EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: SharpEars on September 07, 2020, 09:22:00 pm
-
I am looking at various scopes but all of them have either 1mV or 0.5mV per division as the minimum vertical measurement. If I am looking to measure noise on power supplies that is in the 10s to 100s of microvolts, what would be a good oscilloscope for that (preferably not something that costs tens of thousands of dollars).
Frequency of the scope is less important, since measurements of noise beyond 20 MHz are less critical (i.e., most of the noise is <5 MHz if not hundreds of kilohertz if I was to use it with switching supplies/regulators). A four input scope would be preferred.
I was looking at both the Rigol's 5000 series and Siglents SDS2104X Plus, but they all have the same problem of only going to 0.5mV per vertical division at their highest sensitivity. It would be nice if something offered 100uV per division on a voltage up to 12 V that you adjust the zero point to, to check for noise.
-
You need be aware if max sensitivity is real or some magnification of a lower sensitivity.
Down in the 100's uV is tricky due to noise from all sources however averaging and statistics measurements can help immensely to obtain reliable results.
I tried just what you're proposing to measure the ripple and noise of the linear SPD1000X PSU's that are spec'ed at <350uV and it's not straight forward. :o
-
Define the bandwidth of measurement first, many of the noise sources (and product specifications) change depending on that.
uV measurements are usually "narrow" bandwidth of 10's of kHz or less (DC to Audio, or around a particular carrier).
-
How about building an signal amplifier for the frequency range you are interested in?
https://youtu.be/Xn9LNgVKdqI
-
Most of the times this is done with low noise amplifiers that bring the signal into clear emsurement range of modern digital scopes.
Now, a scope useful for that could be the Nicolette sigma 30, or a tektronix 7000 series with 7a22 plugin. Both of these only found on ebay and similars
-
A good amplifier frontend may help with the very low signal measurements (or an FFT analyzer with a high resolution converter), but if you want dynamic range and somewhat lower noise floor, look towards 10 or 12 bit scopes. Something like the RTB2004 is cheap enough to be "within reach", but offers a 10 bit ADC and a good ENOB and low noise setting that is difficult to match on an 8 bit converter.
-
You would be better off using a spectrum analyzer.
For PSU noise, you probably don't care about the particular waveform, but more about the noise spectrum, and how much in a particular band. That's precisely what a SA will do.
Some SA's don't go to very low frequencies (e.g. < 5 kHz) -- you can use a fast sampling DMM (e.g. Agilent 34411A) to measure there.
-
A proper probe is probably more interesting in this case than the scope itself.
Other options are AC millivoltmeter or spectrum analyzer although the millivoltmeter may not go up to 20 MHz and the spectrum analyzer may not cover the lower frequency range.
Note: DMMs are typically quite bad in measuring low AC voltages. Their lower measurement range is limited to roughly a few percentage of their range thus their sub-mV accuracy will be terrible even if they are able to show any reasonable value (even that can become problematic in the 100µV range). Also their upper frequency limit is quite low even with reasonable BNC converter.
-
You need be aware if max sensitivity is real or some magnification of a lower sensitivity.
Down in the 100's uV is tricky due to noise from all sources however averaging and statistics measurements can help immensely to obtain reliable results.
I tried just what you're proposing to measure the ripple and noise of the linear SPD1000X PSU's that are spec'ed at <350uV and it's not straight forward. :o
Even the little X-E's can do a reasonable job of this and now with a little time to dig out an old post:
Some quick facts/findings from the first I've seen of these:
2s boot time.
Quiet smart fan. 80mm. At no time with the following tests and loads did the fan speed increase.
2 sets of 650mm leads with croc clips, one set with open lug terminals and the other with molded banana plugs.
Output enable soft key driving an output relay. Note, the relay is only bought into ON switching function at voltages above 4.4V settings below which silicon is doing the switching.
19mm terminal spacing.
Overshoot test, high current.
12V set, 8A set.
12V halogen auto lamp, 5A draw.
DSO SDS1104X-E, 10x stock probe, Normal triggering.
(https://www.eevblog.com/forum/testgear/new-spd1168x-siglent-psu/?action=dlattach;attach=478610)
Ripple check.
Load as above.
(https://www.eevblog.com/forum/testgear/new-spd1168x-siglent-psu/?action=dlattach;attach=479366)
-------------------------------------------------------------------------------------------
Overshoot test, low current.
4.4V set, 8A set
Red LED with 1k current limit resistor.
5mA displayed as drawn.
