Question about DSO

1

I recently acquired a DSO (Rigol DS1074Z). I am starting to learn to use it and off the bat I encountered a problem of excessive noise.

I attached two resistors (3m3 and 3k3) to a Li-Ion battery (fully charged showing 4.1V) and measured the voltage across the small resistor. I raised the sensitivity to 1mV and could see that there are a number of frequencies present. Here are some captures.

At the highest rage there is a 100MHz signal:

Then on top of this I see a 4MHz signal

And a beat at 800KHz

I had trouble triggering on the signal, and this becase clear when I changed to 10mV resolution and could see 33KHz spikes present

Moving to the other side of the house (away from the computers) did not change the traces.

And as an aside, the four channels show very different traces. Here are all four attached to the same circuit and set up identically

Clearly, ch1 is cleanest, then ch3 and ch4, then ch2 is the worst. This is at 5mV/div

So, am I doing it wrong (probably)? Is the noise source the environment or the DSO (internal)?

If anyone has experience with these things and plans to be at the space then I can bring this over to get some help.

[later] Changing the battery to a NiMH did not change it, and on a whim I shorted all the probes (attaching the ground alligator to the tip) and got the same signal, with the probes still showing different levels despite all set to 5mV.

BTW, all measurements done with X1 setting on the probes.

I am now even more convinced that I am doing something wrong.

TIA
Eyal

2

What you’re seeing is typical of a scope hooked up in this fashion. You’re talking about noise levels in the order of 10mVpp with a 100MHz bandwidth, that’s really not excessive.

Most of the noise typically comes in via conducted EMI from other switchmode power supplies plugged into the house.

When you get down to these sorts of signal levels on a scope you typically have to use averaging to pull the signal out of the noise. To do that effectively you need some sort of time correlated trigger, that is correlated with the signal in question and uncorrelated with the noise. Then in theory (and it actually works amazingly well in practice) any background noise averages to zero.

I can’t really give you any advice on what you’re doing though, because I don’t understand what you’re trying to do with the scope. Scopes are all about measuring time dependant signals, a DMM will do a much better job of measuring a battery voltage.

I’ll be in the space tomorrow night if you’re around.

3

I want to learn to generally use it. I will want to measure mV level signals so it will be beneficial to find a way to do so.

My current use case is measuring the power usage of an esp8266 setup. The currents involved are around the range of 100uv and 100mv. I intend to measure the low range through a 100x amp and the high range directly. The scope has the functionality required (integrating the area under the trace).

As for the averaging method: it may be suitable, but the event will trigger once every 4-5 seconds and will probably not by 100% identical each time. However for establishing the power usage I expect that the high frequency AC noise is already cancelling itself when integrated.

I will try to be there Wed as well.

4

For what it’s worth I just snapped this on my ancient Tektronix TDS540. It’s a 4-ch 500MHz DSO from the '90s. The probes are shorted and showing only ch1.

I’ve set it to 400ns and 1mV per div in the pic. The cursors are set to as close as 1.25us as I could get so that you could compare the scale with your own pics. At a full 1GS/sec (50ns/div) and 10 pass envelope it seems the noise I have is almost 2mVpp. On “average16” it looks like 1/10 of that.

Perhaps compare and play with the DSO we have downstairs

5

It may just be that the Rigol is a bit noisy or susceptible to EMI. I have a lot of experience with a Rigol spectrum analyzer and while it lacks a lot of the nicer/convenient measurement features and its noise floor isn’t exactly brilliant, it’s also barely 1/5 the cost of even the lowest-spec’d equivalent Keysight/Agilent/R&S instrument - it’d be an obvious choice for a lot of different work at twice the price.

I’d try your DSO at a different location just in case there’s something curious about the current location, compare with the DSO downstairs at MHV side-by-side

6

As mentioned, I tried elsewhere in the house to no avail. I will try at the space next.

I have a small inverter so I could try in outside (using the car) assuming the car (and inverter itself) are quieter.

As for what I am trying to do now, here is a screenshot of an esp8266 doing its thing in 4.8s

as the “Area” (bottom left) shows, the active phase used 402mVs. However, each cycle is different, for example this shorter (4.66s)one

So averaging will probably not work.

