RIGOL MHO98 and MHO900 Oscilloscope Series

[gf]

Martin (and after Iewa)
I'd actually already done some tests to see what changes occurred when working with the 50-ohm input, especially at high frequencies. As far as I remember, only the signal amplitude decreased, but the ringing didn't change at all. But since you've spent time on it, it's worth repeating those graphs at 50 ohms.

That "ringing" looks like Gibbs Effect.

Certainly it is—and presumably, it's primarily coming from the digital interpolation filter of the oversampling DAC (likely a halfband FIR filter with a relatively narrow transition band).

By the way, contrary to popular belief, a discrete-time step (..., 0, 0, 0, 1, 1, 1, ...) does not represent a Heaviside step in the continuous-time domain under ideal Shannon-Nyquist reconstruction. Instead, it results in a sinc-modulated step—described by the Sine Integral function, Si(t)—which inherently exhibits Gibbs overshoot. Consequently, such an output is entirely expected if the AFG sends a discrete-time step to the DAC. If the AFG aims to reduce this overshoot, it must intentionally slow down the digital edges, increasing their rise time.

The catch is, if the AFG relies on plain DDS technology to generate the square wave, the limited wavetable size fails to provide sufficient temporal resolution for the soft edges at lower signal frequencies. This is likely the reason for the observation that the AFG applies the 9 ns soft edges only for frequencies ≥ 1 MHz, while falling back to hard discrete-time steps at lower frequencies. This hypothesis is further supported by the datasheet, which guarantees the specified overshoot only for frequencies ≥ 1 MHz. To get rid of this limitation, technologies like SiFi, EasyPulse, or Trueform are used instead of plain DDS by more sophisticated AWGs.

[mawyatt]

Martin (and after Iewa)
I'd actually already done some tests to see what changes occurred when working with the 50-ohm input, especially at high frequencies. As far as I remember, only the signal amplitude decreased, but the ringing didn't change at all. But since you've spent time on it, it's worth repeating those graphs at 50 ohms.

That "ringing" looks like Gibbs Effect.

Certainly it is—and presumably, it's primarily coming from the digital interpolation filter of the oversampling DAC (likely a halfband FIR filter with a relatively narrow transition band).

By the way, contrary to popular belief, a discrete-time step (..., 0, 0, 0, 1, 1, 1, ...) does not represent a Heaviside step in the continuous-time domain under ideal Shannon-Nyquist reconstruction. Instead, it results in a sinc-modulated step—described by the Sine Integral function, Si(t)—which inherently exhibits Gibbs overshoot. Consequently, such an output is entirely expected if the AFG sends a discrete-time step to the DAC. If the AFG aims to reduce this overshoot, it must intentionally slow down the digital edges, increasing their rise time.

The catch is, if the AFG relies on plain DDS technology to generate the square wave, the limited wavetable size fails to provide sufficient temporal resolution for the soft edges at lower signal frequencies. This is likely the reason for the observation that the AFG applies the 9 ns soft edges only for frequencies ≥ 1 MHz, while falling back to hard discrete-time steps at lower frequencies. This hypothesis is further supported by the datasheet, which guarantees the specified overshoot only for frequencies ≥ 1 MHz. To get rid of this limitation, technologies like SiFi, EasyPulse, or Trueform are used instead of plain DDS by more sophisticated AWGs.

Some of the older AWGs created the squarewave (Heaviside step) with a comparator rather than directly from DAC. The better AWGs can be spotted which have the variable rise/fall times indicating a more complex waveform construction.

It's nice to see someone recognize Heaviside as he gets little to no credit for inventing algebraic solutions of differential equations, we commonly call Laplace Transforms today. Long ago while a adjunct prof, we tried to get a Engineering History course introduced to recognize those engineers lost in history like Heaviside...we failed tho

Best

[Filippo52]

Back to the temperatures. Since no one is working on the old firmware release, I'll do it. I keep all the releases, or almost all of them, so I installed the microSD card with the .25.

I'll repeat the similar table, made at the same time with the same ambient temperatures.

I just took this measurement, the room temperature was 26-27°C.
I monitored both temperatures at different times.

Time                             CHP_CHIP_TEMP                               CPU_AMBIENT_TEMP

0 min                                   40                                                         36

5 min                                   51                                                         44

10 min                                 53,3                                                       46

30 min                                 54,4                                                       48,1

60 min                                 55                                                         48,1

The fan speed is always  at maximum

Filippo

[Hydron]

I did test with the old FW release - see my previous post. Your temps line up very well with mine (just reported the chip temp after it had stabilised, ~56C @ ~27C ambient).

[Filippo52]

947
Sorry, Idrone
I had read, but yours seemed to be based on memory, as did my first statement. Since someone wanted something "safe," I ran the test. It confirmed what I remembered and what you wrote.
61-62°C is the target temperature with the new firmware, compared to the 56°C before. It's true that you have the advantage of a higher but more constant temperature, and since it's a measuring system, that's an undeniable advantage, but when you don't need "precision," 6°C less statistically means a significant increase in oscilloscope's life. 10°C means 50% longer life.
I often don't care much about precision, so I was waiting to talk to Martin 72 to see what he thought about interrupting the PWV wire and running the fan at full power, as it was before. This can be done with a micro switch, which you can take out, and you'd be able to choose based on your intended use.

Filippo

[frozenfrogz]

Just chipping in to the temperatur debate: Today I mounted my MHO984 on a monitor arm with VESA mount. While being very practical in terms of usability and desk space, I saw pretty high CPU temps (compared to yesterdays temps).
I need to do some A/B testing though. Today was a particularly hot day and I do not have reliable data discriminating between the mounting options yet. If there is a significant difference of cooling performance (because basically the mounting plate blocks the suction zone of the fan on the back), I will try with mounting 20-30 mm spacers and share my findings.

[eurofox]

Just chipping in to the temperatur debate: Today I mounted my MHO984 on a monitor arm with VESA mount. While being very practical in terms of usability and desk space, I saw pretty high CPU temps (compared to yesterdays temps).
I need to do some A/B testing though. Today was a particularly hot day and I do not have reliable data discriminating between the mounting options yet. If there is a significant difference of cooling performance (because basically the mounting plate blocks the suction zone of the fan on the back), I will try with mounting 20-30 mm spacers and share my findings.

You have to be careful that the vesa plate is not too close to the air entrance, I put spacers between.

[cncjerry]

last night I  ran selfcal on my 934 upgraded scope.  All probes were removed.  It ran for an hour before I left my office.  This morning, I put Leo's pulser on it and though the pulser outputs about 1.5v P2P, the 934 (actually a 984 now) displayed only the vertical lines.  As I zoomed out, the display was overloaded until I hit 10v per division and then it failed to trigger.   I'm rerunning selfcal now.  Is there anyway to reset it if this fails?  I ran the selfcal as it was reading a little low on P2P waves vs two other scopes.  (5.4 correct vs 5.2 on the 984).

Any other ideas?

thanks

Jerry

edit:

I reran the selfcal, the scope finished in about 45 minutes.  All is normal now with the triggering but the levels are still wrong.  With a 50ohm load, the 934/984 sees it as 558mv flat to flat with the cursors and my two other scopes see it as 850mv flat to flat.  In high impedance, the two scopes read 1.66v and the 984, 1.074v.  This is without a probe - the device is directly attached.  Any ideas?

Thanks again!