| Products > Test Equipment |
| How much noise floor and other things matter in oscilloscope usability |
| << < (20/63) > >> |
| Fungus:
--- Quote from: nctnico on December 25, 2021, 09:49:05 pm ---There are so many lower noise alternatives out there from GW Instek, MicSig and (if you have the budget) R&S as well that it makes no sense to buy a noisy Rigol at all. --- End quote --- Except for the Pesky Fact that OP says he mainly does digital stuff and nobody else makes a 4 channel, 350Mhz, 8GSample/sec 'scope for only $1000. To me it seems like you're arguing that a car with leather seats will allow people to carry more shopping. |
| Performa01:
I think it’s time to do some myth-busting – and providing some hard evidence instead of presenting just wild guesses. Myth #1: “Noise is only important when using x1 probes. It is irrelevant when using the much more common x10 probes, because the high source impedance will exceed the noise of the DSO anyway.” Nothing could be further from the truth. The only relevant effect is the attenuation of the probe, which requires an appropriate vertical gain on the DSO to compensate for it. And of course this won’t matter much at the low sensitivities, i.e. the noise will be about the same with a x1 probe at 10 V/div, a x10 probe at 1 V/div and a x100 probe at 100 mV/div. But if you happen to work with signals much lower than 80 Vpp, you will notice that 1 mV/div with a x100 probe will be noisier than 100 mV/div with a x1 probe and that in a scenario like this, a proper low noise frontend will be much more pleasant to work with. You sure don’t want the excessive noise of an e.g. Rigol MSO5000 when working with x100 probes (e.g. in vintage tube gear). The output impedance of a x10 probe is dominated by an output capacitance of about 100 pF, therefore the noise bandwidth is only about 1.6 kHz. So except for very low frequencies <100 kHz, we won’t see any significant difference in a properly designed general purpose DSO frontend, whether the scope input is left open, terminated by 1 M or 50 ohms or shorted to ground. For low frequencies, things are a lot more complex than just a FET buffer, because of the split path design of all contemporary wideband frontend designs. The practical consequence is, that general purpose (wideband) oscilloscopes generally aren’t well suited for low frequency tasks below about 10 kHz regardless of the probes used. There are specialized instruments for this. Look at the first two screenshots attached. They show the noise spectrum up to 1 GHz of the Siglent SDS2354X (570 MHz bandwidth). First with the input left open in high impedance mode, then the input internally terminated by 50 ohms. There are minimal changes of the spurious signals (because of the different contributions of voltage- and current effects), but the noise changes by less than 1 dB within the 570 MHz bandwidth of the scope frontend. SDS2354X Plus_FFT_Noise_1M_ BW570M_8bit SDS2354X Plus_FFT_Noise_50_BW570M_8bit Btw: please notice, that up to 1 GHz there are few spurious signals and no spur is exceeding -120 dBV, which is equivalent to 1 µVrms. This is another important aspect, because strong spurs near the signal frequency can be at least as annoying as excessive noise. Myth #2: “Frontends with higher bandwidths are always noisy, even when bandwidth limited.” In any modern DSO, the bandwidth limit is an integral function of the PGA – and it sits at its output. So all the input noise gets filtered before the ADC. Of course this is only a first order RC-filter, because other than some popular believe, there is also no such thing as an effective AA-filter (Anti Aliazing filter) in a serious DSO. The most important property of any DSO frontend that is not a toy is constant group delay, and this rules out any “effective” AA-filter. But even with a humble first order filter, the effect of noise reduction is quite obvious. Next screenshot shows the noise floor at 50 ohms input termination again, but this time with 200 MHz input bandwidth limit. Btw, Siglent scopes show all relevant information on the screen, so screenshots should be pretty much self-explanatory. SDS2354X Plus_FFT_Noise_50_BW200M_8bit Compare this with the previous screenshot. At 110 MHz, we already have a difference of 0.8 dB. At 340 MHz it is nearly 6 dB and 7 dB at 560 MHz. That is a difference, isn’t it? The next screenshot demonstrates what happens if the common 20 MHz bandwidth