Frequency response of your DSO

[Mechatrommer]

If a use a constant voltage on my DSO ( Rigol 2072 ) of 0 dbM ( 0,223 V )
on the BNC connector, 1 Mohm 16 pf, i have a straight line from 70 Mhz 0 dB to 200 Mhz -3 dB.

if a use a quality signal generator, 50 ohm, 1 meter of coax, T connector on the DSO with 50 ohm
terminator with a kown specification i get a straight line form 70 Mhz 0 db to 200 Mhz -6 dB.
Thats is making the difference.

if i understand you correctly... no. you'll get 70 Mhz -3dB db to 200 Mhz -6 dB. it doesnt make any difference in the profile curve, everything are just moved 3dB down in every spectrum. and T connector with a dicky 50ohm terminator may not be as good as an inline (pass through) terminator. you may refer to the thread about "coax and bnc"... https://www.eevblog.com/forum/beginners/how-to-better-wire-the-bnc-connector-and-ghz-issue/ and read the whole story...

attenuators. but you also can find such similar enclosured terminator. i've modified one such attenuator (because its reachable from ebay, not quite cheap though compared to the T+50ohm dicky) to get a "conceptually" better terminator. i'm not sure though if an attenuator can be used (as is) in place of a terminator.

an active probe is only more usefull if you dont have such low impedance (50ohm) generator and source loading is an issue. a properly terminated coax is as good or better than active probe, that has been the standard setup (with robust generator). experts are using this method unarguably. end of story... imho.

@Marco: i'll consider a DSO as a black box i dont have to think about its internal (parallel) capacitance. everything in the front end input are considered "lumped" (black box). all i care is a signal feed into it and what it shows on the screen. during real life we dont want to be bothered with this 15pF internal capacitance issue "everytime", all we care is whats on the screen and how accurate it is compared to what is actually feed into it.

[Wim13]

Oke , here is the result of the Rigol 2072 DSO, graph is made with signal of -10 dBm ( 70 mV ) input on BNC.
Measure point every 10 Mhz, used read out of the rms value on the DSO.

The differences between channel A and B are within 0.5 db, so no use to plot.

[Mechatrommer]

Edit: I was wondering if the dip at 30MHz was my generator/cabling or the scope itself, so I tried to measure it visually with my Tek 2230 analogue scope. By eye, it looks to be about 0.5dB down at 30MHz too, so I think this might be a measurement artifact.

since Win13 profile differs from ve7xen esp at -3dB point (~180MHz vs ~100MHz)... the 30MHz notch is probably from your "not so good" coax setup? i dont know. do you have discontinuity of the shielding? have you changed the cable and/or terminator and retry the test? have you tested using 0dBm instead of the maxed 17dBm? i believe providing the image of the test setup will give a better idea if the test result is sound or not.

but i like the idea of profiling both channel, we can see the mismatch level.

[grumpydoc]

Interesting - must find time to graph out the Lecroy.

Don't forget that past 3-400Mhz the cable and BNC's will start to have a significant loss of their own, 1 metre of RG58 has a loss of 0.36dB at 400Mhz rising to  0.7dB at 1GHz and I found a datasheet for "mil spec" BNCs which gives the insertion loss as 0.1 SQRT(BW in GHz) so 2 BNC's plus 1m of coax is almost 1dB down at 1GHz

[tinhead]

right, but as everybody is anyway using something 0.5 and 1m of coax we can ignore it as "common loss".

[grumpydoc]

right, but as everybody is anyway using something 0.5 and 1m of coax we can ignore it as "common loss".

OK, if that's the aim but less good for trying to make absolute measurements - which is why high frequency levelled sine wave generators have a level sensing head to eliminate cable losses - at least things like the Tek SG504 and SG5030 do - I presume others (still?) do this.

[Jay_Diddy_B]

Hi,

I think the reason that Tektronix used levelling heads is that the inputs on the scopes have some mismatches. The VSWR of the 7A29 is specified at 1.2 : 1 That is an impedance of 42 to 60 Ohms.

By using a levelling head the effect of the mismatch is minimised.

The 50 Ohm only scopes are a lot better in this regards than the 1M / 50 ohm scopes. These switch a 50 Ohm termination in parallel with the 1M input and have high impedance attenuators.

The 50 Ohms only scopes use 50 Ohm attenuators and are lot better. There is a thread which shows the internal construction of the 7A29 attenuator.

https://www.eevblog.com/forum/reviews/tektronix-7a29-teardown/msg188496/#msg188496

The SG503 50kHz to 250 MHz calibration generator did not have a levelling head.

Jay_Diddy_B

[grumpydoc]

The VSWR of the 7A29 is specified at 1.2 : 1 That is an impedance of 42 to 60 Ohms.

