HP 54111D DSO

[r010159]

Hello all!

I have purchased a HP/Agilent 54111D. I need some help to understand the specs and how they compare to recent scopes. I apologize in advance for my ignorance and this may be the wrong forum to post. Please point me to the right forum. But the expertise available here is truly impressive.

Let me ask some basic questions on DSO specs. I see oscilloscopes have come a long ways offering performance and features at a fraction of the price of 25 year old equipment. But I suspect the performance specs of this 54111D scope still does have merit when compared to other newly available DSOs.

I am finding Agilent and Tektronics and BKPrecision scopes to offer 500 MHz capability, but at a sampling rate much higher than the 1Gs/s of the 54111D. I find 200 Mhz scopes sampling at 2Gs/s and some even higher. Why the need for these much higher sampling rates to accomplish the 200 Mhz bandwidth? Also there is a difference between repetitive and sing-shot bandwidth. When a scope is specified to have 200 MHz, is this single shot or repetitive bandwidth specification?

The effective resolution of my scope is 6-bits to 8-bits. This is frequency dependent. There are scopes that provide a claimed nominal resolution of 8-bits and even a higher effective resolution, perhaps up to 12-bits. This effective resolution spec must also be frequency dependent, correct?

How does the performance specs of my scope basically compare to that of new scopes on the market? I know on features alone the new scopes greatly surpasses my scope. What are some of these features?

I paid $350 for my huge "boat anchor" of a scope.

Any help would be appreciated.

tucson

Scope specs:

2-channel (w/2 extra trigger channels)
500 Mhz bandwidth repetitive, 250 Mhz single-shot (500 MHz with special probe)
6-bit vertical resolution (8-bit effective at 50 MHz)
1Gs/s sampling rate (2Gs/s with special probe)
Can specify trigger conditions including logical and math functions
Color screen

 

[Samuel Groner]

It is customary nowadays to use a somewhat higher sampling rate than the Nyquist rate, to reduce aliasing artefacts and/or to avoid the overshoot introduced by steep anti-aliasing filters.

Something to check is repetition rate--older DSOs are typically rather poor with this respect, and only the latest generation is coming close to/exceeds the old analog scopes.

Samuel

 

[r010159]

I think I know the reason for oversampling. It is to handle the noise introduced by the oscilloscope. This probably can come from several sources including jitter. The downside is a less accurate representation of the signal since the oversampling involves averaging. Good scopes introduce less noise in the processing of the signal which means less oversampling required which means a more accurate reproduction of the signal.

What do you think?

 

[PedroDaGr8]

I think I know the reason for oversampling. It is to handle the noise introduced by the oscilloscope. This probably can come from several sources including jitter. The downside is a less accurate representation of the signal since the oversampling involves averaging. Good scopes introduce less noise in the processing of the signal which means less oversampling required which means a more accurate reproduction of the signal.

What do you think?

The reason for the high sample rate (not to be confused with oversampling) is as follows.

To quote wikipedia:

For a digital oscilloscope, a rule of thumb is that the continuous sampling rate should be ten times the highest frequency desired to resolve; for example a 20 megasample/second rate would be applicable for measuring signals up to about 2 megahertz. This allows the anti-aliasing filter to be designed with a 3 dB down point of 2 MHz and an effective cutoff at 10 MHz (the Nyquist frequency), avoiding the artifacts of a very steep ("brick-wall") filter.

So this high sample rate is a requirement of the physics/mathematics of signal processing in the digital space.

The reason this isn't oversampling is that the samples are not combined. Each sample exists as its own discrete data point.

On the other hand for repetitive signals, you can keep reacquiring the same point and then oversample the results to get a higher number of bits of resolution for each point.


Jitter is not an issue with oscilloscopes, pretty much ever. Its pretty much a bug-a-boo of audio people where it has achieved almost mythical status. I'll leave it at that. Noisy input stages on the other hand CAN BE an issue, especially in cheaper oscilloscopes. They raise the noise floor making small measurements much more difficult. Most scopes produced though have an acceptable noise floor that falls within +/- 1LSB, even the cheap ones.


Now for your device and how it compares to modern scopes:

The biggest drawback is the 6-bit input. That REALLY hampers your vertical resolution, you will likely see significant quantization of the input in single-shot measurements. As you have 64 units of resolution on the vertical scale. At this time, most systems use 8-bit input stages with high-end (read expensive) scopes achieving 12-bit and higher. 8-bit is acceptable for most uses but increasing the bit number is one area where scopes will head in the next few decades.

Additionally, Digital Phosphor type functionality(also called persistence/intensity grading) which has become common place on many of the newer oscilloscopes such as the Rigol's REALLY helps bridge the gap between analog and digital oscilloscopes. This is something your scope lacks and it gives an analog-esque feel to the digital display allowing you to more easily see glitches, fine detail and the like. In the following image you can get an idea of what i mean. The intensity gradations are something which on older digital scopes would be completely missing.