<OT>
Unlike the data I collect with the PSoC, having 12Mpts at 2MSa/s I can zoom in, like this initial wireless action as it comes out of deep sleep

and I can see how the SDK plays the WiFi hardware when associating with my app. I wonder if the ramp at the end is where power is increased until a stable connection is established.

The long flat middle part is where my app waits for an IP for almost 3 seconds, checking every 100ms.

The idle parts at the end (just before it goes to sleep again) is probably where it waits for ACK after sending packets. I now use UDP but somehow the software (either the SDK or nodeMCU) fails to actually send the packet before calling the “sent” function so I just idle for 50ms before going to sleep.
</OT>

I wonder what I can do to improve the noise. More filtering on the power? Shorter probes? Grounding the channels internally (coupling = GND) does reduce the noise, and any handling of the probe leads (even when they are shorted, externally) makes a huge difference.

cheers

7

I only suggest taking it to the space and comparing side by side with the MHV DSO there because it’s easy to do and helps rule out something local to your house. Perhaps I’m just superstitious, but I’ve wasted days debugging this kind of thing before and if I’d only performed this kind of comparison/isolation test rather than assuming that I surely would not be so unlucky to have an exotic source of interference then I might have arrived at the correct conclusion much quicker :slight_smile:

Another thing to consider is probe quality. Cheap probes make a big difference not only with high frequency attenuation but I’ve found some examples with shamefully high resistance (perhaps defective) shield connection.

Do you have any 50 ohm gear (even a dummy load) to connect into? For that matter, does the noise look different when you configure the channel impedance between hi-z or 50ohms? I’d say that getting the noise to disappear when you are connected to 50ohms or short-circuit is a good place to start.

Edit: I have found a YouTube video of someone with your model Rigol DSO who claimed to solve his noise issue by unplugging his laptop AC adapter.

8

I should clarify that the pic of my scope earlier was with 1x passive probes. See this Agilent doc for some good discussion on probe considerations with low voltage measurements http://www.tek.com/dl/51W_27668_0_MR_Letter.pdf

9

Thanks @csirac2_ ,

I will follow your suggestions and will try the MHV DSO tomorrow.

I do have a number of devices plugged in (phones, computers, radio etc., LED lights?) and all have the potential to cause trouble. I will disconnect these one at a time and see what happens.

I have my old analogue scope (15MHz Trio) and tried those probes (100MHz spec) and saw similar. the same noise.

I may need to use an external, good quality amp (10x or 100x) - I have no idea what this means. This should put me above the DSO internal noise when dealing with mV range signals. Dealing with uV will require even more.

I have some shunt resistors, 0R01 1% that I want to use for measuring current. If I want to use it to watch sub-mA signals (e.g. the esp8266 is supposed to be around 70uA when sleeping, so 0.7uV on the shunt) then I will need more if it is at all feasible.

[later] Looks like an interesting article that you suggest, I will read it.
[even later] Now I know what those small springs that came with the probes are for…

cheers

10

10x and 100x refers to voltage attenuation. The reason this is useful is to extend voltage measuring range and achieve higher bandwidth (smaller signals are not so badly attenuated up the coax due to capacitance), a 1Mohm scope channel impedance may be presented as 10 or 100Mohm load to the circuit under test instead, and expensive “active” probes even have a little high quality opamp in them to improve signal integrity further.

Active probes definitely should try to get them compatible with your model scope. Passive probes are less important in terms of compatibility but the benefit is that your scope will automatically know the attenuation factor and compensate the V scale for you; obviously with generic or incompatible probes you’ll have to multiply by 10x/100x yourself.

I wouldn’t bother with cheap switchable 1x/10x probes, I’d rather a cheap 1x and a cheap 10x - the little switches are noisy

11

Obviously in your case you won’t want any attenuation - but uV scale is approaching radio receiver sensitivity (are you near an AM transmitter station around 800-900KHz?) so you may even want to think about building your own custom properly terminated probe - direct and careful soldering of the shunt resistor straight onto the end of good quality BNC cable with as short hookup leads (or none at all) as possible, the reason is that those ground leads/clips on standard probes act like loop antennas :slight_smile:

Or you could build the shunt into a little op amp jig as you suggested.

But it’d be good to get a cleaner flat line with externally shorted probes first before embarking on these adventures

12

Oh - one more thing that I learnt the hard way - your oscilloscope probe’s ground is likely a path to your mains power ground! This can be very important if the circuit under test also has a path to this same ground, as you can imagine you can create ground loops or just outright short circuits thabks to your probe ground.