(https://www.eevblog.com/forum/testgear/new-spd1168x-siglent-psu/?action=dlattach;attach=478745;image)
Ripple check.
Load as above.
(https://www.eevblog.com/forum/testgear/new-spd1168x-siglent-psu/?action=dlattach;attach=479360)
Another look just now at the same ^ linear PSU SPD1168X that has a ripple and noise spec of 350uV max.
Real electrically noisy SMPS environment although long leads to a 13V 2.1A bulb load didn't exactly help yet with AC line triggering and the wee ripple measured with cursors however proper BNC cabling to the banana jacks would give a totally different result.
AC coupling of course and 1x probe connection.
SDS2104X Plus, 8 bit mode, 20 MHz BW limit. 500uV/div. DC gain accuracy ≤ 3.0%. Cursors and measurements engaged.
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1062200)
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1062204)
-
There is tradeoff between wide bandwidth vertical amplifiers and low noise so even with bandwidth limiting, a wide bandwidth oscilloscope will have higher noise, and this is why you find more sensitive volts/division settings on lower bandwidth oscilloscopes. This suggests looking for an oscilloscope with the minimum sufficient bandwidth.
I think the best option is to design and build a simple low noise preamplifier that has suitable bandwidth, but if you want to buy something, then an old Tektronix AM502 will work with any oscilloscope to provide a differential input down to 10 microvolts/division at up to 1 MHz.
Read Linear Technology application note 70 and application note 124 for some ideas including a list of suitable preamplifiers.
-
There is also the LNA10 from alpha labs that does 10/100/1000 switchable amplification and selectable filter up to 1MHz from about 1Hz. The price is reasonable too.
-
Really these measurements are not hard when a little care is taken.
SPD1168X PSU, 13V 2.1A load
SDS2104X Plus, 10 bit mode, 1x probe, 20 MHz BW limit 500uV/div ch4 and Math averaging = 64.
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1062712)
-
Most of the times this is done with low noise amplifiers that bring the signal into clear emsurement range of modern digital scopes.
Now, a scope useful for that could be the Nicolette sigma 30, or a tektronix 7000 series with 7a22 plugin. Both of these only found on ebay and similars
Fully agree.
A Tek 7A22 in a 7000 series mainframe is a great tool for tracking down sub mV low-frequency noise (< 1MHz). It also has selectable LP and HP filters that allow you to further zoom into signals of interest.
-
I agree the old Tek 7000 mainframe with a 7A22 plug in is the best way to go. I've used this often in the past and a great tool for sub-millivolt work :-+
However, if one doesn't have this Tek equipment available but maybe a newer type DSO like tautech has shown, sub-millivolt work is possible.
Out of curiosity I just gave this a try with a SDS2102X Plus and used a SDG2042X AWG to create a DC 5 volt output on channel 1 and a 4.999V DC output with a 1mv peak to peak sinewave on channel 2. The SDS2102X Plus was set to 500uv/div, 20MHz BW, AC coupled and the Math function used for waveform averaging.
Sure this isn't as good as the old Tek solution could produce, but maybe just good enough for the intended purpose.
The more I fool around with this new to me DSO, the more impressed I am.
Very capable and versatile instruments indeed :)
Best,
Edit: Added something created with an AWG that is 4.999VDC with a 1mvpp ringing waveform to represent a typical switch-mode type PS output.
-
A Tek 7A22 in a 7000 series mainframe is a great tool for tracking down sub mV low-frequency noise (< 1MHz). It also has selectable LP and HP filters that allow you to further zoom into signals of interest.
The AM502 that I mentioned earlier is essentially a 7A22 with a 50 ohm output to drive any oscilloscope, although 7000 mainframes have a vertical output which can be used to allow any plug-in to drive an external instrument and I often do this so that I can use a 100 MHz 7A13 differential comparator with one of my DSOs. The 7A13 unfortunately is not suitable for low level measurements because of high noise from its differential and cascode input stage, although it is still quieter than many modern DSOs.
I agree the old Tek 7000 mainframe with a 7A22 plug in is the best way to go. I've used this often in the past and a great tool for sub-millivolt work :-+
By the time I managed to get an AM502, I had two working 7A22s which came with lot purchases of plug-ins. The 7A22, and of course 5A22 for the 5000 series mainframes, are incredibly useful if you can settle for a maximum bandwidth of 1 MHz.
-
By the time I managed to get an AM502, I had two working 7A22s which came with lot purchases of plug-ins. The 7A22, and of course 5A22 for the 5000 series mainframes, are incredibly useful if you can settle for a maximum bandwidth of 1 MHz.
