June 23, 20192019-06-23
On Saturday, 24 February 2018 03:21:07 UTC+7, Gerhard Hoffmann wrote:> Am 23.02.2018 um 20:15 schrieb Phil Hobbs: > > > So my new-to-me HP 89441A analyzer is due to be here Tuesday. I noticed > > that it doesn't have the serial cable--is it a regular straight-through > > 9-pin cable, or something weird? > > I also had to make the cable myself. > It wasn't much work, and it is known how to do it.Hi Gerhard, i have also bought one of these units, 89441A, no serial cable and for the love of peet, i cannot find the cable connection diagram, and standard RS232 didnt work either, could you help out and show me how to do it please? Thanks! Arjen.