By using a levelling head the effect of the mismatch is minimised.

Hadn't particularly thought about that I confess.

But I think that >500Mhz levelled sources need to eliminate the cable loss regardless- the SG504 is specified as flat to 4% (0.34dB). 1m of RG58 is about the same drop at 400Mhz and more (about 8.5%) at 1GHz so having a sensing head makes sense to keep in spec anyway.

Not sure whether the SG504 actually used RG58 to the head though.

The SG503 50kHz to 250 MHz calibration generator did not have a levelling head.

True but cable loss is less of a problem for the SG503 as it only goes up to 250MHz where 1m of RG58 has a loss of about 0.25dB

[rf-loop]

Here just OP asked displayed response.

Sweep 100s, start 1MHz stop 200MHz, linear.
Owon acquire mode is peak.
Signal generator, HP8644B flatness measured before test only directly from output N-connector so that there was only HP8482H connected directly to generator N-connector, meter used HP437B
I did not find between 1MHz to 200MHz more than +-0.03dB if set 1MHz level as 0 reference.
Real flatness not measured in test configuration. (I have no realiable tools for this).
Test system reactive load together with cable can do whatever... but for this purpose this accuracy is ok imho. Also, this is just how we many times use scope in practice.

[tinhead]

small update : this is what collected so far, look below.
still looking for other ppl/DSO data

[Mechatrommer]

at last! i get the chance to see the ds1102e (ds1052e) respond with greater resolution (still no xls yet? ) thanks tinman for the report. otoh, tekway tried to overcompensate (peaking) their dso to get higher bw.

[tinhead]

otoh, tekway tried to overcompensate (peaking) their dso to get higher bw.

actually no, the first part of frontend (up to AD8370) is not using (in my test setup) the attenuator nor compensation circuit.
The measured response is combination of :
- this part of frontend
- the 20MHz bw filter and it's influence even if off
- the LM6552 output termination
- amount of parallel connected ADCs
- parasitics
- software

When you compare to Rigol DS1000E, the first part of frontend is same, 20MHz filter similar (this is part which does have influense of the total response and partially on the bw),
LM6552 termination same, amount of ADCs diferent (which is why Rigol was not able to produce 200MHz DSO with these ADCs, this is why they used
different ADC in CA models) and software as well different (whih is for these DSOs a big difference - on Tekway i can move the -16dB point between 60MHz and 999MHz,
where of course usefull is something below nyquist frequency).

The result is flat response in compare to Rigols combination of flat and gaussian response. The "overcompensation" is typical for butterworth response, so nothing wrong here.
For gaussian DSO, such response would be however really overcompensated input.

[EV]

Rigol DS2202

Generator HP8642A connected with BNC cable and feed thru 50 ohm terminator.
Output voltage 100mV.
Attached picture and exel file

[EV]

Tektronix TDS 3032

1.
Generator HP8642A
Connected with BNC cable to 50 ohm input of the scope.
Output voltage is 100mV.

2.
Generator HP8642A terminated with 50 ohm feed thru terminator.
This terminator is connected with with Fet probe P6243 to 50 ohm input of the scope.
Output voltage is 100mV.

3.
Generator HP8642A terminated with 50 ohm feed thru terminator.
This terminator is connected with with Fet probe P6205 to 50 ohm input of the scope.
Output voltage is 100mV.

Picture attached.

[grumpydoc]

Finally had chance to graph out the Lecroy 9354TM - this is specified as 500Mhz bandwidth with the internal 50ohm terminator.

Signal generator was a Marconi 2024, 1m RG58 coax, 0dBm signal level, internal 50ohm termination. The signal generator is supposed to be flat to 1dB or so but I don't really have anything accurate enough to check much past 10kHz so I'm rather trusting it blindly.



Almost spooky that the graph shows the -3dB point as exactly 500MHz - though as previously discussed the cable & BNCs are probably contributing about half a dB by 500MHz. Interesting peak at 200Mhz - not sure why that is unless the front end has been artificially tweaked to hit the 500MHz bandwidth. Could be the Marconi, of course.

[JoeyP]

Interesting peak at 200Mhz - not sure why that is unless the front end has been artificially tweaked to hit the 500MHz bandwidth. Could be the Marconi, of course.

1m of RG58 has a propagation delay of 5nS, which happens to be the period of a 200MHz signal. The peak may indicate a mismatch at the scope input and/or Marconi output. Note there's also a dip at 100MHz. The Marconi probably has 10dB of attenuation inserted in its output path at 0dBm output level setting, which should provide a fairly accurate source impedance (unless it's damaged). I would bet the mismatch is at the scope. This wouldn't be surprising since the 50 ohm terminations in most scopes don't have great return loss specs.