An externally hosted image should be here but it was not working when we last tested it.

Lastly, the thing Samuel is referring to is commonly called blind time. It's inherent in all digital scopes and can be very significant (its not unheard of to have blind times of 99.9%). Blind time is negated by waveform update rate, I can't even find a waveform update rate for your scope. Basically, its how many waveforms can the scope capture per second (not how many samples). This pdf here explains blind time very well. It's a complex topic that is a bit hard to wrap your head around. Blind time can also be negated a bit by the super long memories found on modern scopes. Your scope has 8Ksample memory, which actually isn't bad for its age. It's not uncommon for modern scopes to have MSample memory (with Rigol having up to 56Msample memory).

The biggest benefit your scope has is the high frequency input stages. If you need the high bandwidth for repetitive signals then your scope does have some merit there.

 

[r010159]

Wow! You are good! Thankyou very much for your description of issues involved with high sample rates, and also for your analysis of my scope. It looks like the money could of been put into a scope like Rigol.

Well, I know someone who has a boat that can use another anchor.

Just a question. Isn't a sampling rate of 4 times the bandwidth adequate for a smooth slope anti-aliasing filter? This is the case on my scope. Perhaps sampling rate is being used as a "gee whiz" performance parameter.

 

[PedroDaGr8]

Wow! You are good! Thankyou very much for your description of issues involved with high sample rates, and also for your analysis of my scope. It looks like the money could of been put into a scope like Rigol.

Well, I know someone who has a boat that can use another anchor.

Just a question. Isn't a sampling rate of 4 times the bandwidth adequate for a smooth slope anti-aliasing filter? This is the case on my scope. Perhaps sampling rate is being used as a "gee whiz" performance parameter.

Don't get me wrong, the fact is it is a 4-channel, high sample rate, high frequency scope. There ARE uses for it. Just for the general hobbyist its not the best value. There are situations where it has features of value that others do not. For general value, going up in price it goes: Rigol DS1052E->Rigol DS1074Z-->Rigol DS2072A-->used agilent 3000 series or Rigol 4000 series. If I were to buy a scope right now. I'd likely go for the DS1074Z as its a huge value for the price.

As for the 4x sample rate, this is getting to the edges of what I know so I might be a bit wrong or over-simplified. For a pure sine wave, a 4x sample rate is acceptable. The issue comes into play when dealing with non-sine wave forms i.e. square waves in particular. With a 10x sampling rate most square waves will look like a reasonably facsimile of a square wave. Due to a rolling off of the highest frequency components there will be a bit of rounding at the edge but nothing major. The aliasing artifacts are low enough to not dominate the signal. At a 2x rate the aliasing is significant. You have large amounts of over-shoot, pre-shoot, and non-smoothness of the rising edge. A 4x sample rate cuts this dramatically, while an 8x and above cuts it even more. To be honest the differences between a 4x and 8x sample rate are much less than the differences between a 2x and 4x. 2x is virtually unusable. So once again a 10x sample rate is not ESSENTIAL but it gives a marked improvement over 4x in certain scenarios.

 

[r010159]

I found this in an article by Joel Woodward from Agilent titled "How to get more then 8-bits from your 8-bit scope"

[What follows is an excerpt from that article]

If a signal is sampled at a rate much higher than the Nyquist frequency, and then is digitally filtered to limit it to the scope bandwidth, the scope magically gets additional resolution. The 8-bit ADC combined with the digital filter can be made to produce more than the 256 Q levels associated with an 8-bit ADC. Using this technique, the scope can obtain an effective resolution that is more precise than what the ADC would naturally produce. In fact in the example described above an 8-bit ADC can be made to act like a 12-bit ADC given the excess sample rate relative to bandwidth.

Most scope vendors include a setting that allows the scope to oversample and digitally filter the output of the ADC to achieve more bits of resolution. Some vendors call this high-res mode, while other vendors use different names. All major vendors include this capability in their scopes. For scopes with sample rate equal or near to the 2.5 times ratio relative to scope front-end bandwidth, turning on high res mode will cut the overall bandwidth while providing more than 8 bits of resolution. For scopes with a sample rate much greater than the 2.5 times ratio relative to bandwidth, turning on high-res mode will enable greater resolution with no tradeoff in overall scope bandwidth.

So the extra bandwidth provided by a much higher than necessary sample rate can be used for oversampling purposes, thereby providing an effective bandwidth that is more than 8-bits. The excessive bandwidth is not wasted after all. This capability is user adjustable and can be found on many scopes.

Now this makes sense to me. What do you think?

 

[PedroDaGr8]

I found this in an article by Joel Woodward from Agilent titled "How to get more then 8-bits from your 8-bit scope"

[What follows is an excerpt from that article]

If a signal is sampled at a rate much higher than the Nyquist frequency, and then is digitally filtered to limit it to the scope bandwidth, the scope magically gets additional resolution. The 8-bit ADC combined with the digital filter can be made to produce more than the 256 Q levels associated with an 8-bit ADC. Using this technique, the scope can obtain an effective resolution that is more precise than what the ADC would naturally produce. In fact in the example described above an 8-bit ADC can be made to act like a 12-bit ADC given the excess sample rate relative to bandwidth.