I recall using the 7A22 to help measure the noise density at low frequencies. Something about assuming a Gaussian distribution and filling a large part of the scope screen with the DUT noise waveform at low sweep rates. Then setting the BW such that you could estimate the number of vertical graticules (uv/div) and divide by the root of the BW. Recall the peak to peak noise being 6.6 times the RMS. There was always a way it seemed to get those old Tek and HP equipment to do all sorts of measurements, since they were designed by hands on engineers for engineers :-+
Best,
-
SDS2104X Plus, 10 bit mode, 1x probe, 20 MHz BW limit 500uV/div ch4 and Math averaging = 64.
Sorry, averaging does limit the bandwidth (depending on sampling rate).
So what bandwidth do we have really here?
And what is 10 Bit mode: 10 bit ADC or oversampling with a 8 bit ADC which further reduces bandwidth.
I prefer a bandwidth limited low noise amplifier like in AN83 of LT.
https://www.analog.com/media/en/technical-documentation/application-notes/an83f.pdf (https://www.analog.com/media/en/technical-documentation/application-notes/an83f.pdf)
see also here:
https://www.eevblog.com/forum/metrology/lt-an-83-board-(lt1028-low-noise-amp-10hz-100khz)/msg1264284/#msg1264284 (https://www.eevblog.com/forum/metrology/lt-an-83-board-(lt1028-low-noise-amp-10hz-100khz)/msg1264284/#msg1264284)
note that AN83 contains a bug: the 5k62 Resistor should be 5K19 (a 68K in parallel to 5K62) to get a flat frequency response.
with best regards
Andreas
-
SDS2104X Plus, 10 bit mode, 1x probe, 20 MHz BW limit 500uV/div ch4 and Math averaging = 64.
Sorry, averaging does limit the bandwidth (depending on sampling rate).
So what bandwidth do we have really here?
And what is 10 Bit mode: 10 bit ADC or oversampling with a 8 bit ADC which further reduces bandwidth.
The 2 GSa/s SDS2000X Plus models 10 bit mode max BW is limited to 100 MHz in all models plus the 20 MHz BW limit was engaged too.
20 MHz BW is adequate for the assessment of a linear PSU.
10 bit mode in these scopes is explained here:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg2783928/#msg2783928 (https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg2783928/#msg2783928)
-
Lets see how the little 200 MHz 2ch 1 GSa/s SDS1202X-E fares measuring low ripple linear PSU's.
Same as previous setup, SPD1168X 13V 2.1A load. Sub 350uV ripple spec.
Averages, BW limit, AC line trigger.....etc.
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1062856)
-
We specialize in ultra low noise power supplies (under 20uv) suggestions would be an early Lecroy HDO4000 12 bit scope 200Mhz have seen them for around $2400 seriously low noise floor plus repeatable accuracy, the 1Ghz HDO600A has a sub 20uv noise floor (20Mhz BWL) The 4000 is very close to this as well. They are slightly lower than the lab Wavepro 254 which is around 30uv with filters applied.
Quality spectrum analyzer preferable real time, both Rigol and Siglent do some very respectable offerings that will be benefical.
A quality DMM would also be very useful Keithley 6500 or with a similar nv/uv range will help to.
-
I recall using the 7A22 to help measure the noise density at low frequencies. Something about assuming a Gaussian distribution and filling a large part of the scope screen with the DUT noise waveform at low sweep rates. Then setting the BW such that you could estimate the number of vertical graticules (uv/div) and divide by the root of the BW. Recall the peak to peak noise being 6.6 times the RMS. There was always a way it seemed to get those old Tek and HP equipment to do all sorts of measurements, since they were designed by hands on engineers for engineers :-+
I have done the same thing but I used the tangential method of noise measurement which can deliver better than 5% uncertainty. The largest contributor of error is actually the passband response of the oscilloscope itself:
http://w140.com/tekwiki/images/c/c8/Measuring_conventional_oscilloscope_noise_garuts_samuel.pdf (http://w140.com/tekwiki/images/c/c8/Measuring_conventional_oscilloscope_noise_garuts_samuel.pdf)
https://youtu.be/5Rk8I5BT2KU (https://youtu.be/5Rk8I5BT2KU)
Today the best way should be to use a DSO's FFT function but no DSO that I know of supports the noise marker function like a good spectrum analyzer would so at best, you have to figure out the correction factors yourself.
-
Sorry, averaging does limit the bandwidth (depending on sampling rate).
So what bandwidth do we have really here?