[grumpydoc]

1m of RG58 has a propagation delay of 5nS, which happens to be the period of a 200MHz signal.

OK, hadn't thought of that, interesting.

Not sure whether I should be pleased that I found it or miffed that it's there. It's all within ±0.5dB though up to about 225MHz so not the end of the world.

I'd assumed that the dip at 100MHz was the start of being able to demonstrate the roll-off, if you ignore the 200MHz peak it looks like it would roll smoothly from 0.5dB down at 100 to 1dB down at about 250 and 2dB down at 350MHz.

I might map out the other inputs and also the other Lecroy that I have (a 9354T). It could also be useful to try with a 50ohm terminator on a BNC-T piece and compare with the 'scope's termination.

I confess that, although I have a sort of basic grasp of what's going on as soon as we get into the detail or the maths it goes over my head.

[JoeyP]

If you have a 10dB BNC attenuator pad, it would be very interesting to bump the Marconi up to +10dBm, insert the pad directly at the scope input, and do the same plot. If the peak doesn't go away, move the pad to the Marconi's output instead, and repeat the test. If the peak goes away in either condition, it was due to impedance mismatch/reflections. If not, it's part of the scope's response.

[grumpydoc]

If you have a 10dB BNC attenuator pad

I don't but think I might keep my eye out for one. 

The Marconi has the high (+25dBm, I think) output option BTW so the 0dBm output might be more than a 10dB attenuator.

[JoeyP]

I'd be surprised of the mismatch was at the Marconi end - particularly since its range is much higher than the scope's BW. I only included that possibility in case it happened to be damaged. If its output level is correct, then its attenuator is probably OK.

[nctnico]

AFAIK this kind of peaking is pretty normal (it is well within 1dB BTW). If they add a filter which peaks a little bit they can squeeze an extra 10% bandwidth out of the same hardware.

[Someone]

Plotted out the response and input VSWR for one of the MSO-X 3104A units we have kicking about, if I find some time I'll do one of the 200MHz units or some of the Lecroy hardware.

Measured out to 990MHz with a HP 8656B as the source, then used a N9923A out from 1000MHz and for the VSWR. All uncalibrated and I couldnt be bothered dragging out the thermal power meter to track down the reason for that 1dB drift in the result.

[grumpydoc]

If its output level is correct, then its attenuator is probably OK

I have no reason to suspect a problem with the Marconi attenuator - certainly at lower frequencies it appears to be spot on. The graph 0dB line is pretty close to 0dBm - I chose the 50kHz reading as the reference point. The 'scope reading for that was 221.8mV rms which is about -0.07dBm (unless both are off by the same amount, of course).

I do have a Marconi 2305 which I bought to adjust a 2022 which had lost it's RF level cal due to an EEPROM failure - not got round to that project yet so I've not fired up the 2305 but it was bought as working. I might use its RF level meter to check the Marconi output although I think the accuracy is only supposed to be +/- 1dB - maybe not enough to confirm that peak but it should show whether the Marconi is broadly flat in the 100MHz-500Mhz band.

[The Electrician]

Here are some measurements I made of various impedances over a 0-1GHz frequency range, and the frequency response of an Agilent DSO5034 300 MHz scope.

First, here's the Smith chart plot (0-1GHz) of the 50 ohm calibration load after calibrating the network analyzer.  You can just barely see the yellow dot right at the center of the zoomed in Smith chart:



Next, here's the impedance at one end of a 1 meter cable with the calibration load at the other end.  This is a precision 1 meter, armored, expensive reference cable.  Notice that the impedance is very nearly 50 ohms over the whole frequency range:



Here's the impedance at one end of a 53 cm piece of RG-58, average quality, calibration load at the other end, with commercially installed BNC connectors on the ends:



This VNA is a portable unit and it's possible to get the VNA port close enough to the scope input connector to make measurements of the scope input impedance without using a cable, but rather a simple BNC adapter between the VNA and the scope.  Here's the impedance of the scope input, with the input impedance selected to be 50 ohms, swept from 0 to 1 GHZ:



Here's the input impedance of the scope with its input impedance set to 1 megohm and with a 50 ohm pass through inline:



Here's the frequency response of the scope (input impedance set to 50 ohms) without using a cable between the signal generator and the scope.  I was able to get the generator output connector close enough to the scope to do the same trick I did with the VNA.  There's only a single BNC-BNC adapter between the generator and scope:

[ivan747]

Does anybody know where I can build a simple adjustable 10-20MHz sine wave oscillator to test the cheaper scopes like the Velleman HPS-10? Thanks.