Rigol DS1052E Flatness Revisited Using Peak Detector

[Mechatrommer]

I've made a little experiment to observe my Rigol DS1052E flatness using a simple Peak Detector. But result unsuccessfull. I cannot observe any attenuation happening. The result and test setup are attached below (pictures and xls report). The result is graphed in "result.jpg", but the expected result should be like the sketching in the last graph of "expected.jpg". Please give idea on how to improve this, or should i just take the result with grain and salt?

Hypothesis: DS1052E reading will attenanuate at close to its BW limit (50-100MHz)
Method: Vpeak will be assumed as the absolute/real value in reading with theoritically 0% error. By inspecting Vpeak read by DMM, and comparing with DSO reading, attenuation can be estimated from Vdrop (see calculated variable below).

Setup:
1) Hantek 3x25 Function Generator + PC to control
2) DIY Amplifier to boost 3x25 signal (lets just call it 10x25)
3) Peak Detector embedded in 10x25, 1N4148 diode + 1uF SMD capacitor
4) Cheapo ATX PSU to power 10x25
5) some BNC connectors and alligator clips
6) Rigol DS1052E Oscilloscope
7) Uni-T 71A Multimeter

Independent variable:
1) test frequency from 3x25 & 10x25, Sine signal 1Hz - 100MHz

Dependent variable:
1) Vpeak: Peak detect reading on CH2 and DMM (CH2 reading only to confirm DMM reading. DMM reading is to be trusted)
2) Vmax, Vmin: Reading from DS1052E (Vmax, Vmin) on CH1

Calculated variable:
1) Vdrop: voltage drop in peak detector compared to Vmax (Vmax - Vpeak)
2) Vpp: just another useless redundant calculation to better see the signal reading/output

Result/Conclusion:
Vdrop calculated do not demonstrate any consistent attenuation. Hence DS1052E reading does not attenuate up to the test signal (100MHz). Hypothesis proven wrong!

Related links:
Dependence of a scope's frequency response on vertical attenuator setting
Idea for Super Peak Detector

[joelby]

I'm not entirely sure why you're trying to use a peak detector, but is your function generator's output amplitude flat over its range? If you do a frequency sweep and view the output directly on your oscilloscope, do you see any change in amplitude?

It's too late for me to be bothered switching everything on to take screenshots, but I'm quite sure that my DS1102E attenuates signals at the higher frequencies when I sweep using my RF signal generator and a constant amplitude.

[Mechatrommer]

1) my signal generator and dso is not flat/constant and there is no hard prove to saying that or otherwise.
2) my only hope lies in the peak detector, to bring the balance to the force, to tell the truth, of what the signal generator is outputting. and to compare with what the dso is reading.
3) dont tell me about the termination again. its only half the Vmax, half the Vpeak, half the Vdrop half everything. i still cannot see the attenuation.
4) maybe i'm just lost in the cloud of myth here.

addition:
the idea is to convert super mad high frequency in to DC level which is easily measured accurately, with DMM. or maybe i should re-ask
1) where the dso attenuation comes from?

[alm]

2) my only hope lies in the peak detector, to bring the balance to the force, to tell the truth, of what the signal generator is outputting. and to compare with what the dso is reading.

How do you calibrate your peak detector? What's the uncertainty compared to your measured signal level? Do you change the vertical attenuator setting? The large change in amplitude looks like an issue to me. If you change the attenuator setting, you distort your measurements (the attenuator may not be linear at high frequencies). If you don't, the low vertical resolution will kill your accuracy.

the idea is to convert super mad high frequency in to DC level which is easily measured accurately, with DMM. or maybe i should re-ask
1) where the dso attenuation comes from?

There will usually be a deliberate low-pass filter in a DSO to prevent aliasing. Any amplifier will have a roll-off. Bandwidth needs to be limited to limit noise (high bandwidth = high noise). In a perfect design, all components of the system have some roll-off that combined results in the desired result of a Gaussian response with a -3dB point around 100MHz.

[Mechatrommer]

thanx for the reply alm. i thought this is going to be dead.

How do you calibrate your peak detector?

how do i calibrate my peak detector?

What's the uncertainty compared to your measured signal level?

N/A. the only data i have is the 1n4148 datasheet something like this where forward current vs forward voltage graph (If vs Vf) is the most sensible to me as reference (it says Vf around 0.4V to 1.2V on the extreme If). but thats "current" is difficult for me to measure during the test (and designing the circuit for I measurement without affecting the DUT). looking at my result, the voltage drop range from 0.7 to 1V, so i'm just assuming the current will be somewhere 4-100mA (attached graph from datasheet below)

Do you change the vertical attenuator setting? The large change in amplitude looks like an issue to me. If you change the attenuator setting, you distort your measurements (the attenuator may not be linear at high frequencies). If you don't, the low vertical resolution will kill your accuracy.

yes changed. i dont like it but i have to, since the generator will attenuate from ±10V at lo-freq down to ±2V at 100MHz. so in order to get decent range on voltage reading, i have to change vertical scale (from 5V/div to 1V/div)

There will usually be a deliberate low-pass filter in a DSO to prevent aliasing. Any amplifier will have a roll-off. Bandwidth needs to be limited to limit noise (high bandwidth = high noise). In a perfect design, all components of the system have some roll-off that combined results in the desired result of a Gaussian response with a -3dB point around 100MHz.

i was asking this because i concerned that the roll-off is due to the front end circuit, that will in turn affecting the DUT signal. so, is DC signal got attenuated too?

let me give an example. let say we are feeding dso with two signal at CH1 and CH2. and let say the dso is rolling off to -3db @ 100MHz. CH1 is 100MHz sine ±2V (true signal from generator) and CH2 is 2V steady DC. from the dso roll-off characteristic (-3db@100MHz), the CH1 will be attenuated and displayed as sine wav ±1V, but how about CH2? will the screen show a flat line at 1V too? or at 2V (true signal un-attenuated)?

but as i mentioned, i used both CH2 and a multimeter to measure volt peak (detector) at CH2 and as far as i can see, CH2 reading and the DMM is tally.