And what is 10 Bit mode: 10 bit ADC or oversampling with a 8 bit ADC which further reduces bandwidth.
It's fake 10-bits, averaging, etc.
But hey, it's a Siglent! As recommended by the resident Siglent salesman!
I prefer a bandwidth limited low noise amplifier
Yep. An amplifier takes all the guesswork out of it (eg. how much noise is from the scope?), makes it work any any 'scope.
-
Sorry, averaging does limit the bandwidth (depending on sampling rate).
So what bandwidth do we have really here?
And what is 10 Bit mode: 10 bit ADC or oversampling with a 8 bit ADC which further reduces bandwidth.
It's fake 10-bits, averaging, etc.
But hey, it's a Siglent! As recommended by the resident Siglent salesman!
I prefer a bandwidth limited low noise amplifier
Yep. An amplifier takes all the guesswork out of it (eg. how much noise is from the scope?), makes it work any any 'scope.
::) :-//
Where have you been these last 6 years ?
Maybe a brush up on DSO fundamentals is required.
https://www.eevblog.com/2014/04/10/eevblog-601-why-digital-oscilloscopes-appear-noisy/ (https://www.eevblog.com/2014/04/10/eevblog-601-why-digital-oscilloscopes-appear-noisy/)
https://www.eevblog.com/2014/04/27/eevblog-610-why-digital-scopes-appear-noisy-part-2/ (https://www.eevblog.com/2014/04/27/eevblog-610-why-digital-scopes-appear-noisy-part-2/)
-
Another description of a ultra low noise amp with limited bandwidth:
http://dg4rbf.lima-city.de/Ultra%20Low%20Noise%20AMP%20V1.7f%20_WebSeite.pdf (http://dg4rbf.lima-city.de/Ultra%20Low%20Noise%20AMP%20V1.7f%20_WebSeite.pdf)
and the usage for measuring the noise of certain linear regulators:
http://dg4rbf.lima-city.de/Rauschmessungen%20am%20LM723.pdf (http://dg4rbf.lima-city.de/Rauschmessungen%20am%20LM723.pdf)
He feeds the output of the amp to a soundcard and uses a program for analysing the noise. Maybe this is another option for ur request.
sri .. scripts are in german language ..
cheers,
Ronald
-
Where have you been these last 6 years ?
Right here.
Maybe a brush up on DSO fundamentals is required.
https://www.eevblog.com/2014/04/10/eevblog-601-why-digital-oscilloscopes-appear-noisy/ (https://www.eevblog.com/2014/04/10/eevblog-601-why-digital-oscilloscopes-appear-noisy/)
https://www.eevblog.com/2014/04/27/eevblog-610-why-digital-scopes-appear-noisy-part-2/ (https://www.eevblog.com/2014/04/27/eevblog-610-why-digital-scopes-appear-noisy-part-2/)
Um, yes. That's why it's a good idea to amplify the signal before sending it to the 'scope - increase the signal to noise ratio.
-
For bandwidth of 3 MHz there is also 16bit picoscope 4262.. It has 8uV RMS noise at 2mV/div.
With a 100x amplifier in front it can see pretty much anything. Problem is amp noise, if anything.
Without the amp you can reliably measure 50 uV and up. Without amp it is pretty much enough for most uses.
-
Good Call 2N3055,
I have one of these scopes they are for the outlay rather good and have that low noise floor, just keep your lap top off the main lab power circuit when using it.
-
Good Call 2N3055,
I have one of these scopes they are for the outlay rather good and have that low noise floor, just keep your lap top off the main lab power circuit when using it.
I also have Alldaq USB3 isolator, so scope has isolated ground from PC. Isolator comes with nice medical grade PSU, so it's cleaner than what usually comes from laptop, or worse, from desktop PC.
It shaves off few spurs down in the grass. Also, by experimenting, I found one cable ferrite that, with exactly two loops of USB cable trough it, shaves off few dB more on some peaks....
-
I have made something similar with a fully linear psu (we used to make these for audio when the usb computer craze strated) only instead of using Never connected technology these days we use our own propitiatory ultra quiet supply.
As you say does reduce the noise a few points to make it worth while.
Thinking outside the box again 2N3055 good man 8)
-
I also have Alldaq USB3 isolator,
I found one cable ferrite that,
Hello,
which one of the isolators exactly?
that with the 5V adapter: ADQ-USB-ISO-PS
or that with the external 10-30V input: ADQ-USB-ISO-W
Can you also give a pointer to the cable ferrite?