limiter is activated. SDS2354X Plus_FFT_Noise_50_BW20M_8bit At 20 MHz, noise is 2.7 dB down, we get -10.9 dB at 110 MHz, -16 dB at 340 MHz and -17.6 dB at 560 MHz. The SDS2000 series has an excellent software enhanced 10 bit mode, which limits the bandwidth to 100 MHz and lowers the noise floor even more. See the next screenshot. SDS2354X Plus_FFT_Noise_50_BW100M_10bit With this setting, the noise floor fell below the -150 dBV above 200 MHz, so the reference level of the spectrum alanysis had to be adjusted accordingly. Little change up to 110 MHz, but -18.7 dB at 340 MHz and -26.8 dB at 560 MHz make this an excellent low noise mode in applications where 100 MHz bandwidth is sufficient. Of course we can use the 20 MHz bandwidth limiter here as well, as the next screenshot demonstrates. SDS2354X Plus_FFT_Noise_50_BW20M_10bit At frequencies above 20 MHz, noise drops dramatically: About -13.5 dB at 110 MHz, -28.8 dB at 340 MHz and -31.6 dB at 560 MHz. So this proves that it has nothing to do with the genuine bandwidth of the frontend or any other signal paths. I can easily demonstrate that e.g. a modern 2 GHz DSO like the SDS6204 behaves no different in this regard. The following screenshot shows a noise plot of the 2 GHz scope that can be compared to the very first screenshot in this posting. SDS6204_FFT_Noise_1M_BW2G_D1G This is in high impedance mode with open input. Once again, the noise with internal 50 ohms termination is very similar. More interesting is the comparison of this 2 GHz scope with the SDS2354X Plus. Even though the high bandwidth scope produces more spurs (but only very few of them slightly exceed 1 µVrms), the noise is comparable or mostly even better than on its 500 MHz counterpart – within the operating bandwidth of the latter, that is. Of course, at 840 MHz the SDS2304X Plus is already in the stopband of the frontend and noise drops significantly, whereas the SDS6000 has not even reached half its bandwidth, so its noise at that frequency has to be in the same ballpark as the other measurements before. What these screenshots also reveal, is that the sample rate does not affect (excessive) frontend noise. If anything, higher sample rate helps to reduce noise. The 5 GSa/s, 2 GHz scope produces less noise than its 2 GSa/s, 500 MHz counterpart. There is the ADC noise itself, which is the granular noise determined solely by the ADC resolution, not the sample rate. A higher sample rate will spread out the noise energy over a wider bandwidth, but the total energy will remain the same. So if only a limited part of that bandwidth is observed (i.e. FFT zoom feature), only a part of the noise is visible, hence will appear lower than the total noise actually is. High sample rates also do not accentuate high frontend noise. If the sample rate is excessive with regard to the bandwidth of the frontend, it will just produce redundant data and pointlessly eat up sample memory with little to no effect on the result. If, on the other hand, the bandwidth of the frontend exceeds half the sample rate, the noise portion above Nyquist will be aliased back into the Nyquist bandwidth, hence we’ll still get all the frontend noise, even with inadequate low sample rates. |
| G0HZU:
--- Quote ---If, on the other hand, the bandwidth of the frontend exceeds half the sample rate, the noise portion above Nyquist will be aliased back into the Nyquist bandwidth, hence we’ll still get all the frontend noise, even with inadequate low sample rates. --- End quote --- Yes, that's what I find with my old HP scope here when the 30MHz bandwidth limit is selected. It still measures the noise level fine even at very low sample rates. It doesn't matter at all because the signal being measured is noise. The same applies if I deliberately feed wideband noise to the scope front end and use a very low sample rate to try and measure the Vrms of the noise. As long as the bandwidth of the external noise signal is less than the 30MHz bandwidth of the scope it should measure the Vrms quite well even at 100kSa/s. |
| bdunham7:
--- Quote from: Performa01 on December 25, 2021, 10:10:33 pm ---Myth #2: “Frontends with higher bandwidths are always noisy, even when bandwidth limited.” --- End quote --- I think the statement, at least the one I'm thinking of, was that the noise density was higher for higher-BW capable amplifiers. The noise will still be a function of the noise density and the actual bandwidth, so limiting BW will still reduce noise as expected. --- Quote ---In any modern DSO, the bandwidth limit is an integral function of the PGA – and it sits at its output. So all the input noise gets filtered before the ADC. Of course this is only a first order RC-filter, because other than some popular believe, there is also no such thing as an effective AA-filter (Anti Aliazing filter) in a serious DSO. The most important property of any DSO frontend that is not a toy is constant group delay, and this rules out any “effective” AA-filter. --- End quote --- I believe that many very-high-BW scopes now have more than a first-order roll off and thus a flatter frequency response but poorer step response. IIRC either Tek or HPAK or both had/have models where you specify which way you want it as a factory option. I don't think this is specifically for anti-aliasing, but it will have that effect. I can't be very specific because stuff like that doesn't get into my hands much--I have to read about it here on EEVBlog. |
| Performa01:
--- Quote from: bdunham7 on December 25, 2021, 10:44:12 pm --- --- Quote from: Performa01 on December 25, 2021, 10:10:33 pm ---Myth #2: “Frontends with higher bandwidths are always noisy, even when bandwidth limited.” --- End quote --- I think the statement, at least the one I'm thinking of, was that the noise density was higher for higher-BW capable amplifiers. The noise will still be a function of the noise density and the actual bandwidth, so limiting BW will still reduce noise as expected. --- End quote --- Well, the FFT plots in my screenshots show nothing but the noise density. Of course, the total noise is actually higher for the 2 GHz instrument at full bandwidth than what it could ever be on the SDS2kX Plus. I strongly suggest that, other than in the seventies of the last century, for modern semiconductors the noise density remains fairly constant over frequency. In my screenshots it can be seen that it gets rather lower at higher frequencies and four times the system bandwidth doesn’t mean higher noise density at all. --- Quote from: bdunham7 on December 25, 2021, 10:44:12 pm --- --- Quote ---In any modern DSO, the bandwidth limit is an integral function of the PGA – and it sits at its output. So all the input noise gets filtered before the ADC. Of course this is only a first order RC-filter, because other than some popular believe, there is also no such thing as an effective AA-filter (Anti Aliazing filter) in a serious DSO. The most important property of any DSO frontend that is not a toy is constant group delay, and this rules out any “effective” AA-filter. --- End quote --- I believe that many very-high-BW scopes now have more than a first-order roll off and thus a flatter frequency response but poorer step response. IIRC either Tek or HPAK or both had/have models where you specify which way you want it as a factory option. I don't think this is specifically for anti-aliasing, but it will have that effect. I can't be very specific because stuff like that doesn't get into my hands much--I have to read about it here on EEVBlog. --- End quote --- Yes, the genuine bandwidth of most scopes is not first order gaussian – unless they are artificially bandwidth limited. This is also why we don’t get ideal pulse response characteristics and vendors have to specify some overshoot. There can be all sorts of filters in high end scopes – mainly for pulse response equalization, but you can have higher order AA-filters as well. But this is limited to filters with Gaussian and Bessel characteristics, where the transition from the passband to the stopband is very smooth, so you still need substantial oversampling in order to get a useful attenuation to fight aliasing. Nothing gained for todays top models within a series, where the bandwidth is not at least five times lower than the sample rate. Furthermore, higher order filters can get tricky with regard to their sensitivity to component tolerances (and their temperature stability) and they might need alignment in the first place. All this increases effort and costs and puts the risk of different frequency/phase response in different DSO channels. This is probably why standard PGAs for “normal off the shelf” instruments have the bandwidth limiter integrated and designers have decided that a first order RC-filter is the best compromise. |
| Navigation |
| Message Index |
| Next page |
| Previous page |