[Mechatrommer]

made another test without dso. just to see how much loading a dso gave to the source generator on previous experiment. reading of terminated signal also given, just for curious mind folks out there.
from the graph, i can say, the previous test, loading is not much for low freq, and around 30% at 100MHz (1.953 - 1.3159) / 1.953 x 100%. but the point is, even with 30% loading in the previous experiment, the peak detector is there, to monitor the signal output, loaded or not loaded. (and also pls note that this is done today, but the original test is done few days back, the calibration/signal maybe off)

since attenuation is not observed in previous test. there are several possible explanation.
1) if dso is indeed rolling of at 100MHz, it means the 1n4148 performance also rolling off at the frequency, hence the voltage drop consistency observed.
2) if 1n4148 is not rolled off, so does the dso.
3) if dso is rolled off, but not 1n4148. the graph on the first post ie "expected.jpg" should be achieved, ie Vmax and Vpeak converge (come close together), ie Vdrop approaching zero at hi-freg, which is not achieved.
4) if 1n4148 is rolled off, but not the dso, Vmax and Vpeak should diverge (farther away), ie increasing Vdrop.

so... idea?

[alm]

how do i calibrate my peak detector?

With something like a scope, spectrum analyzer or power meter with enough bandwidth so roll-off is not an issue. For example, the 1973 HP 11096A HF probe (which is basically just a germanium peak detector with a resistive divider and compensation to compensate for the non-linearity of the diode) recommends a VHF signal generator and a power meter accurate to 1% up to 480MHz. They set the signal generator terminated into 50 ohm to the desired power level with the power meter, verify that it's stable, and check if the DC voltage from the probe to be consistent with the expected voltage for that power level into 50 ohms within +/- 1dB or so. That +/- 1dB would probably not be good enough to determine a scope's frequency response.

What's the uncertainty compared to your measured signal level?

N/A.

If the difference between Vmax and Vpeak is less than the uncertainty, the results are meaningless.

the only data i have is the 1n4148 datasheet something like this where forward current vs forward voltage graph (If vs Vf) is the most sensible to me as reference (it says Vf around 0.4V to 1.2V on the extreme If). but thats "current" is difficult for me to measure during the test (and designing the circuit for I measurement without affecting the DUT). looking at my result, the voltage drop range from 0.7 to 1V, so i'm just assuming the current will be somewhere 4-100mA (attached graph from datasheet below)

This will cause error at low Vf, but uncertainty also includes for example frequency-dependence of your detector probe, variation in output amplitude of your generator, noise in your acquired signal (eg. from the generator, RFI or ADC in your scope/DMM), plus whatever I forgot .

yes changed. i dont like it but i have to, since the generator will attenuate from ±10V at lo-freq down to ±2V at 100MHz. so in order to get decent range on voltage reading, i have to change vertical scale (from 5V/div to 1V/div)

I would try to change the generator's amplitude while keeping the frequency constant, to get a feel for the linearity of your detector as a function of the amplitude. Of course this may also be frequency dependent .

i was asking this because i concerned that the roll-off is due to the front end circuit, that will in turn affecting the DUT signal. so, is DC signal got attenuated too?

Not sure what you mean. The input impedance will be reduced at high frequencies due to the capacitive loading, which will change loading by something like 25% at 100MHz with 10pF of capacitive loading and proper 50 ohm termination. This may change the behavior of your DUT.

let me give an example. let say we are feeding dso with two signal at CH1 and CH2. and let say the dso is rolling off to -3db @ 100MHz. CH1 is 100MHz sine ±2V (true signal from generator) and CH2 is 2V steady DC. from the dso roll-off characteristic (-3db@100MHz), the CH1 will be attenuated and displayed as sine wav ±1V, but how about CH2? will the screen show a flat line at 1V too? or at 2V (true signal un-attenuated)?

CH2 is independent of CH1 in this regard, so it will still show 2VDC. CH1 should show about +/- 1.4V by the way if it was attenuated by -3dB, since these are amplitude ratios, not power.

since attenuation is not observed in previous test. there are several possible explanation.
1) if dso is indeed rolling of at 100MHz, it means the 1n4148 performance also rolling off at the frequency, hence the voltage drop consistency observed.
2) if 1n4148 is not rolled off, so does the dso.
3) if dso is rolled off, but not 1n4148. the graph on the first post ie "expected.jpg" should be achieved, ie Vmax and Vpeak converge (come close together), ie Vdrop approaching zero at hi-freg, which is not achieved.
4) if 1n4148 is rolled off, but not the dso, Vmax and Vpeak should diverge (farther away), ie increasing Vdrop.

It's hard to tell in my opinion. You're basically dealing with a generator with unknown output amplitude as function of output frequency, a detector with unknown frequency response, and a scope with unknown frequency response. As with any calibration, you need some sort of known quantity for reference. The scope may not roll of by -3dB at 100MHz, it will likely have some extra bandwidth to compensate for manufacturing tolerances, changes with attenuator settings and changes with temperature. For example, for the TDS-220 I measured about -1 to -1.5dB at 100MHz at room temperature, depending on vertical settings.

[Mechatrommer]

thanx alm. i will re-study later on this.