Thanks in advance
Andreas
-
I also have Alldaq USB3 isolator,
I found one cable ferrite that,
Hello,
which one of the isolators exactly?
that with the 5V adapter: ADQ-USB-ISO-PS
or that with the external 10-30V input: ADQ-USB-ISO-W
Can you also give a pointer to the cable ferrite?
Thanks in advance
Andreas
Andreas,
Not a problem. It is ADQ-USB 3.0-ISO. I got that one so I can also use it with 3406D USB3.
Ferrite is an old one I found in a box, it came with some computer equipment god knows how many years ago...
Regards,
Sinisa
-
Good Call 2N3055,
I have one of these scopes they are for the outlay rather good and have that low noise floor, just keep your lap top off the main lab power circuit when using it.
Hello,
one can improve the noise through addition the same signal.
Best regards
egonotto
-
Good Call 2N3055,
I have one of these scopes they are for the outlay rather good and have that low noise floor, just keep your lap top off the main lab power circuit when using it.
Hello,
one can improve the noise through addition the same signal.
Best regards
egonotto
Yes, I see you used well that whitepaper... :-+
-
Yes well even probing techniques matter down in the uV never mind any environmental or scope internal noise.
This screenshot was done with a 1x probe where even LED lighting SMPS would badly impact measurements as the probes reference lead acts as a fine RF loop.......so done in the dark !
12V SLA battery, so pure DC just to see what the bottom level of noise/ripple measurements might be with a SDS2104X Plus.
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1064028)
-
Here is a curious oscilloscope that can see small signals my other scopes miss. It is an HP 1200B, dual-channel, 500 kHz bandwidth relic from the 1970s with Volts/div settings down to 100 uV. Shown in the picture is a 100 kHz sine wave at ~25 uVrms from an RF signal generator. The 1200B has differential inputs plus all the usual dual-channel modes (alt, chop, etc.).
Obviously, the very limited bandwidth makes it a dinosaur for modern electronics. I guess HP was targeting the "audio crowd" back then. The spec says the noise is "less than 20 uVrms measured tangentially at full bandwidth," whatever tangentially means.
I picked it up in broken (no horizontal sweep) condition for $40 a few days ago. The fix involved hooking up a dangling power supply wire and replacing a defective vertical amplifier 2N3440 transistor.
-
To MasterTech:
Thanks for the reference to the paper explaining a tangential noise measurement. I learned something new today!
I did the measurement on the HP 1200B scope and got 9.7 uV which is fine given HP's "less than 20 uV" spec.
-
To MasterTech:
Thanks for the reference to the paper explaining a tangential noise measurement. I learned something new today!
I did the measurement on the HP 1200B scope and got 9.7 uV which is fine given HP's "less than 20 uV" spec.
The 1 MHz Tektronix 7A22 is suppose to be 16 microvolts rms so comparable. That may seem low but the differential cascode inputs that these instruments use extract a severe penalty in noise performance; a 500 kHz or 1 MHz single ended non-cascode input would have much lower noise.
My benchmark though is my 75 MHz 7A18 which I measured at 28 microvolts RMS which is smaller than its displayed trace at its maximum sensitivity and why I laugh at modern 100 MHz oscilloscopes which claim low noise.
-
...I laugh at modern 100 MHz oscilloscopes which claim low noise.
That's just silly.
-
I like Tektronix HiRes Mode for this kind of measurements.
This measurement shows the noise of a XC6201P502 LDO.
Tektronix TDS744A (8-Bit Scope), direct connected with a coax cable, AC-coupling, 20 MHz BW limit:
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1065840)
and when unpowered (this should be the noise floor of scope, probe and environment):
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1065844)
In HiRes Mode the signal is lowpass filtered with 0.44 x Sample Rate. This can be an advantage because there will not be any aliasing artifacts and you can adjust the BW by choosing an appropriate Sampling Rate.
I have no chance to measure this noise with a TDS3032 9-Bit Scope:
(https://www.eevblog.com/forum/testgear/a-good-scope-for-measuring-low-levels-of-noise-(e-g-linear-ps-noise)/?action=dlattach;attach=1065848)
-
If I need to go lower than 20 or so mV, I will normally use some form of external amplifier. I have a few different ones I use for RF and this lower one for DC-100KHz or so. It has a 10,100 and 1000X gain select and AC or DC coupling.
Shown for example is a 100mVp-p square attenuated by 80dB. Amplifier is set to the 100X gain. The scope is shown in full BW, no filters or averaging. Then with 20MHz BW, 10 sweep average.