So another thing to try is use two separate scope channels configured into one delta/differential trace, then you won’t have to worry about this. Or, try to power your project from a battery.

13

This scope comes with passive probes that have a 1X/10X switch. And yes, it is noisy, often making bad contact.

14

  1. I am aware of the ground issue, and even noticed eevblog covering it. I did an A-B setup test that did not help much. I need to test again though. It is difficult to compare because just touching the leads or moving them around changes the noise :-(

  2. My current project measures battery powered esp8266 so this makes the experiments simpler.

15

Ah, I see. A good probe shouldn’t induce such noise when gently moved (especially while you have it shorted), but touching the probe tip itself of course means you are acting as an antenna so this will change the nature of the background noise you see. However I’ve seen what you describe, noise spiking when physically moving the probe cables around, it may be faulty probes or poor connections or both.

16

I have 4 probes (naturally) and they all exhibit this. Actually, having them lying loose (shorted at the tip) shows different noise on each and moving them about changes this again. Not spikes like a bad connection but general change if the amount of noise that stays at the new level once left alone. As an example, if I have a probe placed (shorted) such that the noise is one division wide, holding the lead (not the tip) easily completely overshoots the screen. Very sensitive.

17

I see. With my own setup here it’s only very slight change in noise when I touch shorted probes. However the probes I have on today are just little cheapies, so perhaps I’m just in a very RF quiet environment here.

Just as a sanity check, do you have any of these lying around? Try connecting directly to one input: http://www.jaycar.com.au/Interconnect/Plugs%2C-Sockets-%26-Adaptors/Computer/BNC-Terminator-Plug/p/PP0655

Then try connecting to this dummy load via http://www.jaycar.com.au/Sight-%26-Sound---Home/AV-Leads-%26-Adaptors/Video/BNC-Plug-to-BNC-Plug-Cable---1-5mt/p/WV7300

Then compare with your current probes clipped onto the dummy load.

If you get a nicer/cleaner flatline with the dummy load then at least we can say the scope itself is good. If you get a nicer/cleaner flatline with the dummy load via BNC cable than with the dummy load via your probes, then we can assume that it’s your probe’s fault (insufficient shielding, GND lead is too big, or high resistance shield/gnd connection).

18

The probes that come with the Rigol aren’t bad, not top of the range but you’ll be spending quite a reasonable amount to get any better.

I can bring some cables, AWG, etc along tonight to do some testing.

Reality is though to be measuring these sorts of signals your probably going to need to do some active amplification before feeding into the scope. To get down to the very low currents over a 100mOhm sense resistor you’re definitely going to need active amplification, likely powered from a battery and enclosed in a grounded metal can.

Measuring single digit uVs using a scope is practically impossible without spending a lot of careful effort and/or money (noting that they’re somewhat interchangeable.) I believe there are ways of doing what you’re after without pushing the realms of possibility.

19

Another way of measuring power consumption over time is to power the circuit from a capacitor, rather than a battery. You can charge the capacitor with a roughly known number of coulombs, then run the circuit to expiration.

This is better used for observing relative power consumption (as per the Atmel Picopower paper), rather than determining absolute or instantaneous consumption.

As for preamps: http://aricorp.com/OSP-1.html

Only a thousand dollars :smiley:

20

Oh Man! This takes me back… I had to dig very deep into my bit bucket and retrieve a handful of BNC terminators and a cable.

Least noise is with internal grounding.
Leaving a channel open, or with a terminator, or with a cable or with cable+terminator all look the same, noisier.

When I connect a probe it is much noisier, even when terminated.

Interestingly, it looks like there is some offset in the readings, some channels show the noise above their position, others below and hardly any are centerd (even when grounded or AC coupled).

Moreover, I can see the noise changing when I move my hands closer to the scope (not touching) if a probe is attached.

And some eye candy to back up the above claims:

Internally grounded:

Terminator installed on scope:

Cable attached (same with added terminator):

Not that bad so far.

Probe attached (1X, shorted):

Probe attached (1X, open):

Need I say more?

I will then… I tested my old probes again and they actually do better. Far less reaction to touching and moving and less noise. Maybe the supplied probes are really crap?

cheers