Most scope vendors include a setting that allows the scope to oversample and digitally filter the output of the ADC to achieve more bits of resolution. Some vendors call this high-res mode, while other vendors use different names. All major vendors include this capability in their scopes. For scopes with sample rate equal or near to the 2.5 times ratio relative to scope front-end bandwidth, turning on high res mode will cut the overall bandwidth while providing more than 8 bits of resolution. For scopes with a sample rate much greater than the 2.5 times ratio relative to bandwidth, turning on high-res mode will enable greater resolution with no tradeoff in overall scope bandwidth.

So the extra bandwidth provided by a much higher than necessary sample rate can be used for oversampling purposes, thereby providing an effective bandwidth that is more than 8-bits. The excessive bandwidth is not wasted after all. This capability is user adjustable and can be found on many scopes.

Now this makes sense to me. What do you think?

By the way, I wanted to thank you for pushing deep into this topic. I've actually learned a lot more of the why behind what goes on.

First, one thing to point out. Resolution, which is measured in bits, is not at all equal to bandwidth. Bandwidth is frequency (similar to slew rate in an opamp), resolution is how many vertical steps can a signal be broken down into.

I think one thing needs to be mentioned. The bandwidth of an oscilloscope is NOT a brick-wall. The standard definition of oscilloscope bandwidth is the point at which the response of a scope is reduced by 3dB is called the bandwidth.

As this image illustrates, it doesn't mean you don't see frequencies well above the bandwidth. Ideally, all of these frequencies need to be under the Nyquist limit as well. Otherwise aliasing occurs which will result in distortion of the observed trace.

To get an idea of how this affects a signal. Here are a series of runs captured of a on a scope where Nyquist frequency equals the bandwidth. As you can see they are heavily distorted with lots of overshoot, preshoot, post shoot, etc. You can even see stuff that would make you think that the signal is ringing. This is clearly not a good thing if you are dealing with high speed circuits.

Here we have a 100Mhz square wave on a 500Mhz scope with the sample rate set to 4x the scopes Bandwidth (2GSa/sec):

I have highlighted the two distortion products with colored boxes. They are the slight deviation from horizontal. The large peak on the leading edge is likely caused by a probe compensation issues and is not a distortion product at all. As you can see most of the distortion is gone. This is at the sample rate:BW ratio your scope is capable of. That being said this is on a 100MHz SW not a 500MHz SW. Some of these issues would be exacerbated a bit on a 500MHz SW (which would likely not look as SW like due to attenuation of some of the higher nth order frequency components.

So what happens if we increase the sample rate to 8x on the same square wave, the distortion products should basically disappear. Sure enough they do, all that is left is the probe calibration issue.

So to achieve a truely accurate signal 8x and above is essential, that being said, if all you need is an almost there accuracy then 4x is nothing to slouch at. Based on these images, clearly its capable of at least 95% of what the 8x is.

Note: most of these images came from the Agilent document: Evaluating Oscilloscope Sample Rates vs. Sampling Fidelity

These discussions do not apply though when discussing scopes of around one Ghz and higher due to a change in how they are constructed. One Ghz and above frequency scopes typically have much steeper frequncy roll off on the bandwidth than do the lower frequency scopes. As a result of this steeper roll you don't need as high of a sampling rate to bandwidth ratio The reasons for this steeper roll off are varied (cost, technology, etc.) One side-effect though is that square waves will look less square wave like at the scope BW frequency. If the cut-off is steep enough, they will look like sine waves (there are zero higher-order frequencies).

So on some of these scopes, because of the higher roll-offs you have sampling way above the last available frequency. This is wasted sampling because there is nothing to alias up there. You can use this extra rate to preform a sort of real-time oversampling. This is applicable because of the steeper bandwidth curve and this has to be explicitly programmed in to the scope itself. Another option, is if there is a programable bandwidth limit. You can perform something similar (once again if your scope has the ability programmed into it). Similarly, if you have a repeating waveform, you can recapture the same waveform over and over again. Using that oversampling to impart the extra resolution.

On a related note, some scopes will actually stagger successive aquisitions so that the points don't directly overlap giving you a psuedo-higher sample rate. This is commonly what is referred to as Effective Sample Rate. This is much harder to do than it sounds as it requires VERY precise timing, very few scopes do this flawlessly. Though a lot of the cheap scopes claim they do. Additionally, the waveform must be repeating and without glitches. Otherwise, its more or less useless as the glitches and non-identical nature show up as noise.

 

[r010159]

You continue to amaze me.

Thank you very much for your attention. It looks like 8X sampling is relevant for high accuracy. You have convinced me. And that is not easy to do. I will look closely at the document you have referenced, besides what you have presented here.

Bob Graham